Aaron Sloman is Veebit’s Chief Strategy Officer, the person primarily responsible for making sure Veebit happens from a product development and design standpoint, and happens in line with our business goals. As with all of Veebit’s core team members, he’s not only exceptional at what he does but also a sharp observer of the world we live in.
I wanted to take advantage of Aaron’s insights into search, personalization, Big Data and predictive analytics to put these topics into perspective for the non-technical person. So I asked him some of the questions that I myself had when we first launched Veebit. Given our focus on developing a new personalization model — one that gives individuals a far greater degree of control than anything they’ve ever had before — these are the kinds of things we at Veebit think about every day. In a different way, maybe, so does everyone who lives in the digital world.
I know that you’ve worked in technology since the early days of the consumer web. Can you talk a bit about how things have evolved since your career began?
I’ve been lucky to be a part of several waves of technology over the course of my career. When I started, things like PCs with Netware Lans were replacing terminals. I was able to get other people onto that wave and onto the very first versions of the Apple Macintoshes and Windows machines. I was a teenager then. I had a lot of energy and passion.
As that morphed and grew over the years, Microsoft came out with the Windows NT OS and I had a VPN/Unix background, so everything came pretty easy to me. The first conference I ever went to was Internet World in Boston, and there was this guy, Jim Barksdale, who had this idea to build a thing called a browser and a protocol called HTTP. I was in the room with 25 other people and, you know, wondering how this would become popular. I saw it evolve from there into Netscape.
When I joined Microsoft, I was a technologist. I caught the wave of the Internet. I got put on the .com team and talked to the biggest customers about how they would adopt the new technology. Lots of the technology was not really designed for the Internet, so I had to make it work. We had a lot of workarounds and custom things to make things work properly, but it was fun—and it evolved to Microsoft moving to .net, to the introduction Java, to things like open standards and architectures becoming widespread. Today the new disruptive devices like iPhone and Android phones are just the evolution of that. When someone talks to me about something related to technology, I connect it back to my experiences about how these systems were developed and can quickly understand new development environments, operating systems and architecture approaches.
Is this true of Big Data, too?
To me, Big Data is just data. Only now we have databases that can store more data. Ten years ago, I was working with companies that had Big Data, only we had to put it in more databases. We could ask the same questions: It just took a lot more work to get answers.
I remember in 1998-99 working with personalization engines that would give you a recommendation when you looked at a product. I was using search because we didn’t yet have prediction engines. If you think about it, search is where you ask a question and get an answer. Even today, when you look at a product and are shown three related products, the web site is often doing that through a search platform as the indexes are optimized for those types of queries.
Has search become more nuanced and sophisticated over the years?
I don’t know that search has gotten more sophisticated. I just think, about 15 years ago, it used to be a black box. Now, when we run a search we’re actually running four or five or six different searches and then bringing them all back together. We’re looking to see what people have been been looking at, running a search on offers and deals, and running a search to drive some personalization based on what page they came in from. All of those things together generate a rated and ranked results list and drive a unique user experience.
For the most part, today’s technology isn’t so different from how it has been for a long time. It’s just that we’ve gotten better in our ability to merchandize results and provide smarter insights. The big change I’ve seen is in predictive technology. It’s not really search anymore. It’s going down a new rabbit hole. Predictive analytics is a lot more about looking at trends and patterns and applying that with some measure of probability to provide an outcome. Predictability is, “Hey people who search for this also search for that.” So if I come along and look at a particular item, predictive analytics is about working out the chance I’m going to look at something else in particular. Predictive analytics is all about trying to be one or two steps ahead of the user in terms of what we think they are going to do.
Where does that predicting get done? Who’s paying attention when I’m on Google and searching something? Google? Or is there a big repository in the sky that anyone can tap into?
Everybody has their own repositories and they generally don’t share. You can subscribe to APIs from major search providers and get a subset of the data, but really predictive analytics aren’t created when the user is out on Google. They happen when the user enters the site and you know what they came in looking for. At that point, you get to mine the full experience of what they are doing. Some of the bigger platforms, like Google or Bing, can cookie you and track you around the Internet to see what you are doing to target that into some ad, but that’s an advertising platform, which is different.
So when you search for something on Google or Facebook and you go to another site that has a column of ads and you notice an ad for the same kind of shoes you were looking at, that’s what’s happening?
Yes. I was looking at a wine cooler last week and now every time I’m on Facebook I see ads for wine coolers. They don’t know that I already bought one, if they were using predictive analytics, they would know I am a mission based shopper and when I look for something there’s a high level of confidence I will buy it.
It seems to me that personalization is a good thing for both the end user and the company using it to drive engagement. So why is there such a backlash against it?
There are two concerns around it. First, privacy. Doing predictive analytics implies that someone understands what you are looking for and thinking about, and that someone can present something to you that has a high probability to convert you. Predictive analytics implies other people are learning all about you. Now that Google AdWords knows I was looking for a wine cooler, is there a concern that other organizations know I was looking for a wine cooler? Will those organizations determine that I’m an alcoholic—or worse? The fear is that with all this data out there, combined with a very, very relaxed set of policies, government organizations will begin to monitor me and impose restrictions on me. We just saw the UK come out with new laws, forcing US (and other countries) companies to follow much tougher rules around storing UK citizen data outside of the UK.
