Clouds

Best cloud computing workplace list is a sign of what’s to come

Clouds

It’s no secret that “the cloud” has arrived as a technology, but as if to underscore its prominence, Battery Ventures and Glassdoor today announced a best-of list that should help guide tech professionals looking for a good place to work.

The “50 Highest Rated Private Cloud Computing Companies to Work For” looks at privately held companies with at least 200 employees and includes known names like Asana, Dropbox, Prezi, Sendgrid, Okta, Demandbase and, yes, BloomReach. (See the full list.)

And while, OK, maybe we’re bragging a little bit, landing on the cloud-oriented list is about more than bragging rights. The growth of the cloud sector — and cloud companies providing all sorts of services, operational software, security tools and more — has been astounding in recent years.

But that’s nothing compared to where it is going, as a broad spectrum of companies shed legacy onsite software systems that require those companies to manage upkeep, upgrades and periodic replacements.

Battery General Partner Neeraj Agrawal, who specializes in cloud investing, estimated that enterprises are currently spending about 15 to 20 percent of their software budgets on cloud applications. Within the next decade, he said, that number would likely be closer to 75 percent.

“From my perspective, and I’ve been investing for 16 years, cloud computing is the single largest tech trend out there,” said Agrawal, who is also on the board of Glassdoor, a job and recruitment site that provides employee reviews of employers.

(Battery Ventures is an investor in BloomReach.)

Cloud company acquisitions are heating up

Not only is spending on cloud services on a steep climb, he said, but acquisitions of cloud companies are increasing dramatically. Agrawal said that in 2016 there have already been $50 billion worth of acquisitions involving cloud players.

Yes, some big deals factor into that 2016 figure (Microsoft buying LinkedIn), but the buying activity is still impressive if you look at the $15 billion in annual average activity over the past four years.

There are several reasons for the furious growth, Agrawal said, none of which are going away soon. Hosting software in the cloud allows for rapid iteration and continuous improvement that cloud-company customers benefit from immediately, rather than enduring the unwieldy process of upgrading traditional software products housed on a company’s own servers, for instance.

Buying software as a service gives customers maximum flexibility and requires software vendors to be not only looking for the next sale, but looking constantly to keep existing customers happy so they will renew their contracts.

So what does all this have to do with whether a cloud company is a good place to work? You know the answer, don’t you? In order to be a successful, a company needs to hire the best people. And it needs to keep them.

Scott Dobroski, a Glassdoor community expert, said the company has conducted surveys that show that technology professionals know they can command good pay and land a job at a company that appears to have good prospects. So, the number one thing that keeps tech professionals around and satisfied in their jobs, Glassdoor’s data shows, is an attractive company culture.

“Where the cloud companies and other standard startups really have to try to differentiate themselves, is their corporate culture,” Dobroski said.

No doubt, competition for the best talent is already brutal in Silicon Valley and the technology sector. As cloud computing grows dramatically, the competition for talent will grow with it.

“There is a war for talent,”  Agrawal said, “whether it’s engineering or sales people or marketing or other functions. Ultimately, these individuals have many choices of companies to go to and picking one where employees are rating a company highly is something they think about.”

Now, no one is saying a job-hunter is going to make a life-changing decision based on a best-of list or even a Glassdoor rating. But Sondra Norris, BloomReach’s head of people, said that when job candidates are considering several offers, they are naturally going to find out all that they can about how those already at the companies feel about their workplaces.

“All things being equal, I think it comes down to how you felt when you went through the interview,” she said. “And then you’re going to check with people to find out what they know about what it’s like to work here and be here.”

It only helps, being on a list of workplaces well regarded by those who work where your preferred candidates are thinking of working.

Agrawal agreed that employee reviews matter.

The internet is involved in our decisions, big and small

“There is another kind of strong secular trend that is happening and that’s really toward transparency on the internet,” he said. “Whether it’s a rating review for an e-commerce purchase or a hotel review on TripAdvisor or a restaurant review on Yelp. Every considered decision we have, there is going to be content on the web to help you make that decision.”

And, Agrawal noted, prospective employees are not the only ones who turn to rankings, ratings and reviews as part of their due diligence.

“The other belief we have as investors, it’s pretty simple: Happy employees end up resulting in happy customers, which end up resulting in happy investors,” he said. “Before I make an investment, I read all the reviews. I can go to Glassdoor to see what it’s really like to work there. To me, it’s a really good signal of what’s happening.”

Battery and Glassdoor compiled the 50 highest-rated workplace list using Glassdoor ratings. The list is a rundown of the private 200-plus-employee cloud companies with the highest ratings on the site. There were a few other criteria: Glassdoor defined “cloud company” fairly broadly, though all the “few hundred” companies considered eligible for inclusion conduct cloud-based business. The companies were U.S. based and focused on business-to-business sales. A company needed to have at least 30 Glassdoor reviews for it’s rating to be considered.

And so, how did the top 50 stand apart from all the others? BloomReach’s Norris, who obviously hadn’t evaluated the entire field, said she had some thoughts about what separated them from the rest.

“We are trying to create, and have created, a place that is doing something enduring, something meaningful, that also has a sense of community and camaraderie,” she said. “It’s a place where people actually want to be at work.”

Given the trajectory of the cloud sector, building a place where people want to be is only going to become more crucial as time goes on.

Cloud photo by theaucitron published under Creative Commons license.

Mike Cassidy is BloomReach’s storyteller. Contact him at mike.cassidy@bloomreach.com; follow him on Twitter at @mikecassidy.

 

IMG_8428

Startup founders aren’t crazy — not exactly, anyway

In good times, launching a startup is a leap of faith. In times like this, with investors turning stingy and skeptical, it’s more like a leap of insanity.

And yet they come — the visionaries, the dreamers, the naive, the once-bitten, but hardly shy veterans of the startups gone down in flames. More than 600 of them have come to the Collision tech conference in New Orleans this week in the form of startups that pack the cavernous exhibition hall.

Countless more are here in the form of attendees hoping to pick up advice, partners, employees, customers, an edge, investors’ cash or at least the promise of a meeting. No, it doesn’t make sense on its face, but innovation and entrepreneurship doesn’t come from things that make sense on their faces.

“The mistake is thinking of a startup as a job,” Silicon Valley serial entrepreneur Steve Blank told me at Collision on Thursday. “It’s the world’s (worst) job. It’s the world’s best calling. If you’re not called to do this, don’t do it, because the odds of you succeeding are extremely low.”

Megan Quinn and Peter Pham

Some would say it takes drive. Megan Quinn, a growth-focused general partner at Spark Capital, said “obsession” is a more accurate way to put it. Quinn along with Peter Pham, a startup refugee and the co-founder of the Santa Monica, Calif. incubator Science, presented from the popular Startup University stage at the conference.

“In the early days of a startup, it is about obsessively getting off the ground,” she said. “You have a stopwatch on you. You need to build and you need to talk to customers. And like I said, you have to be obsessive.”

More Collision coverage

Obsessive. And there they were Thursday morning: standing-room only, listening to just what a miserable life launching a startup is. Being a founder means hearing “no,” again and again. It means having family and friends doubt your sanity. It means wanting desperately to give up and then deciding to go on and then wishing you’d given up.

Pham, a co-founder of the confounding and closed down Silicon Valley startup Color, has kept score of the reaction Science’s portfolio companies get when they pitch investors. Dollar Shave Club, which is now has legacy razor company Gillette rattled, was turned down for funding 70 times, he said, before it got off the ground.

“I’ve had 3,000 ‘nos’ in four years,” he said. “And so many entrepreneurs will just break down and quit. It was soul-crushing.”

And yet, as the session went on, more came to the Startup stage, filling more of the standing room. Some came knowing you can’t take “no” for an answer. Some no doubt knew what they were in for. Some might have been a little dazzled by the often-told story of startup success — the “seven-year overnight” success stories,” Pham called them.

“The problem nowadays is that this has become so cool,” Blank says of starting a company, “that everybody thinks they are going to be (Mark) Zuckerberg.”

But very few are going to be Mark Zuckerberg. In fact, almost none, or maybe, in fact, none, are going to be Mark Zuckerberg. And yet they came, obsessed.

“If you’re doing it for the adventure, for the calling, for the passion, because you have the vision that you’re desperate to turn into reality, this is a hell of a career, because there is immense psychic satisfaction and it comes with a non-zero percentage that you could hit the lottery,” said Blank, who Collision described as “an eight-time-serial-entrepreneur-turned educator,” a reference to his experience founding or working in eight companies and his role now teaching entrepreneurship at Stanford University.

It’s the same reason dancers dance and painters paint, he said, just before presenting a Collision session called “How to Build a Lean Startup.”

“It’s like being an artist,” he said. “You’re driven. You’re called to be an artist or a musician or whatever. Yeah, you might get paid, but very few true artists or painters or composers or dancers” do it for the money.

They do it because they can’t not do it. And the undeterred just kept coming to the Startup stage to hear Quinn and Pham talk about charting the startup journey. They came with questions: How do you build a team, find a co-founder, inject a new idea into an existing market, find work/balance?  Insert the sound of a needle scratching across an old vinyl LP here.

How do you find work/life balance? You don’t.

“You will you will gain weight and you will work obsessively,” said Pham, who also worked at Photobucket and BillShrink. “You will sneak that email at dinner. You will stay up until 2 a.m., working, while your significant other is sleeping.”

Reggie LaRoche

Reggie LaRoche, who wasn’t in the standing room overflow, but sitting in the front row, didn’t come to hear that. LaRoche is cooking up a travel industry play that would help consumers sift through the Byzantine pricing of hotels and air flights and he’s looking for thoughts about the best platform on which to deliver his model.

“For me, it’s all about the solution,” said LaRoche, 41, who’s led tours around the world for the nation’s big travel and tourism companies. Now he wants to build up his Signature Travel Club into a full-fledged business. “If you have identified a problem and there is a solution, where people are spending a lot of money unnecessarily because they don’t have the knowledge and the content that’s out there, there is always going to be an opportunity as an entrepreneur to make money.”  

Is he obsessed? LaRoche goes with “passionate.”“I’m actually passionate about solving an issue,” he said. “I want the issue to be solved.”

And let’s face it, “obsessed,” “passionate,” it’s all kind of splitting hairs. And neither LaRoche or those packed into the Startup University Stage area had time for splitting hairs.

After all, they’ve got companies to start.

Photo of Michael Gasiorek of Startup Grind interviewing Peter Pham and Megan Quinn and photo of Reggie LaRoche by Mike Cassidy.

Mike Cassidy is BloomReach’s storyteller. Contact him at mike.cassidy@bloomreach.com; follow him on Twitter at @mikecassidy.

 

Dress-Card_Color

Natural language processing for synonyms: understanding creativity

We all like being creative. Unique. Having our own opinion. Being distinct. We react differently to similar events. We see the world differently: It’s a scientific fact. Being unique is such a human thing that it’s deeply ingrained in natural human languages, which have developed many ways to talk about the same things. In English alone, there are 534 ways to say that something is beautiful. And an astounding 1,261 ways to say that something is strong.

Do you remember the BloomReach “Describe the Dress” quiz from last year? Hundreds of you, the readers of this blog, who took the test, used 148 unique ways to describe the color of the dress, 212 unique ways to describe its material and 216 ways to describe where one would wear it!

Dress-Card_Color

Now, imagine you just got this brand-new item for your online apparel store. What product description would you choose to make sure your customers were able to find this dress on your website? Statistics shows that findability can make or break online businesses. “We found a direct relationship between the user ability to find what they want on the website within the two attempts and the likelihood of them staying on the website,” says Christian Holst of Baymard Institute, a highly regarded web research group.

