Advertising Impact – Not ‘Attribution’

Data-Driven Thinking“Data-Driven Thinking” is a column written by members of the media community and containing fresh ideas on the digital revolution in media.

Elizabeth Zalman is co-Founder at Media Armor, a mobile advertising technology company.

“The more complex societies get and the more complex the networks of interdependence within and beyond community and national borders get, the more people are forced in their own interests to find non-zero-sum solutions. That is, win–win solutions instead of win–lose solutions…. Because we find as our interdependence increases that, on the whole, we do better when other people do better as well — so we have to find ways that we can all win, we have to accommodate each other.”

President Bill Clinton, December 2000, Wired Magazine

At the heart of any digital advertising initiative is that most basic concept of cause and effect: does the program lead to success? This is simple with only one type of media in play: if there is an increase in activity, we infer that sales (or whatever the metric of success is) were due to the lone program. With the plethora of ways to communicate with consumers, it has become a challenge for marketers to understand which ones have most effectively translated spend into success. Enter attribution, and with force: companies have sprung up everywhere whose entire business is telling marketers what led to their sale. The language of ‘attribution’ – meaning to regard one thing to be caused by another – is intuitive, but ultimately misleading as it is currently used in the industry. Whatever the definition of the word, the process known in digital advertising as ‘attribution’ doesn’t actually tell marketers what drove a sale; that is, whether their advertising efforts worked.

As a former colleague of mine laments with vehemence, attribution (in the digital realm) is ultimately a zero-sum game driven by the need to ‘tie-out’ all expenditures to each individual marketing manager’s budget. Rarely does it provide insight into what efforts drove additional revenue. In effect, attribution only tells marketers which pieces of media touched a consumer throughout a conversation, and then assigns each some value. Driven by the marketer’s accountant precept to not ‘double dip’ on a sale, it does not capture or capitalize on the complexity and interdependence of multiple, overlapping media programs.

The failure of ‘attribution’ to attribute cannot be understated, because the ability to infer the cause (or relative rank of partial causes) of a sale is the only valid method for accountability of advertising programs’ productivity. Perhaps we need to rethink our language and home in on the actual goal of advertising evaluations: to infer a cause among multiple, overlapping, competing media.

So how does one make effective inferences amid such complexity? Generally, one must rely on the tried-and-true methodology of test-and-control. At its simplest level, this means that, across time, the behavior of those exposed to a piece of media is compared to a control. Analysts use this comparison to generate estimates of the increase in purchase/engagement likelihood, with more complex mathematical modeling allowing for multiple, overlapping media over time. This strategy provides marketers with an understanding of the particular impact of media or media combinations, and thus enables them to make intelligent spend decisions based on ad effectiveness.

Attribution appears as if it is the solution to multi-channel and multi-vendor digital initiatives, but it is not. The attribution camp is driven by a mercurial interest to maintain the status quo created when the Internet was not considered a true medium for brand advertising. It should instead be driven by the old-school interest of determining what media, or combination of efforts, caused an increase in the desired consumer behavior on behalf of the marketer. The Internet has matured and has rightfully taken its place among the other major media of the late 20th and early 21st century: television, magazine, newspapers, billboards, etc. Measures which ultimately can be used to drive more sales should be instituted, not games that allow vendors used by an online marketing organization to compete for spend.

Follow Media Armor (@mediaarmor) and (@adexchanger) on Twitter.

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. Liz, attribution is an idea. Like all ideas attached to technology, you have to start somewhere. For the longest time, marketers thought it as all about clicks, and in fact, last clicks, so you would buy tons of media to support that last click. This invariably led to putting more money into search marketing and lining Google’s coffers. But what attribution studies so far have revealed is that other channels have a role to play as introducers and influencers to those who are not “actively in market.” The key is to understand lift and then intelligently assigning credit. Where attribution is at today is only focused on assigning credit because 1) the data was there for the taking and 2) there are economic reasons for doing so as you alluded to. But in the long run, attribution is understanding how the conversion lifecycle/funnel works online and offline and optimizing on that.

    • Derek, I understand it’s an idea. However, saying that attribution actually tells you how the conversion funnel works is inherently incorrect. It’s a way to show what touched a consumer; it does not indicate whether your advertising had an impact on increasing consumer behavior.


  2. Attribution as described here seems to be something of a straw man. While many of the points ring true for last-touch models, or models where the value assignment methodology is driven by intuition rather than campaign data, better attribution modeling techniques exist.

    Test and control methods are certainly necessary, but as the number of dimensions that an advertiser can buy across multiplies, managing hundreds and thousands of A/B tests becomes unrealistic (not to mention expensive). This is where more data-driven attribution modeling can help.

    Attribution is a concept, and it can be executed poorly or well. Methods of attribution must be understood and validated. In and of itself attribution is neither good nor bad, and can’t be dismissed out of hand.

    • Mmmmm, my point is that attribution != success. Attribution says what touched someone, not whether it worked. Test/control may not be easy, but it certainly is the only correct way I can see to understand the proper impact of a media mix on consumer bheavior.

      • Test/Control is certainly valid for calculating lift, but to look at the complex interactions you discuss necessitates exponentially increasing numbers of A/B groups. The most valid test in the world doesn’t help if it can’t be implemented. Reducing A/B groups to single buys fails to control for the possibility of highly correlated exposures between groups.

        Whether or not attribution is purely a touchpoint log, or if it actually shows the influence of media is dependent upon the method used. To say that the entire concept of attribution is unable to be extended into causal analysis is a somewhat myopic view.