Saturday, January 26, 2013

What does Facebook’s Graph Search mean to analytics?

What does Facebook’s Graph Search mean to analytics?

Paid search has been dominated by Google so far. The company holds enormous amounts of search data that it is able to sell as part of its advertising packages. On top of this, Google runs its free web analytics engine that integrates with its search platform to deliver even better ad placements and results. Advertisers are able to see what people do not just on Google, YouTube, and in Gmail, but also take advantage of Google’s reach across third party web sites in its current ad presence. Google can help to build intimate customer profiles based on these browsing habits, and turns that data into very, very accurate advertising placements based on users’ web history. (Salesforce is a good demonstrator that customer profiles are worth big money) Power plays to fight Google’s foothold in analytics data, search, and search advertising, even those made by Microsoft, have fallen flat so far.

The massive volume behind Facebook’s user base puts the company in a theoretical power position to leverage personal information. The site holds an extensive network of information and social connections that it has never properly leveraged in the past. Running a search had previously only turned up sporadic friends and few relevant results, yet Facebook has seemed relatively uninterested in reworking its data to improve search results or ad revenues.

Press reaction was heated after the publicity surrounding Facebook’s public launch. With 1 in 6 people on the planet (a substantial ratio of whom are active participants in the global economy, rather than an isolated one, which in turn could drive ad sales) on the service there is a wealth of data to be examined and mined on the back end. People put fairly intimate public profiles on the site, build photo libraries with family experiences, make friends, and actively communicate on Facebook’s servers. All of this is fair game for search analysis, and can allow the site to suggest search results that a friend likes. Just as importantly, it is possible to filter out results based on what friends don’t like, or customize results based on other users who also like ‘Game of Thrones’ and Mexican food in the San Francisco area (Facebook’s chosen example).

This information is predicted to have an excellent impact on analytics. Facebook should be able to use customer profiles to deliver extremely targeted results. This should result in excellent placement for things like restaurants or attractions, and should allow Facebook to charge a premium for ad results because friends and friends of friends can be accurately profild and targeted.

What will the results consist of?

Zuckerberg’s stated goal is that:

“We are not indexing the Web,” Zuckerberg said. “We are indexing our map of the [social] graph.” Users can navigate through the 240 billion photos on the network, the trillions of user “likes,” and the connections between users. (Bloomberg)

This is a brilliant concept with a huge flaw. This is a massive network. The data is overwhelms belief. Yet it only consists of data within Facebook. While most news articles fawn over the idea that Facebook will be able to rival Google, the engine currently relies upon core results results that are billed from Bing. Microsoft was in the news for Bing’s results when it first launched, yet it was not the most favorable reporting:

By now, you may have read Danny Sullivan’s recent post: “Google: Bing is Cheating, Copying Our Search Results” and heard Microsoft’s response, “We do not copy Google's results.” However you define copying, the bottom line is, these Bing results came directly from Google. (Google)

Suddenly the issue becomes clearer; Google has been collecting information on its users for longer than Zuckerberg has been programming. The wealth of data contained in their servers extends across the web; the site runs a mapping service that contains location data on every business in the world. The service is supplemented by a review engine. Search users are tracked through YouTube and Gmail profiles to directly monitor viewing habits and conversations. AdSense, DoubleClick, and any site they connect to can trace exactly what users are seeing in enormous portions of the web. Google may not be directly serving the page views on these sites to boost its comScore rating but it certainly benefits from the user data it collects in each and every interaction. It’s worth going deeper here to see where benefits come from.

If Zucks likes it then it must be good for me because he's rich and I want to be rich too!

There is a big component here that is another problem; companies have been buying ‘likes’ on Facebook for years and have turned these trillions of impressions into a garbage pile. If a company bought a ‘like’ so a user would enter a raffle, is it valuable for this or any other company to re-pay to advertise to this person’s friends or to friends of their friends? Facebook has poisoned its ‘likes’ data with bad data through extensive advertising sales, which could make useful analytics data very, very difficult to leverage for beneficial search results.

Why is a customer profile important in search?

Any field a user enters information into on their own should be valuable to shaping results. Case in point; a frequent early complaint about online dating sites like eHarmony was that they would pair vegetarians to hunters. Modern user habits that are monitored and controlled through analytics should actively eliminate these types of mismatches. The concept is simple on the surface but becomes difficult in aggregate. It could be valuable to take everyone who says in their Facebook profiles that they like Tibetan prayer rugs, own a dog, and drive a Subaru, and then list advertisements for Purina’s new vegan dog food as a relevant search result. If this user were shown advertisements for a Christian Singles dating service it may not be a valuable use of the space.

Facebook holds a huge amount of this cross-referential data and has built extensive profiles on all of its users. They also collect information from users who ‘check in’ at events and locations, which can help to build a large map of businesses around their individual business fan pages (that in turn contain huge amounts of business data). This data has shortcomings though; roughly 1% of Facebook users are estimated to interact with advertising. On top of this, almost half of Facebook users say they find Facebook’s interface boring and minimize Facebook interactions. (CNN) This is backed up by the numbers from my last post; 85% of posts are interacted with by less than 5% of users. (Business Insider) Suddenly Facebook’s enormous market advantage could become a huge liability as it depends on the habits of a minority of users.

Mind the gap

The concept that most social users are ‘lurkers’ is backed by hard data, (Wikipedia, Reddit) and it’s actually pretty cool to think about. Facebook has a problem when people don’t fill in their profiles. You may be friends with Pat and Pat may like base jumping, but you could be an avid botany lover who never actually filled in your ‘activities’ section. When Facebook suggests that you graphically line up because you’re friends, they aren’t getting the same type of information that Google has on all of the searches you’ve run in the past.

Instead of being able to look at your history and say ‘You don’t want to see an ad about parachute accessories, you should see an ad about a local gardening store,’ the best Facebook can do is show you what Pat likes. The idea of getting a contextual search answer is great, but analytics need to have more clean data than Facebook can access if they are going to deliver good results.

This is where Facebook holds a great advantage it doesn’t seem to like; the company is ripe for partnerships. The data Facebook holds on users could build phenomenal success on top of Bing’s search if it can leverage all of Bing’s resources. While the information and results are stuck to what’s inside of Facebook there are limitations here. The question is whether Facebook will truly partner with Bing to improve its analytical reach and build a foundation for search results so it can have an excellent set of data to draw from. Until the response engine is broadened and more information is gathered on lurkers, Facebook is going to be playing the search game with a short deck.

You Can't Just Google It!
by: SweetSearch

Edit: 1/27/13
Maybe Graph Search's answer engine will solve first world problems like these.

Some additional, and substantially more bullish, readings on Facebook's new search engine:


  1. Great article Jeff. It is such a common assumption that simply because FB has so many users it is a gold mine to marketers. As you have clearly shown consumer behavior and social behavior are do not necessarily correlate.

  2. Very informative and a great intro to something I'd never thought about before. I like your suggestion about FB connecting with other companies (such as Bing) to fill in its information gap.