With so many ways to describe one and the same thing, is it at all possible to match online shoppers’ unique and creative search queries to the products your e-commerce company sells? According to last year’s e-commerce site search usability report by Baymard Institute, only a few e-commerce websites can do that successfully. The rest, 70 percent of websites, require users to search by the exact product name the website uses. For example, at a site selling “wool blankets,” typing “blanket” into the search box would get you what you need. Typing “bedspread” or “quilt” would not.

One might argue that “blanket” and “bedspread,” or “mountain bike” and “downhill bike,” or “laptop case” and “laptop sleeve” in fact describe slightly different products. But these products are so highly related that the majority of online shoppers typically don’t care. In the world of e-commerce, these phrases are synonyms, or phrases that can be used to describe similar or related products.

It looks bad if your site carries “espresso machines,” but when an online visitor types in “coffee maker,” he or she gets a “no results” page. It’s even worse if the online shopper mistypes the brand of your best-selling coffee machine as “Keurick” instead of “Keurig” and gets the “no results” page again. Unfortunately, 18 percent of e-commerce websites don’t provide useful results if their visitors mistype just one letter in the product name.

There are many more pitfalls on the road to seamless understanding of the language of your customer. Take grammatical forms like “top” and “tops.” For us, it’s obvious that these words mean the same thing, just in different quantity. It’s not that obvious for a site search algorithm. Or consider spelling variations. Is it “fitbit,” “fit bit” or “fit-bit”? Or abbreviations. Is “SF Giants” gear the same thing as “San Francisco Giants” gear? The site search has to know how to deal with all these things in order for online shoppers to have a pleasant online shopping experience.

From a technical viewpoint, there are several approaches to dealing with synonyms. Traditional ones include stemming, using dictionaries and manual editing. More innovative approaches are based on machine learning and statistical natural language processing techniques.

Stemming helps solve cases like “top” vs. “tops.” The stemming algorithm identifies the “stem,” or root, of grammatical forms of the same word (“top-” for “top” and “tops”) by cutting off endings from the right end of the word. The problem with stemming is that it’s not flexible enough and doesn’t understand the meaning of the words. For this reason, it will mistakenly treat “legs” and “leggings” as having the same stem “leg-” and, therefore, being the same word.

Traditional dictionaries like Oxford English Dictionary, available online, provide a good starting point for dealing with common spelling variations, abbreviations and synonyms like “kids” vs. “children.” However, language models based on dictionaries are incomplete and don’t reflect how people actually use the language in search queries. Dictionaries might not be useful for emerging synonym pairs like “fitbit” vs. “fit bit” or “sleeve” vs. “case” in the context of laptop accessories.

4041872282_287bea50a1_z (1)

Manual editing is the easiest way out, but requires a lot of time and effort. The procedure is self-explanatory. It allows site search managers to manually create matching rules for pairs of words that should be treated as synonyms. Based on BloomReach’s analysis of major site search providers, this is the most common approach to dealing with conceptually related phrases like “running shoes” and “sneakers,” or “hoodies” and “sweatshirts.” But just imagine doing that by hand for thousands of products! And then re-doing it once search trends change, as they often do.

The cutting-edge approach to synonyms is based on analyzing volumes of data from multiple sources, including site search queries, retailers’ product descriptions, as well as Web-wide collections of text data. Using the variety of data, any phrase can be represented in a mathematical form (for example, as a number of contexts in which it occurs). After the text data was converted into this structured quantitative form, machine learning algorithms are applied to test pairs of phrases for relatedness and calculate their “similarity vectors.” The higher the similarity score, the more confident we can be that the two phrases mean the same or related things. This approach helps identify synonyms of any type (spelling variations, abbreviations, related concepts), and does this automatically, with speed and scale. Moreover, this approach automatically detects any changes in the way online shoppers use the language. For example, it will detect that people started increasingly using “iPad” to search for any tablet. Or that “frozen toys” now mean the same as “toys from the ”Frozen” movie,” and not “toys that were put into the freezer.”

Emerging machine learning and natural language processing technologies are amazingly liberating. If you are an online retailer, these technologies make sure your products get found online. If you are an online shopper, the same technologies allow you to easily navigate the sea of online information and to be able to find what you are looking for, even if you describe it in your own, only known to you, way. Technology allows us to remain creative in the way we express ourselves while shopping online, and to be unique, which, after all, means being human.
The photo “Sculpture: OMG LOL by Michael Mandiberg / Eyebeam Art + Technology Center Open Studios: Fall 2009 / 20091023.10D.55420.P1.L1. / SML” by See-ming Lee published under Creative Commons License.

13798927135_200e425260_z (1)

Natural language processing for site search: make your site speak the language of your customers

Site search is rarely on the minds of digital marketers as a white-hot trend that is key to customer satisfaction, increased sales and their companies’ survival. Neither is it widely discussed in reports on the future of e-commerce. However, it’s impossible to ignore the statistics.

Poor site search capabilities and navigation are among the top 12 reasons e-commerce sites could lose customers. A 2014  survey of UK consumers by Rankspace discovered that 34 percent of online shoppers give up browsing a website within 10 minutes if they can’t find what they want. Finally, a survey of B-to-B e-commerce professionals found that 45 percent of respondents view “advanced site search/navigation” as a key capability that drives satisfaction of online buyers.

The importance of search for online shopping can’t be overstated. There is this existential challenge: on the one hand, we, consumers, are all unique individuals describing the things we want in a million unique ways. Literally a million. Google reports that every day there are at least 500 million queries that have never been seen by the search engine before. Of course, not all of them are related to shopping. Yet the share of search keywords integrated with Google Shopping already constitutes 16% and keeps growing.

On the other hand, in retailers’ product catalogs each product has one and only one name. Those names can be as quirky as IKEA’s “Knislinges” and “Ullgumps” or as long as “Vintage 1950’s Floral Spring Garden Party Picnic Dress,” but there is just one for each product and that’s it. Do consumers need to know the exact name to be able to find the product they are looking for? What if they don’t? Bridging this naming gap, bringing these two sides – consumers and products – together is not an easy task.

A remedy for search engine fatigue

Online search still remains very much keyword-based. That means that whenever you type in the words that describe what you are looking for, the retrieved results will contain the exact search terms you used. With keyword-based search, if you type in “copy machine,” you will see pages having the words “copy” and “machine” on them, but not a “multipurpose printer.” Although a multipurpose printer might be the thing you were actually looking for.

From a language viewpoint, keyword-based search recognizes the form of the words, not their meaning. It may understand that “machine” and “machines” are actually the same word in two different grammatical forms. But since it doesn’t understand the meaning of the words, it can’t logically pair groups of words related to the exact search query used. Users will have to come up with related searches (like “printers,” “copiers” or “xerox machines”) on their own.

In the early years of Internet search, the inability of search engines to understand what people were actually looking for gave rise to a so-called “search engine fatigue”: 72.3 percent of Americans experienced it when researching a topic on the Internet.

A remedy for search engine fatigue came in the form of semantic search. In contrast to keyword-based search, the goal of semantic search is to understand the intent behind a user’s search query and find information based not just on the presence of the words, but also on their meaning.

9044745765_b49f107f65_z (1)Most leading search engines now use some elements of semantic search in their algorithm. The most dramatic shift towards semantic search happened when Google announced the Hummingbird algorithm in 2013, a major revamp of their search engine. With Hummingbird, search users can now ask Google questions in natural language and receive relevant answers in return. For a query “Who’s the President of the United States?” Google now returns a card about Barack Obama and links to pages mentioning Barack Obama. A keyword-based search would return pages ranking high on the keywords “President” and “United States”, which may or may not mention the current U.S. president.

Semantic search for e-commerce: when “dress” is not always a dress

Consumers are increasingly looking for Google-like search experience everywhere on the Internet and on e-commerce sites. With semantic search becoming a new standard, they want e-commerce site search to know what they want and be able to speak their language.

The main element of semantic search is the ability to understand natural language queries. In an e-commerce setting, it also means being able to accurately identify products and their attributes. When online shoppers search for a “dress shirt” chances are they don’t want to see women’s “shirt dresses” in the search results. In this query, “shirt” is an actual product and “dress” is its attribute. Similarly, they may not want to see casual shirts, because they typed in “dress” for a reason, and a results page with any kind of shirts wouldn’t be relevant for them.

Screen Shot 2015-08-04 at 10.58.50 AM

Search engines that understand natural language queries are better at handling long-tail search queries. They seamlessly parse the queries, usually representing phrases and even full sentences, find their meaning and are capable of directing search users to the exact product they are looking for. Common sense suggests that a consumer looking for “black leather Jessica Simpson high heels size 7,” is more likely to convert than those looking for “women shoes.”

Another critical element of semantic search is the ability to understand synonyms. If an e-commerce website doesn’t match “blow dryers” to “hair dryers,” “hoodies” to “sweatshirts,” “downhill bikes” to “mountain bikes,” frustrated online shoppers may decide to go find what they need somewhere else. The result of such a poor consumer experience, of course, is lost Web sales. A 2014 benchmark study of site search usability by Baymard Institute found that an alarming 70% of websites use site search that doesn’t support synonyms.

Understanding concepts is the third major feature of semantic search. In our world, things can be related to each other in many different ways. Often, when people don’t know what solution to look for, they search for products by problem or symptom. “Symptom searches”, highly applicable to e-commerce sectors like drugstores, health & beauty, cleaning, tools and supplies, are one of the 12 most popular types of search queries identified by Baymard. As Baymard report demonstrates, websites that don’t support symptom searches return any products with “dry” in their description, like treatments for dry skin, for a query “dry cough,” which is not even close to what the search user had in mind.

13798927135_200e425260_z (1)

It’s in the best interest of e-commerce websites to implement semantic site search. According to Total Retail Report, websites with keyword-based search report 40 percent abandonment rate. For websites with semantic search however this number is as low as 2 percent.

Natural language is a very nuanced and ambiguous tool. We, humans, use it masterfully when we need to express what we need, and, probably with a slightly less precision, when we try to understand what other people want. Until recently it was unthinkable for machines running search engines to be able to understand natural language and the intent behind it. With the latest advances in technology, it has become more real than ever to create a site search system that speaks the language of your customers.

Site search is still an area in need of considerable improvement, even for the nation’s leading e-commerce sites. The good news? There is a huge opportunity to make online shoppers happier. E-commerce sites need to make sure they don’t miss it.

Photo of Superman by Steven Guzzardi, photo of The Hellish Diva – Lovely Shirt Dress (LIITA Hunt) by Tigist Sapphire and photo of Best of British vintage retro products by Paul Townsend published under Creative Commons License.

405077721_91411d162d_z

Know your customer: Natural language processing for e-commerce and retail industries

This is the second post in the series on natural language processing (NLP) focusing on solutions for the e-commerce and retail sectors.

You know the drill: You’re looking at a full Saturday of shopping at Williams-Sonoma to update your kitchenware. Naturally, you turn to your gourmet friend for recommendations and she delivers. But can we also turn to a machine in the hope of finding similar help?

As futuristic as it might sound, that is exactly the grand vision behind an app, Natural Selection, which relies on IBM Watson’s cognitive computing technology. The app allows online shoppers to ask questions using natural language and returns a set of personalized offers. The creators of the app hope that it will help e-commerce retailers better understand the intent and shopping preferences of their customers.

Understanding consumers has always been high on retailers’ agenda. The rise of the Internet and social media drastically changed the way customers shop and interact with companies and brands. The sheer scale and intensity of customer communication can be overwhelming. A much-cited example is Wal-Mart, which is estimated to collect more than 2.5 petabytes of data every hour from its customer interactions – that’s about 20 million filing cabinets’ worth of text, every hour.

