Factual CEO: Machine Learning Could Upend Local Search | Street Fight

Factual CEO: Machine Learning Could Upend Local Search

Factual CEO: Machine Learning Could Upend Local Search

Gil ElbazGil Elbaz has built his life around numbers. The Los Angeles native spent time as an engineer at IBM in the early 1990s and later made his name building the technology behind AdSense, the product that helped Google turn a popular product into an exceedingly profitable business.

These days, Elbaz runs Factual, the L.A.-based startup that provides the local data that underlies some of the biggest names in mobile apps. Elbaz will sit down with Laura Rich to discuss further at Street Fight’s Local Data Summit on March 5th in Denver.

We caught up with him recently to talk about the future of search, what role Yelp may play in it, and why he thinks machine learning will create a game changing product in the next 18 months.

You spent a decade in the search industry working at Applied Semantics and then Google after the acquisition. For better or worse, the search experience hasn’t changed much since you left Google. What will drive innovation?
In general, innovation in [search] typically comes down to connecting the dots between various silos of data. And I think the development of a lot of new data APIs, whether that’s in social or elsewhere, has made it easier for developers to innovate. To me, some of the interesting things are the apps that can help you reduce the amount you interact with a device — in some cases zero clicks. I’m still a big believer that people want an easier experience on mobile and they want to do the least amount possible.

How does mobile figure into the equation? A lot of these new vertical sites do not seem explicitly unique to mobile.
Search is an interesting category on mobile because you don’t have the same real estate to have multiple tabs open; there’s just a lot less tolerance for maybes and probablys. There’s less browsing and more desire among consumer to know what you’re about to click on is what you want and is useful information for you. Consequently, we’ve certainly seeing a lot fewer blue links, and more and more, we’re seeing less people use general search in favor of vertical alternatives.

But in the next few months, you’re going to see a few new players in mobile search that are going to make us rethink the process. People have gotten used to the fact that you don’t really need a search engine on mobile. You just go to your favorite app for your favorite use case. But I think the pendulum is going to swing in the other direction; it’s just a matter of allowing that new innovation to thrive by making sure the right APIs are available and the VC community believes there can be a new winner in mobile search. I think that company is going to be very interesting

You mentioned a shift from searching for links to searching for facts. Do you imagine a time when we look for applications on mobile the way we search for webpages online?
Even though there’s a switch to wanting the information itself, I think there’s plenty of opportunity to index the knowledge that’s [in apps]. It’s just a question of leading with the answer and then using the link as a way to secure the opportunity. For example, you can lead with the answer that there’s an open reservation at your favorite restaurant at 6:30pm, and then link to the OpenTable deep page where you can complete the reservation. You still need the link, but you lead with the answer.  In that way mobile search can look pretty similar to web search.

Yelp is an interesting case because it emerged during the transition from web to mobile. It’s hard to tell whether it rode the growth of mobile or succeeds because of the maturity of the web. Do you see Yelp as an image of the past or vision of the future?
On one hand, Yelp has this amazing content: 70 million or so mostly quality reviews. And they’ve been around this space a long time. But they haven’t really tried to push the needle very far forward in terms of machine learning and recommendations. One of the reasons is that they’ve had a model that worked — so why mess with the magic?

But I think when you’re doing anything around machine learning and recommendations, you want it to be really good or else it gets annoying. Doing that takes a tremendous amount of data and machine learning know-how.

But I believe you’ll see a game changing technology over the next 12-18 months — something that when you see it, you’ll look back and not remember how you could manage life without a helpful assistant that’s always looking out for you.

Back to Yelp quickly. Would those recommendation algorithms simply sit on top of the content, analyzing sentiment in the way we read reviews today? Or does that type of future need its own type of data? Does content become less relevant in the age of data?
The reviews are valuable but they’re even more valuable if you can turn them into patterns and actions. If you can know that someone has a better time in a noisy, trendy place, you can offer a better recommendation. But to figure that out, you have to know the places they’ve been and then marry that with data about the ambiance of a location. That information might be hidden reviews but it might not be structured yet. The question is who’s going to do that work.

Why hasn’t Yelp, or really any other company, made progress in understanding our behavior — not just the quality of the destination?
For one, I think a lot of us [in the industry] focus on technologies that move toward advertising where the impact is extremely measurable. If you improve by a few percentage points, there’s revenue you can generate. Frankly, a lot of apps do put a lot of emphasis to figure out if personalization of advertising messages can make their model stronger. But a typical app doesn’t have the engineering resources to focus on user personalization, which is considered a nice-to-have more than a need-to-have. In the end, it probably gets cut.

But I’d like to think that it’s just a matter of time before consumers set a higher expectations of what they expect out of an app. Soon, they will these apps not to miss anything.

Let’s talk a bit about recommendations and machine learning. What’s driving that technology beyond, as you put it, a nice-to-have?
The bigger narrative is that the mobile device, through these apps and through the sensors on the phone that are getting much more intelligent. And this intelligence is manifesting itself in improved search results, better data, and better suggestions. But it’s not always a response to a query: it might be a notification that reminds you of something relevant to you, sort of like what we’re seeing with Google Now.

In terms of what’s driving change: there’s been a huge growth in the number of apps that are willing to collect location information. That’s something that’s taken some time to become a norm — developers were worried about power management and other user concerns. And its taken a while for app developers to come up with good justification for getting this location information.

Often, it’s a killer application that makes a given technology take off. But sometimes it’s just a matter of refining. Which bucket do you see machine learning falling into?
A lot of the times it’s some sort of killer product that brings an emerging technology to the forefront of our consciousness, that’s something that can impact every part of our lives. As someone who’s been thinking about data forever, I still can’t believe that health industry works in the way it does. That people can make so many mistakes about their health without receiving a push notification telling us were doing something wrong is astounding to me.

Steven Jacobs is Street Fight’s deputy editor.

See Gil Elbaz speak at Street Fight’s Local Data Summit on March 5th in Denver (click here for more info and tickets).