Blog post

How Google Got Game (Theory)

By Martin Kihn | September 04, 2013 | 0 Comments


So you’re a coach. You have a team, a sweet season. You look back and ask, “How do I know which player contributed what to all this sweetness?” and a vision of a 90-year old economist named Lloyd Shapley appears . . . and you wake up and realize you’re just a digital marketer lost in a fantasy world.

As Freud said, a dream is the fulfillment of a wish. Turns out Lloyd Shapley has an answer for you, too.

Shapley is a game theorist who devised a formula called the Shapley value in the 1950’s. It describes a way to assign credit among a group of players who cooperate for a certain end. It has been used to value individual members of sports teams, polluting manufacturers . . . and now, digital marketing tactics.

Shapley value is the gizmo behind the attribution feature Google just added to its Google Analytics Premium product. This new offering is a big improvement on the already-impressive marketing funnel and rules-based attribution available in Google Analytics (free and premium) and vaults Google closer to the enterprise-class orbit occupied by stand-alone providers such as Visual IQ and Adometry.

Truth is, most of your target customers are probably exposed to different messages on different channels. Yesterday a banner ad and a TV spot, today a search ad, two more banners, an e-mail and a magazine spread. More money, more problems. You’d like to know what’s working and what isn’t — that is, what’s contributing how much to that final sale — and thus: attribution modeling.

Solutions such as Google’s pack in as much data as possible about a particular users’ exposure to your ads — what, when and where — and look at user-level data across vast swarms of humans. It then constructs a best-fit description of the info it has, saying things like, “Display ads contributed 21.2% ($45K) to the campaign’s success.”

But as Google’s resident evangelist Avinash Kaushik recently warned, “There are few things more complicated in analytics . . . than multi-channel attribution modeling.”

So in the words of a musical team from the pre-Google era, don’t believe the hype. At least, not all of it. There are still important things no attribution model can do:

  1. Include all relevant data — offline media and online media viewed on different devices are generally not in the model (although Google’s Universal Analytics is starting to bridge this gap)
  2. Measure brand or social impact — attribution works best with e-commerce variables such as online sales; it doesn’t incorporate success metrics important to brand advertisers, such as consideration, or positive impact on social conversation
  3. Go very deep — the models work best at the channel level (e.g., paid search vs. e-mail) and get less useful, even irrelevant, as you drill down into detail (e.g., sites, placements, sizes, creative versions)
  4. Tell you why — you may find out e-mail contributed a lot more than you thought, but you’re on your own figuring out why
  5. Be easy to act on — without knowing why something worked (or didn’t), you may find yourself with more questions to deal with (e.g., “Would a cat have done better in that video banner?”), before you’re comfortable shifting budget around

In the end, an attribution model is a lot like the handicap on your favorite sports team: informative, sure, but no guarantee of future success.

Upcoming Events: Please join Gartner’s digital marketing analysts for a presentation and networking with colleagues in content and data-driven marketing. Click on your city below for more information:

  • Boston, Tues. 9/17 & New York City, Wed. 9/18 – “The Three C’s of Content Marketing”
  • Chicago, Wed. 9/25 & San Francisco, Thurs. 9/26 – “Driving Customer Engagement with Data-Driven Marketing”

Comments are closed