Given the customer-facing nature of retail business, it’s not surprising that, as an industry, it contributes nearly one-third of the growth of the text analytics market. E-commerce companies enjoy a large base of customers who increasingly express their needs, attitudes, preferences and frustrations online. Social media listening has become an important tool for e-retailers who want to understand consumer shopping habits, predict product demand or monitor trends to create sticky marketing messages.

Consider the analysis of Father’s Day tweets that uncovered gifts that kids were considering for their dads. That could drive creation of real-time marketing campaigns. Or how about the recent study that found that Thanksgiving preparation involves a lot of stress related to awkward conversations or interactions with family members. Marketers who mention holiday stress relief in their messages could resonate well with their customers during the lead up to Thanksgiving. The analysis also found that people talk a great deal about being hungover on Black Friday. As pointed out in the study, pharmaceutical brands like Advil or Tylenol could leverage this trend mentioning “hangover remedies” in their real-time marketing campaigns on that day.

Customer experience management is another big application of natural language processing for the retail industry. After Nordstrom started analyzing volumes of customer feedback through comment forms, surveys and thank-you cards, they found that many of their in-store customers had trouble locating Nordstrom salespeople who wore street clothes and nametags rather than uniforms. In response, Nordstrom dressed their associates in branded, brightly colored T-shirts, so customers could easily spot them. Within two days of the pilot, the company saw a 30-point jump in the key metric they use to evaluate the effectiveness of their sales staff.

So how do these big companies spot these trends, technically speaking? To analyze recurring themes raised by customers on social media, in free-text comment forms, product reviews, emails, etc., machines should be able to categorize all this text data into business-relevant topics (also called “text categorization”). One way to do it would be to manually create rules describing when a certain phrase or text can be assigned a specific topic. For example, we could create a rule that whenever customers mention “price,” “cheap,” “expensive,” or “sale,” their comments should fall under the “Price” category. Alternatively, two machine-learning techniques, clustering and classification, can be used to automate the process. Clustering groups similar data points together. Classification automatically categorizes data points using pre-defined categories. All three methods have their advantages and drawbacks, so they are typically used in combination in a business environment.

A technique used to monitor customer attitudes, opinions and feelings towards what’s being discussed is called sentiment analysis. If topic extraction demonstrates what people are talking about, sentiment analysis uncovers positive and negative drivers of those conversations.

405077721_91411d162d_z

Listening to the voice of the customer and managing customer experience is essential for e-commerce and retail companies. But how can an organization claim they know their customers if they don’t even understand what consumers want when they shop on their websites? Since we mentioned Nordstrom earlier, let’s take them as an example. Generally their site search results are nice and relevant. But not for a query “foot cream.” As of today, none of the top results for this query is an actual foot cream. Instead, a puzzled online customer sees several types of Barefoot Dreams blankets and an eye cream.

Foot cream is not the hardest example. People use language creatively and can describe one and the same product in many different ways, like “red lipstick” versus “lipstick for a date night.” Interpreting search queries phrased flexibly in natural language rather than using rigid keywords is a formidable task for machines. However, the better they do it, the better e-commerce and retail companies can serve online shoppers, converting them into satisfied and loyal customers. Our analysis at BloomReach, which looked across a number of merchants at million of purchases, found that visitors using site search account for up to 45 percent of online sales revenue.

Improving search capabilities is at the top of the list for business users of text analytics solutions. Twenty-nine percent of respondents to analyst Seth Grimes’ survey on the state of text analytics in 2014 named “search, information access and question-answering” a top business application of text analytics/ language processing solutions.

Natural language processing becomes all the more important in the booming field of artificial intelligence (AI) — a field of computer science which studies how to create computers capable of intelligent behavior. Apps that can answer shoppers’ questions in natural language are an early example of this. With further advances, AI machines are expected to increasingly take care of online shopping on behalf of their owners. Gartner predicts that by the end of 2016, mobile digital assistants will be conducting $2 billion in online shopping. At first, digital assistants will be allowed to automatically fill in address and credit card information. As these applications earn the trust of their owners, they will gradually be taking on routine, repeatable tasks like buying paper towels every three weeks or getting new water filters. If this trend continues, in the  foreseeable future, a substantial percentage of visitors to online websites might be machines, and not humans.

E-commerce companies can respond to this trend by doing what some have already been doing quite successfully: proactively taking on the role of digital assistants or virtual concierges for their customers. This requires the ability to recognize online shoppers, understand what they are looking for and know their habits and preferences. In other words, it’s time for e-commerce enterprises to communicate with customers as if on every visit there were an attentive and helpful salesperson on the other side of the screen.

Photo of “Fridge Words” by John Fife published under Creative Commons license.

3231178720_5e2c1c45a8_z

What marketers must know about natural language processing

This is the first post of a series on language technologies. Why do marketers need to know about it? Because their business relies on how well they understand their customers. In today’s digital and mobile world this means listening to what customers say on social media, in free-form messages, even in search queries. Businesses are drowning in text information. Here’s where language processing solutions come in handy, extracting actionable signal from noise.

In the spirit of the upcoming Father’s Day, did you know that daughters send more Father’s Day tweets to their dads than sons do? And that apparel, together with special meals, alcohol and barbecue tools and meat are the top four gift ideas mentioned in the tweets? Uncovering insights like these, which are invaluable for any retailer counting on a big Father’s Day, would not be possible without Natural Language Processing (NLP).

Natural language processing, also called “text analytics”, is the technology allowing machines to understand what people write or say conversationally at a scale and speed that greatly exceeds the abilities of human experts. In their Father’s Day Twitter analysis, a text analytics company, Luminoso, looked at 92,450 tweets. It would take an average person, working eight hours a day, about 21 days, or an entire work month, to read through all those tweets. And we haven’t even gotten to the work of generating insights.

Big Data became a buzzword in the business world a long time ago. Since then, organizations have started learning how to draw actionable insights from data that’s quantitative and structured. Now the time has come for Big Text.

Big Text is this enormous universe of documents, e-mails, free text forms, social media posts, product reviews, call center logs, you name it, generated in natural language. The documents are inherently vague, ambiguous and context-dependent, because that’s what human communication is. This is unstructured data that is not presented in any kind of pre-defined data model readily understood by machines. According to a new IDC study, unstructured data composes up to 90 percent of all digital information. Can you say that you read a book if you only understood 10 percent of it? Can you say you know how your business operates, if you can only measure 10 percent of it? Organizations that are not leveraging unstructured data risk losing business to more data-focused competitors. And, again, most of unstructured data comes in the form of natural human communication.

Twitter

 

How hard is it for a machine to make sense out of a phrase in natural language? Imagine you are selling dresses and your potential customer types “fabulous cocktail dress for a wedding” into Twitter, Facebook, her blog or your site search field. We, humans, understand what this person is looking for because we know that “dress” is a piece of clothing and “cocktail” is its attribute, and not a drink, and “wedding” is a formal occasion, which usually has guests, guests who are expected to dress up. How do we teach machines to understand this query so it can facilitate discovery?

When people think about machines understanding human communication the first thing that usually comes to mind are virtual personal assistants like Siri or smart question-answering systems like Jeopardy!-winning IBM Watson. However natural language processing is much more pervasive. Say, when a customer routinely searches for products online, language technologies are quietly working in the background to generate relevant search results. To understand a search query machines do “query parsing”, breaking up the query into words and understanding how they relate to each other using statistical and machine learning heuristics.

For instance, to understand whether a product called “cocktail dress” would be relevant for a customer shopping online for a bridesmaid dress, the machine may calculate the similarity of these two phrases by looking at all the contexts in which both phrases are used. This approach is called “distributional similarity”: the higher the number of similar contexts, the higher the probability that the two phrases will have the same meaning.

There are some great natural language processing service providers in the market who can deal with all sorts of tasks related to text analytics. But for businesses, having an available technology is just part of the solution. The largest challenge lies in the organization’s need to identify use cases and answer the question of how they could benefit from language technologies.

The market understands this and tries to sell business solutions instead of pure language processing technologies. Seth Grimes, a renowned expert in this field, observes: “Text-analytics technology is increasingly delivered embedded in applications and solutions, for customer experience, market research, investigative analysis, social listening, and many, many other business needs. These solutions do not bear the text-analytics label”.

Voice of the customer analytics/customer experience management remains the biggest driver for adoption of language processing solutions. Alongside with analyzing Father’s Day tweets, it may bring businesses much more tangible results. With the help of text analytics, a leading hardware technology company managed to reduce the number of social media messages a specialist at its social media command center has to go through daily from 5,000 to 450 (the rest being replied to automatically or filtered out as spam). That’s an over 90 percent workload reduction!

It’s no surprise that Seth Grimes predicts that the first text analytics unicorns (startups with $1 billion or higher valuation) will be in social media analytics or customer experience management space.

More organizations are expected to benefit from natural language processing solutions on a larger scale in the years to come. According to a recent market report, the natural language processing market is expected to grow at an annual rate of 18.4% and be worth $13.4 billion by 2020. Given the speed with which the volume of unstructured data increases every second, it’s in the best interest of organizations to be able to convert this data from waste into asset. Whenever businesses have to take action based on vague, ambiguous, ever-changing human communication, natural language processing technology will have to be part of the solution. And all the better, if in the process, Father’s Day becomes a little happier for dads and retailers alike.

Photo of Twitter by Jeff Turner published under Creative Commons license.

 

IMG_5524

Collision conclusion: Big data for the masses is near

Sometimes at big, chaotic tech conference it’s best to look for inspiration in unlikely places — like the title of one of the dozens of sessions meant to draw from among the mob scurrying from demo to meeting to to talk to funding pitch.

And so it was at the Collision conference in Las Vegas this week, which offered a session called, “It’s the Data, Stupid.” No, it’s hardly an epiphany that data is a big deal. But that wasn’t the point of the session nor of the many conversations I had with attendees about the state of data today.

IMG_5524

What’s clear is that we are nearing a tipping point with big data, a new era in which the tools, technology and practices exist to democratize data. More and more frequently, the insights that analyzed data provide can be used by teams throughout an organization to make immediate decisions that help move a business forward.

Listen to Amanda Kahlow, founder and CEO of 6Sense, talk about the new ways businesses view the potential of data.

“Five years ago, people were investing in the back end technology,” Kahlow, whose company provides predictive analytics for marketers, said during a Collision panel that I moderated Wednesday. At the time, technologies like Hadoop were all the rage and received all the buzz, she said.

“Today,” Kahlow continued, “it’s all about the applications. How do we use this data to drive something forward? And how do we use it got get measurable results?”

Yes, in Jeopardy-like fashion, she phrased her answer in the form of a question. And it’s true that the democratization of data has a ways to go. The vision of just the right insight being available to just the right person at just the right time is a work in progress. But plenty of tools are out there and they are being put to good and profitable use.

Take the example of marketers, as Richard Frankel, president of Rocket Fuel, a programatic media-buying company, did during the aforementioned “It’s the Data, Stupid” session.

“For marketers, the job is changing,” he said. “It used to be a ton of people guessing and doing research: of the torrent of data, which tiny bits could they use? At human scale, that was all the data they would spend time with. Now, with all the technology that’s come along, all the assumptions of how much data could be used and how fast it could be used have gone out the window.”

Machine learning, natural language processing, the ability to wrangle unstructured data and render it meaningful has changed everything.

It’s as if data is moving through an evolution that is familiar in the world of technology. When the first personal computers were developed. Visionaries were wowed. A complete computer that you could fit on a desk! For others, the question was: But what can you do with it? Then came spreadsheets and word processors, which helped democratize the PC. And then the Internet, which broke everything wide open.

