Google Goes After Data-Driven Attribution To Upend “Last-Click” Modeling

Bill KeeMarketers are looking for better methods to assign value across the entire customer journey and Google appears to be answering the call.

Today, building on its Multichannel Funnels and Attribution Model Comparison tools for Google Analytics, Google has rolled out Data-Driven Attribution globally for Google Analytics Premium customers.

Bill Kee, product manager for attribution at Google, spoke with AdExchanger.

How are you bringing more business intelligence into Google Analytics?

At a high-level, it’s the evolution of our attribution offering. Advertisers today have the ability to measure and collect data about their advertising, but actually understanding which aspects of their advertising campaigns are working and which are not is very difficult, particularly in a multichannel context.

With the Multichannel Funnels product and the Attribution Modeling tool that we’ve had for a while now, we’ve exposed a view that lets you see all the steps that a customer engages with along the path to purchase. They might see display ads or click on search ads or engage with social prior to making a purchase and you can get insight into all of that.

Multichannel Funnels wasn’t about saying, ‘Which of those things is contributing how much?’ but it was giving you insight in to what that journey was. With the Attribution Modeling tool, we had added this ability to compare different models for distributing credit to those touch points, so if a user had seen a display ad and clicked on a search ad and clicked on an email link, for instance, by applying an Attribution Model, it would give equal credit to all of those things or it would give different amounts of credit to the first, middle and last interaction.

We think that’s extremely useful for getting perspective beyond the last click, but we weren’t in the position of making any kind of recommendation as to which of those models was more appropriate or more accurate. With Data-Driven Attribution, we are introducing a model where we are, deriving from data, an estimate for how the credit should be distributed. It’s moving from a world where you have many different possible models to compare with to one for customers who have large enough data sets, to actually derive a model that gives an estimate and point of view on how the different touch points contributed.

There’s a cross-platform tie-in here with Google Display, DoubleClick, YouTube, etc. Any tie-in to Google Product Listing Ads and commerce?

Absolutely. Google Analytics has a strong integration with AdWords. All the interactions that a user may have along the path to conversion, when those involve search clicks, we’re able to provide really good analysis around every level of granularity. In the case of things like PLAs – all of the detailed information about that ad and what triggered it.

We’ve seen so much enterprise M&A movement in digital marketing and Web analytics, from Adobe/Omniture to IBM/Unica/Coremetrics. What is Google’s enterprise strategy?

We really see this as being a key part of doubling down on our Premium strategy. Premium is an area we launched 18 months ago and we’ve really seen a lot of momentum there and this is continuing that. Over the last couple of years, we’ve invested a lot more into the people, product and marketing side of that business and really, that’s our offering for enterprise-class products. We’ve seen a lot of growth there and we’re up to hundreds of customers at this point, which is pretty remarkable in an 18-month time period.

Are there any meetings of like minds between your data-driven attribution line and Google’s big data product, BigQuery?

Absolutely. That’s an area where we have a direct integration, where customers can take some of their data and begin to analyze it within BigQuery. It’s still relatively early, that offering, and we’re still exploring it, but we’re working to build that out as an even more robust part of the product itself. The idea is, with large sets of granular data, our customers can take advantage of that highly scalable capability to do SQL-like queries within BigQuery, which would be very difficult to achieve on their own. We frequently see that for large enterprise customers, when they try to do ad-hoc analysis, they’re limited by what their spreadsheet program can handle in terms of the number of rows. We see that as being something that can solve a huge issue for a lot of customers.

What’s your core, data-driven approach to attribution?

Data-driven attribution is really a key moment both for the development around our sophisticated marketing capabilities, as well as for the premium product as well.

When customers look at these types of techniques that are available in the industry, one of the challenges is that they feel very “black-boxed” and it’s hard for the organization to see how the model is actually working and assigning credit. One of the key features we’re introducing is this concept of Model Explorer, which allows the customer to dig in to the model themselves and see how it’s assigning credit, and it really helps them get around this black box problem.

The model will give a valuation to one marketing campaign and you can actually dig in with Model Explorer and see how that credit is being distributed. In terms of our cross-platform integrations, with AdWords and the Google Display Network and the ability to tie YouTube TrueView ads in to this view, you’re able to understand what the contribution of seeing a TrueView ad prior to conversion is. It’s a big capability for customers who are trying to understand what role video advertising plays in the return on investment they’re getting from their ads. The cross-platform integrations are a key part of this.


Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. I’m so looking forward to seeing more on this. It favours the value-driven media businesses over the sales-driven organisations that focus on easy to game metrics rather than underlying value and that’s going to be interesting.

    My questions:

    1) What data can it easily take in other than from Google properties? At a simple level will it make impression and click trackers available for insertion in non-Google campaigns?

    2) When will it be possible to feed optimisation decisions from the attribution modelling automatically back into the buying channels in real time? This would move us from optimising to correlation to optimising to uplift and then the game really changes…

  2. Seth Ulinski

    John W. – my guess is the strongest application (and insights) is currently within Google properties (closed system); even within the Google universe proper mobile tracking/analytics is likely challenging. This is where open analytics platforms using “cookieless” device recognition technologies becomes interesting. Still, significant progress by Google in breaking down the proverbial silos.