Bing's Stefan Weitz Thinks Search Is About to Change Dramatically | Street Fight

Bing’s Stefan Weitz Thinks Search Is About to Change Dramatically

Bing’s Stefan Weitz Thinks Search Is About to Change Dramatically

Stefan-WeitzFor decades, we have filled the web with information about our lives. Photos of cats, our opinions on news, reviews of the businesses in the neighborhood — information created by people to reflect the world.

Stefan Weitz, director of search at Bing and author of the new book, “Search: How the Data Explosion Makes Us Smarter,” says that dynamic may be changing. An Internet of connected devices, replete with sensors measuring everything from our habits to our heartbeats, is yielding a trove of information created solely by machines. For the first time, he argues, these data will reflect the world as it is — not our interpretation of it.

Weitz will present in a keynote at Street Fight’s Local Data Summit next Thursday in Denver. We caught up with the longtime Microsoft veteran to talk about how an explosion of machine-collected data will reshape the search business, creating new ways for consumers to find local information and forcing the industry to rethink one of the most successful business models in the Internet-era.

What’s the most important shift in the way we should think about our relationship to data occurring today?
Until now, so much of the world has been created in an analog environment and then, maybe, digitized into a format that would be used later. But now, what’s happening is that the entire planet is captured digitally — everything we do now is born in digital and that’s a huge differences from even five years ago.

The result of of that massive amount of data is that machines can begin to understand the world as it is versus the world as we describe it. That’s a big difference. These machines, which have been sort of stimulus response mechanisms, have become aware of their actually place in the world. That’s a profound difference in terms of their capabilities. So, we will be able to bridge human capabilities with these mass machine cognition in ways we’ve never seen.

Web search is arguably the most important product in the history of the Internet. What does that change in insight — from what we describe to what actually exists — mean for the way we think about search?
For the past 15 years, search has essentially been a very good URL-to-keyword mapping system.

Going forward, you’re going to see this dramatic change in this notion of “doing a search.” The notion of you being the stimulus for search will likely fade. Instead, any change of state — the fact that it’s 15 degrees in Chicago when you wake up or that you turned right on Madison — will passively spark these implicit searches that will do things on your behalf.

What implications do these changes have for the way the search industry is structured?
In general, search becomes more ambient and less about the destination. Traditionally, everyone made money in search by delivering you to a destination. But because search itself is going to be so different, it means all of the players will have to monetize that user intent in a different way. [Search engines] in the end are tremendously expensive to build and run, so you do have to figure out way make money — and a good amount of it.  But the old model that drove so much revenue — mapping a keyword to an ad — is likely not it.

When we think about local search, content has dominated the way we discover businesses. Do you see this explosion of data as a threat to the semantic review?
We’re going to this of proliferation of new data points and models that can indicate your actual satisfaction with a place versus the reported satisfaction with a place via a rating system. Today, it’s all based on stars and words; but the future, even today, is the ability for search systems to figure out what it is we really do care about versus what we we report through reviews.

In local, we’ve seen a deluge of new startups that are building products for specific industries — like home services, for example. Do you see verticalization as sustainable strategy?
I think the trend stems from a realization by these companies that they need a business model. So they’re looking at industries like home services, where you traditionally can get a commission off the contractor, or they’re looking at restaurants, where you can get a commission off a reservation. They’re almost lead generation systems. A lot of the verticalization is coming from the realization that they need to make revenue. They’re looking for the low-hanging fruit that has high audience appeal, high competition, and a tendency toward lead generation in the past.

What we’re seeing now is the taking of the 1.0 economy and moving it into the websphere. Whether it’s the long-term play? I don’t think so. I think you’re going to see pressure on a lot of those companies as they find that the very expensive bounty some of these sites extract from the service providers might not be sustainable in the long term.

A big part of the pitch of these companies is the ability to tie into the actual transactions. Do you see commerce and search converging?
It’s tricky because on the one hand you want search to be the unbiased broker, so the second you bring commerce in into core algorithmic results,  not clearly marked as ads, it starts to get a little bit wonky. Search has traditionally been a little skittish to do too much direct commerce inside of the results.

We are seeing increasing examples of more comfort. However, I think it’s more about the task completion metaphor than the commercial metaphor. How can we get people to go from idea to an action more effectively — and in many ways that’s best solved through a commercial transaction. Other times its something else. Either way, its best thought of in terms of task completion.

Earlier, you outlined a search model similar to Google Now where content is pushed to users via a machine that can understand passive behavioral data. How close are we to having those technologies work for the mainstream consumer.
We’re shockingly close to being able to model certain behaviors. If we collect traffic patterns for a few weeks, we can predict with an 85% confidence where you’ll go when you get up any part morning. Humans are remarkably pattern-driven. That being said, it’s still 15% of the time when we’re getting it wrong. That’s a very big failure rate depending on the importance of the action we’re recommending. But as the resolution of our lives increases ever more so, I think that will easier for us to model.

Steven Jacobs is Street Fight’s deputy editor.

See Stefan Weitz’s keynote next week at Street Fight’s Local Data Summit in Denver — click here for more info and tickets.