16315439180_7a4da20bfa_z

 

Hundreds of companies big and small are scrambling to be the word processors and spreadsheets of the big data era. The choice and different approaches can be daunting for those who are leading companies struggling with their own data needs. But few, if any, doubt that it’s time to do something to make data more valuable for workers throughout an enterprise.

The pace of competition won’t allow for the old ways, in which a mid-level manager comes up with a question he or she needs answered. Then it’s: Send the question to the analytics team and wait for an answer. By the time an answer comes back, the world has changed.

There is no question that we’re in an inflection point,” says Stuart Frankel, CEO of Narrative Science, a company that uses artificial intelligence to generate narrative summaries of data. “Where we started with the data, it was about data capture. That’s really what the data has been about — monitoring, tagging etc. There was almost this blind lust for data.”

The thinking, says Frankel, who appeared on the panel with Kahlow, was that if we just gathered as much data as we could, we’d figure out what to do with it, eventually.

“It turns out the collecting of the data and the aggregating of data is ultimately one step and the first step in what I think will be many” on the way to where the focus on data pays off, “because we can make better decisions.”

To look at the PC evolution analogy, it may be that we’re still in the early spreadsheet era when it comes to big data.

“We are just hitting the cusp of what is about to hit us and about what data can do,” Kahlow said. “Data is going to change everything that we do. And  every industry and every vertical is going to be disrupted. Today is a different world. I’m really excited to see what’s to come.”

And rest assured that it is coming.

Photo of Richard Frankel and panel by Mike Cassidy. Photo of IBM PC by Paul Sullivan published under Creative Commons license.

Mike Cassidy is BloomReach’s storyteller. Contact him at mike.cassidy@bloomreach.com; follow him on Twitter at @mikecassidy.

 

Collision

Socks and underwear? Disruption at Collision knows no bounds

When you think of disruption in 2015, I’m betting the first things that come to mind are not socks and underwear.

Collision

But socks and underwear disrupters are here at the Collision tech conference in Las Vegas, a gathering of entrepreneurs, tech visionaries and the money people who fund their visions. Besides the brains behind Adore Me, an e-commerce lingerie play boasting a huge inventory and a subscription model, and Sensoria socks, which uses data to provide personalized coaching for runners, there are data scientists, app makers, artificial intelligence experts, business minds and — not incidentally — big companies that buy little companies and venture capitalists.

While the fledgling conference, which grew out of Europe’s Web Summit, is a place where ideas and the money to make them real do collide, it is also something of a celebration of the disrupters.

The conference was built on the idea of disruption and it’s being held in a neighborhood that is working to disrupt its former reputation of being the wrong side of town.

And so, it is very exciting, all this disruption: Ubeam is working to charge our electronic devices with ultrasound energy. Moveable Ink sees a world where every marketing email is tailored expressly to the individual receiver. Playtabase wants our smart homes to be controlled by gestures. Isaac Phillips, a fellow I met while walking to the convention  hall, is launching a startup in Mexico City that serves up ads to smartphone lockscreens in return for paying part of your mobile bill.

All of which is great, but how do these companies — which are at various stages of maturity — move beyond being disrupters and into being businesses? And not just those companies, but all the hundreds of startups at Collision. How do they get traction or “cross the chasm” as business author Geoffrey Moore would say?

If the answer were easy, you’d all be out of here, celebrating your first or tenth million. It’s not easy, but roaming through the cavernous tent city that is Collision 2015, some clues emerged. First, a disrupter is a certain kind of person: a visionary, sure, but also persistent, flexible and extremely confident. Moore writes in “Crossing the Chasm” that conferences like Collision are thick with disruptors, or visionaries as he calls them.

“Visionaries are defining the future,” he writes. “They are easy to strike up a conversation with, and they understand and appreciate what high-tech companies and high-tech products are trying to do. They want to talk ideas with bright people. They are bored with the mundane details of their own industries. They like to talk and think high-tech.”

One more thing that disrupters are: generous with advice. What follows are a few bits of it that I gathered from some of the disrupters I ran into at Collision.

IMG_5517

Meredith Perry, CEO of Ubeam, the ultrasound energy company, said that persistence is a must.

“This company should have died about 15 times over already,” Perry said on stage while being interviewed by venture capitalist Mark Suster, of Upfront Ventures. “It takes that almost delusional mentality, where you will not accept failure. It’s that kind of grit and determination that you need to succeed.”

She said that it’s good to remember that new ideas are often ignored, or ridiculed.

“They don’t think it’s possible or they think it’s not going to work,” she said. “People thought electricity was dangerous. They thought trains would make you vomit. People are just afraid of things that are foreign to them.”

Al Baker, CEO of Playtabase, the gesture-control guy, says having an outside-the-box idea isn’t enough. You need to build a solid business case, maybe especially when your idea is a bit out there. Consumers and companies don’t ask for disruptive products, because they don’t know they exist — or that they’re the answer.

“The responsibility of disruptive companies is to be a thought leader as well as a tech leader,” Baker told me. “It’s almost a burden.”

Davide Vigano, CEO and co-founder of Sensoria, which makes socks that track runners’ running form (as well as other metrics), didn’t get hung up on his team’s original solution or a narrow product. The Sensoria crew’s original idea — take existing flexible electronics and build their technology around them — literally fell apart in the wash. They needed to come up with specially designed sensors and circuits to stand up to machine washing. And while building a system that connected socks to a smartphone app was a good way to prove their general idea, why stop there?

Instead, the company has opened its technology to sports apparel brands and others, who are working on their own applications, including clothing embedded with sensors to track the progress of patients recovering from physical injuries. And Vagiano sees a day that Sensoria’s platform will be used by various brands to optimize a golfer’s swing or a scuba diver’s technique.

“As a developer we are saying, ‘We have built a pressure sensor, but there are multiple other opportunities,” Vigano said. “We fully understand the power of platforms and developers.”

For Adore Me founder and CEO Morgan Hermand-Waiche, the secret to online lingerie is having the discipline to stay on top of every aspect of the business.

“We had to really try to be smart at every level, from a logistics standpoint, from a supply chain standpoint,” Adore Me’s Hermand-Waiche said. “Smartness at every level was the key to succeeding in everything we do.”

Vivek Sharma, CEO of Moveable Ink, the personalized email company, points to simplicity.

“Focus on simplicity and the user experience and also that consistency,” says Sharma, suggesting that in general customers should know what to expect from you. And perhaps more importantly, Sharma has some advice that might not seem that disruptive on the surface: Know your customer and “that you, as a brand, are learning about them with each interaction.”

Then again, think about your own experiences and maybe the notion of a company that really knows you and really listens seems a little more out there.

Photos of Collision and of Meredith Perry and Mark Suster by Mike Cassidy.

Mike Cassidy is BloomReach’s storyteller. Contact him at mike.cassidy@bloomreach.com; follow him on Twitter at @mikecassidy.

6042015816_7753477e76_z

Meetup: Strategies for elastic Solr in the cloud

The challenge of scaling a search platform, like the BloomReach Personal Discovery Platform, is a little like the fading art of plate spinning.

All kinds of priorities — many of them competing with each other — vie for your attention while you do your best to keep everything running smoothly without letting anything crash to the ground.

It’s the nature of juggling a search platform’s heterogeneous workload. You are serving hundreds of millions of documents while battling to maintain low latency and high throughput, all while maintaining efficiency and keeping costs as low as possible.

There are real-time jobs, such as handling queries from our customers’ customers as they search for products online. Those tasks demand extremely low latency for a satisfying user experience. All the while, various batch jobs are plugging away — time-sensitive tasks such as quickly reflecting changes in a retailer’s inventory and pricing, or the work of building thematic pages, which requires a mind-numbing number of product queries across a large number of customer sites. And then there is the need to provide comparative analysis: How many products did a given retailer have available yesterday? How does that compare to today? What’s selling? What’s not?

The jobs are a mix of read-heavy and write-heavy tasks that are potentially pitted against each other — optimizing for one could diminish the ability to do the other. In order to balance the workload, you could say our goal was to spin the four plates that are core to our system’s success:

  • Heterogeneous workloads at very large scale
  • Latency-sensitive tasks
  • Time-sensitive jobs
  • The need to optimize cost and utilization

BloomReach has tackled this problem with a home-grown elastic Solr infrastructure that operates in the cloud. I first described our approach at the Lucene Solr Revolution 2014 in Washington D.C. in November. You can find my deck from that presentation on my LinkedIn page. I’ll be presenting our elastic Solr solution again at BloomReach in Mountain View during an evening SF Bay Apache Lucene/Solr meetup on Feb. 11.

To summarize: BloomReach’s SC2 infrastructure dynamically shrinks and grows the number of search servers handling the workload while providing cluster replacement, cross-data-center support and disaster recovery.

Given the challenges facing us, we opted to build a new cloud-based architecture that was highly reliant on its elastic nature. The system includes a production cluster that is provisioned based on real-time traffic requirements. That cluster is complemented by an automated, elastic Solr system for batch jobs, which grows or shrinks based on demand.

Consider the costly and inefficient alternative of building a system capable of handling peak traffic demands. By definition, the peak is a rare occurrence, meaning any system built to handle that occurrence represents a tremendous waste of money.

Instead, BloomReach relies on its custom API to provide dynamic resource allocation, which determines:

  • How many clusters a particular job needs.
  • How long that job will need those clusters.
  • Whether a job is actively using the assigned clusters.
  • Whether a cluster should be terminated.

The system relies on a home-grown HAFT (highly available fault tolerant) library to copy data among clusters, including transferring data from the static production cluster to those in the dynamic elastic system, ensuring that the data flow is bi-directional.It’s a good thing to be able to move nimbly in all directions whether you’re spinning plates or scaling a search platform. The key thing we’ve discovered is that with the right tools and systems in place, you don’t have to spin those plates alone.

Please join us at BloomReach in Mountain View on Feb. 11 to learn more.

Featured photo of plate spinner by Nina Haghighi published under Creative Commons license.

Nitin Sharma is a member of BloomReach’s technical staff.

 

 

alan-archie

Alan Emtage looks back (briefly) at his search engine Archie

If you don’t know who Alan Emtage is you can Google him — thanks in some part to who Alan Emtage is.

In some ways the search engine that Emtage developed in 1990, Archie, is more famous than he is, which is fine with Emtage, who seems genuinely happy to have played a part in the history of Web search. And why not? Web search has changed the way we do business, live our lives, learn our lessons, connect with each other, work for change.

“So, I look back on it, it was an amazing time,” says Emtage, surveying the early days of Internet search. “I got to work with most of the pioneers, of who people think of as the pioneers, of the Internet: Vint Cerf, Tim Berners-Lee, Jon Postel, the list goes on.”

alanEmtage, 49, was a student at McGill University in Canada when he came up with a way to index and search computer files over a network. These were the days before the Web and hyperlinks. Archie relied on File Transfer Protocol (FTP) and was used mostly by academics and the technically-inclined. The system was designed primarily as a searchable archive of computer programs. The search engine was named Archie, in fact, because Archie is Archive without the “v.” Hey, why not?

I was lucky enough to speak with Emtage while doing some research on the history and evolution of site search. (Stay tuned.) I caught up with him by phone, interestingly enough, as he visited Amish friends in eastern Pennsylvania. Here was one of technology’s leading lights speaking to me from a corn field in a community that prefers to keep technological intrusions to a minimum.

Despite its somewhat obscure beginnings, Emtage’s Archie is widely considered the first search engine and the principles behind it — build and index a catalog of information, then search the catalog to find where the information you’re looking for resides — are essentially the same ones used by contemporary search engines.

Technological evolution rarely takes a linear path and search engines today are the products of a lot of innovation and creative thinking. But there is little doubt that the Web-based search engines that followed Archie — names like AltaVista, Inktomi, Yahoo, Google and Bing — can trace their foundations back to Archie.

