Big Data and Local Search: The Netflix Precedent

This article has been written before. Back in June, BuildZoom CEO Jiyan Wei wrote a piece for Search Engine Watch titled How Big Data Is Fundamentally Changing Local Search. In it, Wei covered two areas in particular where the influence of “big data” can be felt in the local search arena: first, in the drive (spearheaded by Google) to improve the relevance of search results, which Wei thinks is driven by analysis of massive amounts of user activity; and second, in the increasing openness of local publishers like CityGrid and Yelp toward sharing content with innovative start-ups. So big data, for Wei, means user statistics driving better search experiences, and open data leading to more innovation. Sounds good so far.

Local search start-up Loku got a good deal of press in 2011, including an article in these pages, due to its claim to have launched the “Google for local” by using “big data tools to make sense of local search through dynamic analysis and a clean presentation of hyperlocal content.” The idea was to present aggregated content from around the web in the form of social mentions, blog posts, check-ins, local media coverage, and the like, in a manner that was meaningful to those searching for local places and recommendations. Loku positioned big data as the opposite of search — as a means to bring content to users rather than a gateway to content located elsewhere. The difference between this concept and old-school curated search was the scalability brought by machines trained to sift through massive amounts of data to find what users would be interested in.

Cut to early 2013. The bloom is somewhat off the rose as far as big data is concerned, and yet its true potential for local is likely still emerging. I can tell you that as much as I’ve heard and read about big data and its implications, the concept didn’t really hit home for me until I encountered a piece on Salon.com last week titled How Netflix is turning viewers into puppets.You’ll detect from the title that author Andrew Leonard has an ax to grind, but the article is a well-balanced discussion of the staggering amounts of data that Netflix collects on user activity — when users pause movies, what they rewind to watch again, which movies are abandoned in the first few minutes, how all these data points link up to user tastes — and how Netflix is turning to this data to help with the design of original programming.

We all know about the very sophisticated recommendation algorithm that Netflix has developed over the years that suggests movies based on prior viewing habits and ratings. Now Netflix is using the same engine to drive creative decisions. As Leonard reports, the new House of Cards miniseries, a remake of the BBC original from 1990, got its start when Netflix analysts realized that viewers who liked the original series were also fans of movies starring Kevin Spacey and movies directed by David Fincher. “Therefore, concluded Netflix executives, a remake of the BBC drama with Spacey and Fincher attached was a no-brainer, to the point that the company committed $100 million for two 13-episode seasons,” Leonard writes.

Just how many of our passive online activities can be made into data points to be examined for purposes of marketing and product development?

It remains to be seen whether this gamble will pay off, but the point should be clear. We’ve entered an era when product decisions can be made not by analysis of demographics or user testing but by extremely fine-tuned measurements of current user activities designed to predict what new activities they will be eager to engage in. For Netflix and companies like it, the question is whether we will soon be seeing movies where scenery, actors, music, and other elements, once the province of creativity, will be geared to appeal to our semiconscious preferences. For local search, the question becomes, Just how many of our passive online activities can be converted into data points to be examined for purposes of marketing and product development?

We already know that companies like Google and Facebook are collecting such data today and using it for things like ad placement and multivariate testing. What I don’t think we’ve seen is a truly innovative attempt to create local discovery tools that tell users about products, services, and local businesses they might want to explore, based on their preferences and activities and those of other users. The model would be similar to the Netflix or Amazon product recommendation system but would extend to all local activities with an online component.

Sure, Facebook’s new Graph Search is an attempt to do something like this, but it is tied very closely to the activities of one’s friends and other publicly identifiable users, in keeping with Facebook’s foundation as a social network. Big data is really about opening up all user activity for cross-referencing — not just the activities of your friends and not just likes and ratings, but every action performed online or on a mobile device by someone in your vicinity. It is, in fact, the opposite of social because big data depends on the ability to draw broad conclusions from anonymous sources.

Moreover, analysis of user activities should be able to tell companies which local search products to build. This already happens in a blunt sense, so that the most popular search categories like restaurants and shopping drive the design of mobile apps today. But the point of a more refined analysis is that it can tell us things we don’t already know, such as the promise associated with combining a BBC television series with a seemingly unrelated actor and director. What product discoveries might be made by performing the same analysis on local search user activity?

Just as with Leonard’s discussion of Netflix, we might wonder if this is a future we want. Tools that predict and cater to every passive whim could be wonderful or they could be unspeakably annoying. But given a properly constructed feedback loop, user data will also tell us exactly where to draw that line.

Damian Rollison is vice president of product and technology at Universal Business Listing, a company dedicated to promoting online visibility for local businesses. He holds degrees from University of California, Berkeley and the University of Virginia, where he worked at the Institute for Advanced Technology in the Humanities. He can be reached via Twitter.