We’re seeing a strong backlash against data being collected. There are people who will take the time to turn off cookies, anonymize themselves. The private VPN market is getting huge: You can jump on one and nobody will know you are using the Internet from your home in, say, Southern California.
The second concern is that the personalized experience being offered to you is really just the experience created by the person who is paying the most. It’s not actually what you want, it’s what the merchandiser thinks you should get. So, I’m looking for a wine fridge, and I’m getting bombarded by ads for low-priced, high-volume wine fridges, but that’s not what I’m interested in. And now I have to cut through all kinds of clutter to find what I want. What’s happening is that people with the bigger budgets are controlling the market. Give them a really good internet ad-buying strategy and watch them sell the #&*! out of their product.
So it’s not just about the accuracy of the advertising but also the objectivity of the recommendations coming through—which may be influenced by who’s paying to get access to you.
If a user really wants predictive analytics, and if the engine is truly smart, then it will recommend the best solution, not the solution paying the most cents per click, or the solution from the company with the best marketing view, or the solution sponsored by the political party trying to learn the behaviors of a certain set of voters to influence them. So companies like Apple who are not in the ad business, with iOS9, are starting to block advertising and pop-ups on websites. You could say that Apple is trying to play privacy guide and influence what you see in a different direction.
Given concerns with privacy, why hasn’t a platform grown up around users themselves building their own profiles, with a high degree of transparency around what’s being gathered about them?
I think it’s probably because doing so would require many big companies to collaborate. And the companies don’t want to. Take single sign on. Facebook is winning at that, with Google a close second. You see Microsoft using Live, Apple using iCloud, and Google using your Google Plus account. And then you have Facebook. Let’s say you’re visiting Best Buy to buy a Microsoft Surface and assume your login credentials will be the same as your Facebook account. But no. Microsoft would never allow that. They want to retain ownership of the user, too. So personalization linked by identity isn’t connecting up across the internet. It’s because there’s something of a turf war going on for your data.
Going back to the notion of influence and recommendations, I suppose it can be like talking to a salesperson who will map your preferences to what’s available. Invariably, he won’t be presenting you with all possible options. So, in a way, any recommendation you receive is biased, yes?
Yes. Even search is biased, Search is a marketing tool.
When it comes to influence, we hear a lot about how people trust recommendations from their friends. Is that why Facebook is becoming the aggregator of everything—of friends, of news, of advertisements? Is that why I can now search Facebook as well as the Web from within Facebook?
Let’s say you’re out to dinner and a friend says to you, “I just bought these shoes and wow, are they comfortable.” Then you ask, “Can I write that down?” Afterwards, you go buy them. Using that one degree of separation and something like a social network on Facebook gives you that same capability. People buy from people they trust, that’s an established fact, and they trust recommendations from friends above all. So if one of my friends reviews something, I’ll trust their review before anyone else’s. It’s all about reputation. It’s a huge part of digital business, and it can really influence a buyer’s decision.
Where does that leave the expert? The thing about getting a recommendation from a friend is that if you like your friend’s new shoes you’re likely to buy them—even if you hadn’t wanted new shoes to begin with. A search engine relies on you to come in and say, “I’m looking to buy new shoes”. Your friend just shows up—wham.
If you think about it, that’s the difference between predictive analytics and search. When I go to Amazon now, they do a pretty good job of simply saying, “Here are some ideas for you. Be inspired.” It’s driven by what I’ve searched for, sure—but also by what I’ve bought before. They know I’m a tech guy because I’ve bought nearly thousands of dollars worth of tech stuff online in the last few months. So they know to show me a new version of a phone. They know to show me the new version of a Roku before it comes out to see if I want to be an early adopter. Amazon knows because I’ve fallen into certain categories. They don’t need to rely on search. They just know who I am, what I buy and what consumable products I need to re-buy.
Or not, right? I mean, 85% of the people who are interested in duck hunting might also be interested in fly fishing but I’m in the 15% who aren’t…
This might be a more helpful example: You could infer that I may like camping because I like the outdoors, so I’m more willing to invest and support political causes to save the environment. The inferences stem into different segments, and the confidence intervals in the systems get lower and lower, but when we get enough hits in the same categories coming from three or four different angles then we get enough confidence that we’ve identified the right segment. People fall in multiple buckets, right? They aren’t just duck hunters. They’re teachers, have kids, etc. That’s the frustration, right? The inference is only as good as what the engine knows about you.
So if we have an engine that knows you intimately, the predictions and recommendations should be more or less spot on.
Yes, which brings us back to Veebit.
Aaron Sloman is Veebit’s Chief Strategy Officer and oversees all aspects of product development for the company and its clients, bridging the gap between business strategy and technology implementation.