Is Emtage proud to have been a part of that evolution? Of course, though he hardly defines himself by that early accomplishment.

“For me, it’s getting to be a bit of ancient history,” says Emtage, who works with Mediapolis, a New York City firm that builds websites and offers consulting to Internet startups. “It was 25 years ago and 25 years is half my life. It’s not something I dwell on.”

But sure, it’s nice to be reminded of his role and to remember now and then, like when a blogger calls to pick his brain about search.

“I’m spared searching for an answer that a lot of people seem to ask themselves as they get older: What legacy am I going to leave,” Emtage says. “I’ve left a legacy. At a certain point it lead to an enormous industry and enormous companies and it’s sort of hard to imagine.”

And maybe it’s only appropriate that the man who moved search so far forward is spared from the need to hunt around for the significance of his own work.

Photo of Alan Emtage courtesy of Mediapolis

Mike Cassidy is BloomReach’s storyteller. Contact him at mike.cassidy@bloomreach.com; follow him on Twitter at @mikecassidy

IMG_2732

Churchill Club retail panel says Amazon can be beaten — maybe

It isn’t often that a panel discussion hosted by the august Churchill Club has me thinking that I’ve stumbled into a book club debating the latest dystopian novel from the author of the “Hunger Games.”

But there on the SAP campus in the hills above Palo Alto recently, a crew of retail experts was talking about dominance, survival, agility, staying hungry and living by one’s wits. The topic was the future of retail — online and in-store — framed by the combative title “Battle for the Buy.”

The short summary: If you’re not Amazon, you’re dead. OK, if you’re not Amazon or maybe Walmart, you’re dead. OK, maybe you’ll make it if you’re not Amazon or Walmart and you come up with a killer customer experience. Oh, and you’re going to need machines to make it happen. The world is moving too fast; inventories are too large, and changing too quickly; and individual consumers describe products in too many different ways for humans to possibly keep up.

IMG_2732Oh, one more thing from Brian Walker, the senior vice president for Hybris, an SAP-owned e-commerce platform provider: “I think it’s also about business processes and being willing to fail fast while you learn how to engage with your customers in a new way that is mobile first, digitally enabled.”

Got that?

Survival for the unAmazons is simply going to take a new way of thinking, Walker and others on the panel said.

“Technology is an enabler,” Walker said, “but it also has to do with the processes and the mindset within an organization, developing more of a one-to-one relationship with your customer, talking about merchandizing in that same construct, getting beyond simple personalization engines into thinking about driving a more holistic engagement with customers, where you are giving them a reason to come back.”

The talk wasn’t all gloom and doom (as you can see in the video of the lengthy discussion below).

The panelists offered potential solutions. And if any group would have ideas about where retail is going and how retailers should navigate the future, it would be this crew. Besides Walker, the panelists included Venky Harinarayan, a digital commerce veteran, who worked at Amazon and @WalmartLabs; Ratnakar Lavu, Kohl’s senior vice president for digital strategy and Patricia Nakache a general partner at Trinity Ventures, who focuses on e-commerce among other things.

 

Harinarayan was perhaps the staunchest of the Amazon-will-conquer-all-crowd. He pointed out that it’s nearly impossible to beat Amazon on retail’s three key differentiators.

“You’re not going to win the selection process with Amazon today,” he said. “You’re not going to make it up on price or convenience anymore. You truly have to be honest with yourself about how you’re going to play that game with Amazon.”

But Walker and others said, not so fast. Brick and mortar stores and online retailers can turn to the customer experience. In-store beacons are coming — technology that will let store employees know that a particular customer has arrived, provided that that customer elects to check-in. Sales associates can be armed with information about that customer’s likes and other information that would help the associate anticipate the shopper’s intent.

Call centers can be reconfigured to cater to in-store shoppers, customers who have specific questions about what’s right in front of them, or what they can’t find. Physical stores can morph into centers that are part showroom and part fulfillment center, serving as warehouses for shipping or pick up.

The possibilities online are even more promising. The personalization that Walker mentioned? Walker and Ratnakar agreed the technology is all but here to provide one-to-one personalization, treating each customer as a distinct individual with individual preferences.

Yes, there is still work to do, Lavu told the crowd. But, “I do think for the first time, actually, the technology is there. Whereas if you asked the traditional retailers in the past, like five years ago, we would have said, ‘Absolutely not.’”

But with big data advances, such as Hadoop, he said, retailers can analyze large sets of data, use predictive algorithms and build a one-to-one understanding.

And on the consumer side, technological advances are both increasing the flexibility that customers have and raising their expectations regarding their shopping experiences. Take the explosion in mobile phones and tablets. Mobile is growing wildly and has tremendous potential, Nakache said.

“You want them to be able to shop right then and there,” she said of consumers. “You need to design an experience that is simple and as frictionless as possible.”

Given smartphones’ small screens and tiny keyboards, customers need to find what they are looking for quickly and with minimum effort. And if they start their search on mobile, they don’t want to have to start their quest all over, should they later move to a tablet or laptop.

“By the way,” Walker added, “the customer just expects consistency, relevance and context. The device? The screen you’re using? That’s just a mechanism,” he said. Adding later: “It’s not just about converting on mobile devices. You have to tie all those channels together and it has to happen right now.”

It all sounds so urgent, the way Walker puts it. And after listening to him and the others at the Churchill Club, it’s hard to argue that it isn’t.

Photo of moderator Avery Lyford of Qubell and panelists Harinarayan, Lavu, Nakache and Walker by Mike Cassidy.

Mike Cassidy is BloomReach’s storyteller. Reach him at mike.cassidy@bloomreach.com and follow him on Twitter at @mikecassidy.

HilaryMason 5.12.14

Accel’s Hilary Mason says big data, or whatever you care to call it, is just getting started

Hilary Mason is not hung up on the terms — big data, thick data, heavy data, dark data, generative data, smart data, juicy data — that have surfaced (thanks LinkedIn followers)  to describe the best use for the massive amounts of digital information that is now at our disposal.

Mason, a data science superstar, prefers to think of sophisticated data crunching as a superpower with the potential to turn any of us into super heroes.

“With any of these terms, it generally takes me awhile thinking about it to come to some understanding,” she said when I asked her about the idea of “thick data” vs. big data. “Big data is one where I don’t like the phrase because it doesn’t really matter how big it is.”

I met up with Mason at the Anita Borg Institute’s annual Women of Vision dinner. She was the keynote speaker at an event that honored Harvey Mudd College President Maria Klawe, whose success encouraging women in tech I’ve written about; Tal Rabin, an IBM researcher and Kathrin Winkler, who leads EMC’s sustainability efforts.

Mason had just finished explaining to the crowd of nearly 1,000 at the Santa Clara Convention Center in Silicon Valley that the ability to rapidly calculate answers to vexing problems was a powerful force that we are just beginning to use.

Mason says data science provides us with superpowers

Mason says data science provides us with superpowers

“When I think about the way that people will use it,” she said of data science, “if everyone in the world were using this thing, how does the world change? What capabilities does humanity really gain?”

Mason, who was Bitly’s chief scientist and is the data scientist in residence at Accel Partners, makes regular appearances on lists of 40 (to watch) under 40 (years old). She’s a big data optimist, whether she cares for the particular term or not.

In her view, big data has the power to make us more creative, insightful and productive. Big data is a partner to us mere humans; a tool that makes us better at what we want to do. And, of course, she is right about that as long as we know what we need data to do for us.

I’ve thought about the question as one of human vs. machines. How much of our intellectual power should we give over to the machines — and when precisely should we intercede and substitute our intellect and instinct for machine learning and processing? It’s a question that Andrew McAfee addressed provocatively in the Harvard Business Review, arguing that human intuition is a real thing and also a faulty thing.

Maybe human vs. machine is the wrong framework

Maybe human vs. machine is the wrong framework

For her part, Mason wonders whether human vs. machine is the right way to frame the question. Despite her deep technical roots, she doesn’t see the questions around data as yes or no, black or white, on or off, 1s or 0s.

“It is the human augmented by the machine, which is why I chose superpowers as the theme of this talk,” she tells me. “The systems, they’re dumb, right? They just run code; so what code are we going to tell them to run? It’s not so much a matter of the human vs. the machine, as it is the human using the machine to process a volume of information that the human unaided simply could not even comprehend.”

Remember, she says, big data is nothing new. We’ve been surrounded by data and lots of it for a very long time. What’s new is the ability to analyze the data at speeds that were once unimaginable.

“Understand that what big data really means is to be able to count things in data sets of any size, rapidly,” Mason says. “And the advantage we actually get from that is not a technical advantage, well it is, but it’s actually a cognitive advantage. I can now run a query and get a result back in 30 seconds. That might have taken 30 hours 10 years ago.”

But having the answer is only part of the equation. We still need humans to tell the story that the data is composing.

“I think the real progress in data science is in telling these stories that come from data,” she says, “and the stories are things only humans would know are interesting.”

It’s an idea that is gaining traction in academia. At its core, it just makes sense. Data, even fully-crunched data, is relatively meaningless without a narrative that answers questions like: So what? Who cares? What next?

In the past year Tricia Wang offered a thought-provoking look at the issue in Ethnography Matters and Christian Madsbjerg and Mikkel Rasmussen presented their thoughts in the Wall Street Journal.

Mason has her own favorite example. She told me about an assignment she gave to a class she was teaching at NYU. She asked students to examine the publicly available data on New York City bridge traffic (how much, on what bridges, during what times etc.). Simple stuff, right?

“Even in that, there was this weird anomalous week. So a computer could have told me that that was an anomalous week,” she says. “The traffic dropped to very low numbers, then to zero for several days.”

She said her human brain first went to conspiracy theory: The government was withholding the information for some reason. “And then the intelligent human in me said, ‘Oh, that’s Hurricane Sandy.’ So there is that example of the algorithm and the human interplay.”

It’s the sort of interplay that Mason and many others are convinced is going to become increasingly important as our data tools become more sophisticated.

Look at the way professionals in all sorts of fields are familiar with spreadsheets, she says. Now think about stepping that up to more sophisticated data wrangling.

“One of the things that’s wonderful about data science as a field right now is that people come to it from many different kinds of backgrounds,” Mason says. “So, I know data scientists who are psychologists, who are political scientists, who are economists, who are mathematicians, who are computer scientists, who are physicists, who are chemists, right? And so, perhaps, as we sort of refine what it means to be a data scientist, we can keep that kind of open-mindedness in the training and then have a pretty broad pool of people who might choose to try that kind of position.”

In short, maybe it’s not so important what we call the vast sea of data out there as long as we understand that it’s our data and that we now have the superpowers we need to do some amazing things with it.

Photo of Hilary Mason by Mike Cassidy, photo of young super heroes by Valentina Powers, image of the machine by Frédéric Bisson published under Creative Commons license.

Mike Cassidy is BloomReach’s storyteller. Reach him at mike.cassidy@bloomreach.com and follow him on Twitter at @mikecassidy.

2.4Netflix

Netflixing the Web

Netflix knows me. The recommendations they serve up for me are terrific and clearly speak to their understanding of my personal tastes. That is a pretty huge leap in the movie watching experience when you consider that not too long ago, I used to walk through the aisles of Blockbuster Video hoping to find something that was half decent and in stock. How does Netflix do it? Data. Data on an individual customer’s behavior and a deep understanding of the movies they carry. In a sense, that’s what BloomReach aspires to do for our customer’s sites. We are Netflixing the web.

netflix iphone (shardayyy)Let’s go back in time briefly and recall what we did to find a decent flick at Blockbuster. They had their selection of VHS tapes DVDs organized into a small handful of categories – drama, comedy, action and children’s. Oh and if you lived in a progressive urban area, maybe they’d have an indie section. That’s it. Then it was up to you to wander the alphabetical aisles (I’ll bet if you could look back at Blockbuster’s rental data, movies that start with a-f titles were far more popular than those in the later half of the alphabet purely because overwhelmed movie renters gave up before getting deeper into the alphabet). The Blockbuster experience never learned from your behavior in order to save you time or delight you with recommendations.

Enter Netflix and their data driven approach to personalizing the customer experience. The magic of their method is certainly in the front end – learning more about you each and every time you engage with their site. But all of that is for naught if they don’t fully understand the attributes of the very movies they carry. Sounds simple right? It is if you are Blockbuster and have a few huge buckets. But that’s not nearly good enough if you have a massive catalog that people must navigate on their TVs, laptops, tablets or phones.

Instead Netflix uses an army of movie watching experts who watch and describe movies with an exhaustive list of tags. Those tags, which cover attributes including the obvious and the obscure – from lead actor and genre to location and “stoner”. These attributes are then combined to create “microgenres” that are extremely niche. For example, “violent action thrillers starring Bruce Willis”, “African American crime documentaries” and “tearjerkers from the 1970s” are microgenres. With focused niches like those in the mix, the obvious question is how many microgenres Netflix has?

76,897.

Soak that in for a second. Blockbuster had maybe ten genres. Netflix has 76,897. Better still, they learn from you and show the movies that you’re most likely to enjoy. This makes discovery of movies relatively effortless, dare I even say, enjoyable. I don’t recall ever enjoying browsing Blockbuster’s racks.

photo (2)The BloomReach approach to personalization is similar but different.  We don’t collect information about you, precisely, but we’ve observed and interpreted so many interactions and the language people use in those experiences to be able to infer intent based on observed anonymous behavior across devices and channels. Take the Dynamic Categories we create on the Neiman Marcus mobile site for example. We use an individual customer’s behavior and our deep understand of the products Neiman Marcus carries to create Dynamic Categories. Those categories put attributes and categories together in unique ways based on the products that individual consumer is looking at. During the last three months, over 3,000 Dynamic Categories were created on Neiman Marcus’ mobile site. On the Just For You page of the site, those Dynamic Categories were mixed and matched to create 30,130 totally unique experiences.

Like Netflix, we aim for the customer to feel like the retailer “knows them” by delivering the right mix of products that suit their personal taste without revealing private and personal information. After all, a shopper gives you a lot of clues about who they are and what they like before you even know their name. Why not use those clues to personalize their experience to what they want? The result is very unique categories such as “New This Week in Sale”, “Ties & Pocket Squares by Brioni”, “New Apparel by Gucci”, and “Retro Sunglasses by Prada.” Creating, grouping and showcasing these categories is a Netflix approach to a world of online shopping that is often stuck in a Blockbuster paradigm.

There is one point where the Netflix and BloomReach models diverge. Netflix employs a small army of people to watch and tag their movies. BloomReach, on the other hand, uses an algorithmic approach to extract the attribute information about our customers’ products and consumer behaviors. At the scale we work in – millions of products across over 100 merchant websites – a manual approach would never scale.

Netflix on an iPhone image by Flickr user Shardayyy used under Creative Commons License.

revenue-search

Generate Big Revenue in Paid Search With Big Data

It’s no secret that one of the fastest-growing ballooning marketing budgetary line item is paid search. Since its inception and – more recently – when Google introduced product listing ads in May 2012, paid-search costs have steadily increased. It’s shocking to hear that the top 10 AdWords spenders are shelling out more than $100,000 a day, with Geico reportedly leading the way at almost $400,000 per day!

While most in the digital commerce world are expecting PLAs to grow in usage compared to text ads, both are considerably important. But, what’s behind the curtain, after a consumer clicks on an ad – whether text or visual? Usually, you can expect poor experiences that have no ability to learn from the consumer’s behavior or intent.

Most retailers would want their customers to have an experience that is just as personalized, intuitive and comprehensive as a personal shopper armed with knowledge of individual behavioral patterns, history, preferences combined with a thorough understanding of every product available. After all, how better to provide exactly what someone wants in the quickest time, while offering opportunities for complementary purchases? But, if you are a retailer with 100,000 different SKUs for potentially millions of different existing and prospective customers, that collection of data points seems impossible to analyze and act upon in a timely fashion. Or is it?

Already, many of the world’s top search engines have done an incredible job with anticipating individual consumer’s queries and predicting the most likely match that fits their intent. Using big data, machine learning and advanced algorithms, it’s almost like Google is our confidant who’s worked with us basically emulating a real-life experience. Doesn’t it logically follow that this level of technology should be accessible to consumers on the other end of the click?

The answer to that problem is exactly what BloomReach Paid Search tries to answer. Every consumer interaction on a retailer’s site reveals intent. Think about when a consumer goes to a dressing room. They try on a garment, but it doesn’t quite work for any number of reasons. A salesperson can bring something in another size, color; can suggest a sale item or similar item that others like the consumer have bought. They may even remember the last time they were in the store. A retailer can have access to all of these intent cues, if interpreted correctly. Consider a customer who moves away from certain racks, toward others, or lets you know “definitely not this one” as bounce rate or time-on-page. Consumers that quickly bounce off of a landing page or convert more on a particular suggested product page are providing small digital cues where an intuitive system can learn.

Then add to that a complete understanding of attributes of every product across a site together with knowledge of consumer interactions and an understanding of constantly changing language models throughout the entire web. Continually collecting and synthesizing these billions of data points make up the core big-data power of BloomReach’s Web Relevance Engine.

The brands that provide premium and superlative experiences to understand their customers’ intent and provide the best content will hold the competitive advantage. As I mentioned in a December 2013 article for Search Engine Watch, technology is the great equalizer. Any brand who invests in big-data technology like BloomReach Paid Search can be just as successful as the behemoths like Amazon. Companies have focused gargantuan resources to widen the top of the marketing funnel – attracting as many as possible to a site, but have left the lower part of the funnel too narrow. In other words, they have dedicated money, research and data resources to predicting and optimizing for consumer searches, but have failed to use a data-driven approach to deliver onsite experiences for prospects that match their intent.

Like I said earlier, so many companies provide poor paid search experiences after investing lots of money. Search nearly any long-tail keyword, and you’ll see a bad landing page experience that requires clicking numerous times for the right produces, assuming anyone can take it. It’s like paying for signage outside a store to market new blue jeans and then having the customer walk in to find t-shirts on the shelves. The customer then is expected to search the bins until they find the advertised jeans. The probability of them returning and their perception of that company are severely downgraded.

Like all tenants of change, the first step is to admit you have a problem. Here are a few questions to understand that will help you gauge your paid search experience (also discussed further in the Search Engine Watch article):

1.       What keeps customers onsite if they don’t find what they want?

2.       Does a site suggest products intelligently for the business?

3.       How does a site adapt as its customers change the way they act?

Recreating the in-store experience at the size, scale and velocity of the constantly evolving online marketplace is not easy, but not impossible. It is a big data problem, one that only a big data solution like BloomReach Paid Search is working to address.

BloomReach_Ready_Badge

Announcing the BloomReach Partner Ecosystem!

BloomReach_Ready_BadgeBloomReach can do a lot to empower online marketing managers, but we also know that we can’t be all things for all people. That is why we partner. Over the past year as we have been quietly building up our partnerships eco-system in order to bring more value to our current and future clients. We have been fortunate to create partnerships with some of the best companies in the world that, like BloomReach, value being smart, innovative, and having a customer first mindset.

We currently have four types of partners, each addressing different challenges facing the online marketer. The first three types are all certified technologies that are “BloomReach Ready”, and include eCommerce platforms, mobile platforms, and tag managements solutions. Each of these have pre-wired integrations which allow for easy and efficient deployment of BloomReach’s solutions.

Our eCommerce platform partners are Demandware, eBay Enterprise (formerly GSI), IBM Websphere Commerce, and SAP Hybris. Through these platforms, sites can manage and adapt to the change digital marketing ecosystem. Putting a ready-to-go solution such as BloomReach Organic Search on a site is simplified for any site using these platforms.

John Mesberg, Vice President of IBM B2B & Commerce Solutions, described the opportunity as:

“With consumers’ heightened expectation for a highly personalized shopping experience, retailers are looking for new and innovative ways to deliver personalized and relevant content to attract and engage shoppers. Together, IBM and BloomReach are helping retailers meet these shopper expectations, increase site traffic and drive sales by engaging customers. Through an enhanced organic search process that automatically adapts to shopper preferences, customers can find the products they are looking for quickly and easily.”

Our mobile platform partners are BrandingBrand, Mad Mobile, Moovweb, and Skava. These platforms enable retailers to have a relevant mobile site for their brand. With the BloomReach Mobile, sites using these mobile platforms can improve their revenue per visit (RPV) with a personalized discovery experience. Through personalized search, More Like This visual browsing, social trending, and cross-device personalization users are given truly unique and engaging experience.

Our tag management partners are BrightTag, Ensighten, TagMan, and Tealium. These solutions expedite deployment of BloomReach Organic Search or BloomReach Mobile.

Our fourth group of partners are Certified Experts. These Certified Experts come from some of the top digital marketing agencies such as 360i, Ovative Group, iProspect, and Stone Temple Consulting. With Certified Experts available, these agencies help clients maximize the value of BloomReach. They are trained on Landing Page Manager and use it to create new pages for their clients sites. Dale Nitschke, Managing Partner at Ovative Group and Former President of Target.com, had this to say about the opportunity the Certified Experts programs provides Ovative Group and their customers:

“Whether the primary goal for brands is to acquire new customers, offer a more seamless cross-channel shopping experience, or grow traffic and sales, those that continue to increase their ability to access and make use of data from both inside and outside the organization will have a significant competitive advantage. BloomReach has the best data engine available to deliver more relevant and personalized experiences across devices and channels, creating higher revenue per visitor.”

Screen Shot 2013-12-04 at 9.57.23 AM

Teaching Machines to be Human: Content Extraction

Imagine that I’ve just described the following product to you:

“Sail the vast and salty sea in the yacht-perfect style of this adorable A-line dress! You’re bound to have a grand old ‘mari’-time in this frock’s deep navy, tank-like top, which features flexible knit fabric adorned by a trio of brassy decorative buttons. The striped cotton skirt flaunts lines of red and blue, and gathered pleats create a flattering and seafaring silhouette. Outfit yourself in this nautical frock alongside strappy, gold-detailed sandals, an anchor-shaped pendant necklace, and UV-protected shades, and take a coastal cruise with your best crew!”

What does the product look like? How would you categorize it?

It’s surprisingly difficult to picture a product when all you’re given is words. And yet, online, often all we have to work with are words. At BloomReach we use advanced Content Extraction technology to paint pictures from words. Deep understanding of content paired with our understanding of query intent enables our customers to give shoppers exactly what they are looking for at a scale never before achieved.

Why bother with content extraction at all? With content extraction, BloomReach can leverage all of the content you have to create rich definitions of what you offer to customers. Product and inventory understanding is normally limited by the manual tagging and feeds that retailers create to define their catalogs. While structured data like tags and feeds are immensely useful, their ability to fully describe a product is limited by the creativity and language of the human that creates the feed. It is impossible for a single person, or even a team of people, to define every product in all of the ways that it could be described. It would simply take too much time and too many resources to even understand the way that human language changes.

BloomReach uses the structured data and then layers the other content that is on the site to build a rich understanding of your products. For instance, we look at the product descriptions on each page (information that is sometimes more rich than the feed) to understand what other information could be gleaned for the product.

Let’s return to our example to see how our content extraction algorithm extracts relevant information from text to paint a rich picture of what the product is.

Here is the product page for our mystery dress:

Screen Shot 2013-12-04 at 9.57.23 AM

And here is how our system breaks down the product description into relevant (green) and irrelevant (red) content.Understanding_Content_with_BloomReach_Attribute_Extraction_10.14.13__1_.docx

From that, we parse what is and isn’t important.

Understanding_Content_with_BloomReach_Attribute_Extraction_-_Google_DriveThe first thing we do is identify that this product is a “dress.” Once we know the product type, we can separate signal from noise by only pulling attributes that we have learned are dress related. We use all of the data sources – such as information from crawling the page, the merchant’s feed, our web-wide understanding of the product “dress” – to generate an overall understanding that this product is: An A-line, mid-length, striped dress that is sleeveless with a tank-like top. We know that it has pleats, brassy decorative buttons, and is nautically themed in the colors of deep navy, blue, white, and red. We even know that it is available in plus-sizes and is made of cotton.

Just as important as knowing what the product is, we also know what the product is not. We know that it is not a necklace, or sandals, or shades, or a yacht. It is not a picture of the sea and isn’t strappy. Nor is it old or vast or salty. Without the power of our incredible engine to understand the context of the content, we would have a terrible explanation of what this item really is.

At their core, our content extraction algorithms borrow heavily from current and past work in information extraction, natural language processing and named-entity recognition. At a conceptual level our system is comprised of Extractors and Formatters.

Extractors are used to extract named entities from both structured and unstructured text using rule-based extraction (pattern matching and vertical specific heuristics) as well as stochastic approaches. Most of our extraction is still dictionary-based and we do use human input to ensure the quality of extraction.

Formatters are used to format the extracted data and augment it with additional information. For instance a parent color formatter would add blue as the parent color where the extracted color was aqua. Additional types of formatters may add synonyms, unit standardization, singular and plural forms of words, etc.

Over 4 years of processing several billion keyword queries, 1 billion consumer interactions and over 10 billion web-pages daily has led us to build a comprehensive understanding of entities and entity relationships that form the backbone of our system.

Understanding content is the foundation of the BloomReach Web Relevance Engine and powers all of our big data applications. It has taken time but with each day we build a better engine that understands new entities and entity relationships. For instance, over time, our system understands that in certain contexts different words mean different things. For instance, if we’re looking at a product page for a cake, the word “chocolate” refers to a flavor. However, if we’re looking at a product page for a rug, the word “chocolate” refers to a color.

Our content extraction understanding makes it possible for us to understand verticals as varied as clothing, furniture, appliances, beauty and even travel and real estate. Each day, we process more data to form a deep understanding of the way the world is described around us. And we use our big data applications to take that understanding and use it to help retailers connect to customers who want what they have.

SEO-for-PPC-success-350x278

Driving Revenue in Paid Search with Big Data

Two years ago, when BloomReach emerged from stealth mode, we revealed our core technology, the Web Relevance Engine, and three applications – BloomSearch (organic search), BloomLift (paid search/SEM) and BloomSocial (social media). In the two years since, we’ve learned in four ways:

  1. BloomSearch became BloomReach Organic Search to bluntly state its focus. Today, more than 100 leading ecommerce sites rely on BloomReach Organic Search to get more of their content found – especially in long-tail organic search.
  2. Marketers want to manage the content quality on their pages and create their own new organic search landing pages, so we developed Landing Page Manager with actionable Content Quality Metrics giving customers more visibility and control.
  3. BloomSocial was an experiment that taught us about the influence and utility of social media. It informed our “What’s Hot” capability in BloomReach Mobile.  “What’s Hot” shows visitors the products generating the most social-media traffic on Pinterest and Facebook from the most influential consumers. Try it on the mobile sites for Neiman Marcus and Deb Shops.
  4. BloomLift initially delivered significant conversion lift for its beta clients by routing long-tail paid search queries to more relevant pages. However, it was limited to a small proportion of their paid-search campaigns, and our beta customers wanted to use it more broadly.

After analyzing paid-search traffic for one of our more complex and branded Internet Retailer Top 50 clients, we observed that more than 67 percent of the query-to-keyword phrases did not lead to a relevant landing page that matched the consumer’s intent because there is no practical and scalable mechanism to create, test and optimize those pages. We rebuilt our paid-search product to provide a big-data-driven tool that would provide insight, management and analytical capabilities for any paid search landing page – even those with few clicks per month. Many companies are wasting money driving traffic that doesn’t convert, and the hardest pages to optimize are the ones where you have the biggest opportunity to differentiate.

This month, we previewed BloomReach Paid Search for our existing customers. BloomReach Paid Search understands all the merchant’s content, queries and consumer behaviors to allow paid-search managers to efficiently improve paid-search landing page experiences for every click. BloomReach Paid Search enables marketers to easily identify and continuously deliver relevant pages for segments of even one.

BloomReach Paid Search will allow marketers to:

  1. Manage their paid search landing pages to maximize relevance – even for lower volume queries – where A/B testing segments and rules don’t have enough volume. Check out http://visualwebsiteoptimizer.com/ab-split-test-duration/, and see how long it would take to reliably test a query for 100 clicks a month, roughly 4 visitors per day.
  2. Surfaces new landing pages and/or allow marketers to dynamically re-rank content on existing page to match queries to landing pages, while easily complementing your bid management platform.
  3. Automatically identify negative keywords where the merchant doesn’t have matching product – either to drive buying decisions or to negative keyword match.

BloomReach Paid Search represents the next step in our quest to drive the best digital experiences possible for consumers and more return for marketers. Stay tuned for our wider launch and customer case studies.

demandware link

BloomReach Becomes Demandware LINK Partner

We are proud to say that BloomReach now has a pre-built integration for BloomReach Organic Search ready for the Demandware Commerce platform. This is a significant achievement both for BloomReach and the many retailers that rely upon Demandware’s LINK partner network to validate quality partners and significantly reduce integration time and costs.

BloomReach Organic Search will offer any Demandware customer the ability to harness the web’s collective intelligence – more than 1 billion consumer interactions and 1.2 billion pages interpreted per week– to capture demand by matching consumer searches with the most relevant content. The application brings between 20 to 80 percent lift in organic search, driving more product discovery on a website. An independent Forrester report released in February stated that the application could provide up to a 196 percent return on investment (ROI) with a payback period of 2.2 days for established e-commerce brands in addition to bringing 60 percent new customers not attracted through other marketing tactics.

This is extremely important in a time when marketers have seen a significant drop in organic search data and insight for action. Not long ago, an organic-search query passed right through to your analytics system, giving you a quantifiable click volume and value (assuming your analytics provided conversion/revenue data). And there was a lot less competition for organic search traffic from other merchants and other content. The focus was justifiably on head and torso term optimization. For example, you could bring up high-volume queries with low conversions with better content or landing pages.

Unfortunately, search marketers no longer have such clarity into their organic-search traffic for two reasons: an increasing percentage of traffic is “no query” and a huge portion of mobile traffic is not categorized correctly, often failing to attribute the traffic to a search rather than “direct.”

In addition to lacking data, real estate is also a growing issue with the space available for organic-search giving way to paid search and product listing ads. At the same time, the majority of online search is still organic, so it’s critical to optimize content for the long-tail organic search consumers, who are often new customers, to grow and maintain e-commerce revenue centers.

The Demandware LINK Technology Partner Program provides Demandware clients with a rich of set of pre-built integrations to cutting-edge commerce technologies and applications that can unlock revenue, generating opportunities and enhance the brand experience. By reducing the cost and complexity of integrations, Demandware LINK allows retailers to adopt innovative third-party technologies quickly and cost-effectively, allowing them to accelerate time to market and attain a faster return on their investment.

BloomReach is proud to be offered with the best of the best, and we look forward to working with Demandware to help more companies get found!

big-data-matrix

Vote BloomReach As A Top Big Data Startup!

As we build the premier big data company for marketers, BloomReach is continuing to rack up industry awards and acknowledgements. One of those distinctions is veteran technology journalist Jeff Vance’s Startup 42 for Big Data, a list of the currently 42 early-stage big data companies making an impact in the tech industry, which is on its way to 50. As the data industry becomes more mature and the market becomes more diversified, great new organizations are formed almost every week, making BloomReach’s placement an important visibility milestone for our company and its awesome collection of growing talent.

BloomReach recently was just honored to be selected as a challenger for a spot in the list, and part of the criteria is a public vote for your choice of top contenders. While we have some good competition, we’d like you – our community of supporters and followers – to vote for BloomReach to secure us a spot. The public vote only represents a portion of selection, but it goes a long way to indicate to Jeff and other industry pundits that BloomReach is a company making an impact in the world of big data, e-commerce and overall technology.

Startup 50 was started after Jeff wrote two related articles about mobile startups for Network World and CIO. After issuing a call for companies through a popular reporter resource, he received more than 150 recommendations. Since putting together an initial “Final Forty” list, venture capital firms and startups alike have strived to have their interests represented on the list.

Armed with a new approach to highlighting new and exciting tech companies to the public light, Jeff’s list has evolved into a top industry distinction and a go-to source for journalists and analysts to understand the different players in big data, among other markets.

With the recent launch of our mobile big data application, BloomReach is barreling forward in 2013, with much more planned for the rest of this year and early 2014. I encourage everyone who follows BloomReach to cast a vote in our favor to ensure that our momentum is noticed by the people and companies who can benefit from our services.

You can follow Jeff Vance’s Twitter (@JWVance) to keep up with his top-notch coverage of technology in a variety of publications including Forbes, CIO and Datamation.

press coverage

Media Responds to BloomReach and Launch of Mobile

When I graduated from LSU in public relations, I would have never thought that I’d build a career in what we be considered “deep tech.” From semiconductors to enterprise software down to the test-and-measurement companies that support enterprise, I’ve touched and represented just about every technology “genre” that Silicon Valley could offer the world. Through the course of that experience, you become a little hardened to innovation (especially coming from PR firms), because you never know what’s walking through the door next.

“What now? What do you do?” Those are the obligatory questions and often skepticism that I found myself asking hundreds of hopefuls. If my journey ever led me to make the leap back to internal PR, I presumed I’d never join a startup. Yep, I said it. Like a journalist who receives hundreds of pitches about industry firsts and next-generations, the value proposition and the competitive differentiators are often hazy, if not completely lost.

However, little did I know that on my first day as a manager for a leading startup PR firm, I would meet a company that would change my perspective – BloomReach. Let’s just say that things “blossomed” since then. From the moment I had my first meeting, I was genuinely astounded by the intelligence, the commitment and challenge that BloomReach encompassed. Depending on what you read or who you trust, you’ll hear that only 10 percent of startups succeed, but BloomReach was clearly different – a company bound for success. OK, so I played a little favoritism while working for them externally, but fast-forward a little under a year later, and I threw my foregone conclusions out of the window and joined the winning team.

That brings us to today after the extremely well-received launch of BloomReach Mobile. Covered by and introduced to a great matrix of business, marketing, search, IT and general tech media outlets, BloomReach Mobile offers a solution for two very important groups at a very important time – consumers and online businesses. As a prime example of this critical junction, check out one of our top venture-capital backers, Ajay Agarwal from Bain Capital Ventures, as he discussed BloomReach Mobile on the nationally broadcasted show Bloomberg West with Emily Chang.

With consumers moving in droves to mobile devices, the overwhelming majority of experiences that await them quite literally stink. Plus, patience for poor experience is downright pathetic. The opportunity for businesses to monetize one of the fastest-growing segments of commerce is tremendous. I’m proud of our dedicated company and its leadership, again exemplified in an excellent interview by BloomReach CEO Raj De Datta in his appearance on TechCrunch TV.

The long-term success of any PR program can only be as strong as the people and the technology behind it, so I know that I made the right decision. Without being anymore long-winded, the media reception for BloomReach Mobile was excellent, and we are confident that it will change online shopping, as Forbes writer Lydia Dishman put it.

 

red-suede-shoes-example

Explaining the Uniqueness Score in CQM

When you land on a category page, you want a good selection of choices that fit exactly the intent (or keyword search) that brought you to that page. You are looking for what you want specifically, and you’d like the retailer to present a unique page that is organized with their best, most relevant products for you to choose.

Delivering a unique experience is not a new concept to marketers. SEO professionals know that search engines value unique content because it provides better user experiences by guiding visitors to more unique and closer-matching landing pages. However, many e-commerce teams don’t know that overlapping pages too closely also creates friction in funneling users to the best possible pages, resulting in bounces instead of conversions. To help marketers better provide unique experiences while tracking overlapping content, the BloomReach Continuous Quality Management (CQM) technology quantifies this content overlap through a Uniqueness Score.

The Uniqueness Score measures how unique the content and set of products are compared to the closest duplicative page on a site. A higher Uniqueness Score means that the overlap of content and products is minimal compared to the next closest page.

For example, consider two pages: “Red Suede Shoes” and “Blue Suede Shoes.” The “Red Suede Shoes” page has 12 products, 10 of which also appear on the “Red Leather Shoes” page. Conversely, the “Blue Suede Shoes” page has seven products, only one of which appears on another page, “Blue Leather Shoes.” In this case, the “Blue Suede Shoes” page provides a more unique experience and receives a higher Uniqueness Score.

Naturally, there will be some product overlap between pages, which is fine. Not everyone searches in the same way, so you need to create clusters of pages with somewhat related content to better match user intent. While building out these pages, the Uniqueness Score can help make sure that the pages you do create are differentiated. A healthy balance of unique pages that are topically related to one another is key to providing customers with a quality site experience.

stats-flux

Looking at the Flux Scores in CQM

There’s a temporal component to a quality web page. Marketers, visitors and search engines appreciate a page that is consistently “good,” meaning the content is what they expect, and there are enough different products to satisfy their desire for choice. So, if a page provides a high-quality experience one day and then a poor experience the next, there could be a problem.

The Flux Scores in BloomReach’s Continuous Quality Management (CQM) technology help marketers measure and act on these fluctuations. Flux Scores reflect how often the content on a particular page is changing. If the content on a page changes too frequently, the page may deliver an unpredictable experience to visitors.

Let me give you an example. Say you have a “Dark Wooden Desk” page that has 10 products on the page today. Tomorrow, six of the desks go out of stock leaving only four left on the page. Two days later, 10 products are added to bring the total products to 14. One week later, 12 products are removed from the page because they are discontinued, leaving only two products.

With so much variability, a page like this can drive an online marketer crazy because it will have performance implications: conversions and revenue will be up one day and down the next as the number of products and the amount of content rises and falls. Also, search engines appreciate content consistency, so pages that fluctuate too frequently may experience an impact to organic search traffic.

No matter what comprises your product catalog, you want a consistent, high-quality experience for customers and search engines. Flux Scores can help you understand where opportunities exist to create more consistent content.

Marketing With IT; Not Marketing Versus IT

Whether you’re in the mailroom or the boardroom; the front lines or behind the curtains, technology advancements have caused unnecessary rifts within organizations. The one thing that hasn’t changed about businesses is that every employee should be in it for the good of the organization. Recently, I shared my thoughts in a leading marketing publication iMedia Connection on how marketing can work together with IT to help companies keep pace with the growth of technology.

I’m sure many of us can remember a time when marketing and information technology could carry on for years in the same organization and never even see each other. Sure, they may have shared a “C” position in the boardroom, but the two might as well have spoken different languages. Subsequently came the rise of the Internet – the need to tract, analyze and adapt to consumers that existed as a collection of “1s” and “0s,” but the hardware and software to act on this data was rudimentary and expensive. Then, the limitations of housing the data virtually erased with the advent of the cloud and other forms of cheaper storage – which blew off the lid of marketing intelligence. The CMO became voracious for data as technology became more sophisticated. Who would have thought that in a few short years, the CMO could outspend CIO for technology?

This is the technology territorial struggle taking place within organizations, which is also gaining more and more attention in both fields. But, which area has the responsibility for the growing demand for technology in marketing? While big data has not traditionally been part of the CMO’s role, CMOs will either have to exert more authority or talk in terms that are important to CIOs.

The CMO is quickly becoming the major data consumer in business, responsible for gathering insights from all the unstructured data available. CMOs also are becoming big users and buyers of technology, which makes them more strategic. Marketing is accustomed to quick experiments, whereas IT thinks in terms of 18-month-long projects. Marketing is often the pebble in IT’s shoe, constantly nagging for one-off projects, quick fixes or data analysis.

Conversely, marketers don’t necessarily factor in the total cost of determining return on investment, nor do they worry about the impact on performance, reliability or security. CIOs might evaluate the platform based on cost savings, productivity increases and impact to the overall infrastructure, whereas marketers look at the marginal revenue gained for bringing in new customers. The central challenge is balancing these two valid, but often disparate objectives in a way that minimizes total cost of ownership and maximizes ROI.

Clearly, there is an evolution taking place, and CMOs can work in partnership with their fellow C-level executives for a win-win. I believe the most successful companies will use technology and expertise to carve out ways to act in two-week increments, implementing large-scale projects that move at the speed of marketing with the reliability and performance of IT. We are already seeing the rise of hybrid roles – a marketing technologist as some have called it (a lot of companies even have their own “Chief Marketing Technologist”). On the other side, data scientists specifically dedicated to handling terabytes of data. Now more than ever before companies are tracking consumers every move – combining intelligence from brick-and-mortar locations to interactions on the web. The emergence of cross- or multi-channel initiatives is quickly becoming a box that both CMOs and CIOs must check for their boss – the CEO. And it should be. Consumers are coming from the Web, the street, social media, mobile devices, emails and soon-to-be-huge voice controls, so the amount of available data is only going to grow exponentially.

Most marketers are ready to pounce on every platform, but by opening up these doors, you expose yourself to a ton of security risks and performance issues that hurt other critical functions of the enterprise. It is the function of IT to scrutinize every aspect of technology integrations accounting for all threats – things that take time which marketing never seems to have. However, there are ways to bridge this gap and both work toward the ultimate common goal – growing a business (and keeping a job!). Here are a couple of successful approaches for marketing when approaching the situation.

Successful Approach 1:

Bring IT into the conversation early after determining that major data-driven initiatives or strategies are needed to effectively compete. When marketing starts to explore technologies to help scale their campaigns, inviting someone from the IT team to join the evaluation and ask their questions early will save everyone time, identify issues marketing might not consider and align everyone towards success. Sending requests “over the transom” breeds frustration and misunderstanding. Identifying key integration concerns is not a conversation for the boardroom – it should start when evaluating different vendors or specific technologies. You should question any marketing technology vendor that claims to be “plug-and-play” yet bring serious revenue generation.

Successful Approach 2:

Consider dedicating or hiring marketing technologists to the marketing team who can liaise or even come from IT. These people are responsible for assessing the technology ramifications of a marketing solution, estimating the implementation issues and often can handle the implementation itself. Face it – most of the time, the language is inherently different, and our world is changing at a pace that only the winners will keep up with. You may remember that there was a time when a “CIO” was a not-so-common position.

Successful Approach 3:

CMOs and CIOs should have a standing meeting to discuss marketing-driven initiatives and understand IT initiatives. Progressive CIOs are allocating time and teams for rapid response to business needs so that the teams on long-term initiatives – re-platforming, new applications – and are not distracted by the repeated requests of marketing.

The bottom line is that successful organizations are adopting better communication and cooperation channels within the organization – not just those at the cutting-edge of technology adoption. This “data thing” isn’t going away, and even SMBs are stepping up their game in big data. As the recent technological disruptions have demonstrated, no one is safe and your competitors are almost assuredly evolving, so are you?

BloomReach Adds ShopLogic Team to Expand Big Data Marketing Applications

When we founded BloomReach more than four years ago, our vision was to create an online experience that was excellent for consumers by providing relevant content, while allowing retailers to get more of their products discovered on the increasingly noisy web. This grand vision doesn’t stop at search, but extends across all channels, especially as the avenues from which consumers come keep increasing.

However, bridging the digital divide between consumers and online retailers is an extremely difficult, data-intensive problem that requires the best talent with broad skillsets from all fields of ecommerce. Today, I am proud to announce that BloomReach has brought on the co-founders behind ShopLogic, the promotions management platform company. When we met CEO Kevin Chan and CTO Dennis Maskevich a little while ago, we recognized the talent and industry knowledge that the two offered and began discussions to bring their expertise to help us meet BloomReach’s significant growth. Kevin and Dennis represent the highest caliber of dedication and drive, and we all decided that the best way to fully realize our shared vision for online commerce was to build upon BloomReach’s technology with them as part of the BloomReach team.

Since launching in 2011, ShopLogic provided an intelligent way to manage and optimize promotional marketing using customer interaction and purchase data. The company was highly successful with many mid-sized B2B customers and helped them to achieve an 11 percent increase in net revenue. Driving greater return for e-commerce marketers through better visibility and profitability is quite a feat with the digital marketplace being as competitive as it is, and we are elated to have Kevin and Dennis onboard as we plan to have a breakout year!

ShopLogic originally launched with seed funding from big data VC firm Data Collective and AngelPad, a mentorship program started by ex-Googler Thomas Korte. Before founding ShopLogic, both Chan and Maskevich worked at Adchemy, a prominent online advertising technology company. ShopLogic’s customers are still able to use its platform, but the technology will operate independently from BloomReach’s platform.

The technology and quality of BloomReach is measured by the excellence of its team, and along with everyone at BloomReach, I encourage you to stay tuned in the coming months for many more exciting announcements.

CQM – Understanding the ‘Content’ Score

“Content is key” could be the most-used, yet least-measured saying in the online world. Ok, so you need content, and it should be good content if you want visitors to engage, subscribe, buy or otherwise convert. But how do you quantify the quality of your content? We’ve looked at this challenge and developed “Content Scores” as part of our CQM technology.

Judging content quality on an editorial, video or photo-sharing site likely revolves around views. But with ecommerce, it’s not that simple. Put yourself in the shopper’s shoes, and you’ll see what I mean. To a shopper, quality content means the page contains what they are looking for, and they have some options in order to find that perfect pair of jeans, non-stick grill pan or bridal-shower gift. You expect the page you land on to wow you with great choices.

So, if we start with a consumer’s high expectations and work backwards into what we would need to measure to ensure the content is high quality, we end up focusing on two things:

1) The relevance of the products to the page theme – basically, how well do the products on the page align to that page? For example, imagine 2 pages with the following titles: Green Lace Dress and Little Black Dress. The Green Lace Dress page has 10 products, 8 of which contain the words “green lace dress” in their titles and descriptions. Little Black Dress has 4 products on the page, and only 1 product contains “little black dress” in the title and description. In this case, the Green Lace Dress page is a better user experience and has better content scores because it has deeper, more relevant content vs. the Little Black Dress page.

2) The number of products on the page. Are there enough products matching the theme to give the shopper the selection they would like? That shopper looking for “minimalist running shoes” is expressing intent to research a purchase, hence the somewhat broad search. If the page only has 1 or 2 pairs, that’s unlikely to satisfy their curiosity as to their shoe options.

High Content Scores will indicate that a page is optimized to match the shopper’s desire for a selection of choices that all suit their needs. After all, if you were the shopper, isn’t that what you’d expect from a great site?

Upcoming posts will cover CQM Uniqueness, Flux and Behavior scores.