Re-Examining Attribution

Attributing credit across a multitude of marketing efforts is one of those sticky problems in digital analytics that seems to generate a whole lot of controversy. This is a topic that comes up with nearly all of my clients and is one that both Eric T. Peterson and I have been researching and writing about for some time now. My latest findings on attribution will be published in a whitepaper sponsored by Teradata Aster, titled, Attribution Methods and Models: A Marketer’s Framework, but you can tune in to our webcast on January 16th, to get the high notes.

While some pundits will argue that attribution is not worth the trouble and that all attribution models are flawed, others contend that attribution simply requires a healthy dose of marketing science, which will enable marketer’s to reap benefits tenfold. At the risk of opening up a whole can of Marketing Attribution worms, I’ll offer my Marketer’s Framework for Attribution, which is a pragmatic approach to organizing, analyzing, and optimizing your marketing mix using data. But first, let’s define marketing attribution:

Analytics Demystified defines Marketing Attribution as:

The process of quantifying the impact of multiple marketing exposures and touchpoints preceding a desired outcome.

The first question that you need to ask yourself is whether or not you really even need to include attribution in your analytical mix of tools, tricks, and technologies. I offer this as a starting point because attribution isn’t easy and if you don’t really need it, then you can save yourself a whole lot of headaches by short-cutting the process and offering a data-informed validation of why you don’t want to mess with attribution.

The approach I offer is shamelessly ripped-off from Derek Tangren of Adobe, who blogged; Do we really need an advanced attribution marketing model? Derek encourages his readers to answer this question by looking at their existing data to determine what percentage of orders occur on a user’s first visit to your website vs. those that occur on multiple visits. I bastardized Derek’s idea and applied it to help marketers understand how many visits typically precede a conversion event. While Derek offers a way to do this using Adobe Omniture, I’ve created a custom report within Google Analytics that does virtually the same thing. I call it the Attribution Litmus Test.

My version is a quick sanity check for those of you running Google Analytics to determine the number of conversions that occur on the first visit versus those that occur on subsequent visits. To use this, you must have your conversion events tagged as Goals within Google Analytics (which you should be doing anyway!). If you’d like to run the Attribution Litmus Test on your own data within Google Analytics, you can add the Custom Report to your GA account by following this link: http://bit.ly/Attribution_litmus_test. Remember that you must have goals set up in Google Analytics for this report to generate properly.

So now that you’ve determined that Attribution is a worthwhile endeavor to pursue for your organization, let’s dive into the Framework. According to a study conducted by eConsultancy, only 19% of Marketers have a framework for analyzing the customer journey across online and offline touch points. Yet, the reality of consumer behavior today illustrates that multi-channel marketing exposures and multiple digital touch points are commonplace. As such, Marketers need a method for understanding their cross-channel customers in a systematic and reproducible way.

Step 1: Identify Your Data Sources

The first step in utilizing an Attribution Framework is to identify and input your data sources. Because advanced attribution requires understanding marketing effectiveness across all channels, it means that you must acquire data from each channel that potentially impacts the customer path to purchase. Typical digital channels may include: display advertising, search, email, affiliates, social media, and website activity.

Step 2: Sequence Your Time Frame

All attribution models must consider time to understand which marketing exposures occurred first, and also to discern the latent impact of exposure across channels. This requires that organizations sequence their data. While numerous data formats will likely go into the model, we’ve seen the greatest success when attribution data is stored and aggregated within a relational database.

Step 3: Apply Attribution Models

The actual attribution models will determine how you look at your data and make determinations about which marketing channels, campaigns, and touch points are effective in the context of your entire marketing mix. There are five models that are commonly used in the attribution world: First Click, Last Click, Uniform, Weighted, Exponential. To learn more about these models, tune into the webcast where I explain each in more detail.

Step 4: Conduct Statistical Analysis

After the data has been prepped, sequenced, and cleansed; this is typically where Data Scientists conduct general queries, apply business logic, and run what-if analyses against the model. At agencies that specialize in attribution modeling like Razorfish, they have an advanced analytics team comprised of data scientists that attack the data. They’re looking for correlations to identify if users are exposed to marketing assets A>B>C, are they likely to take action D?

Step 5: Optimize Marketing Mix

Of course, the ultimate goal in utilizing an attribution framework is to make decisions that impact your marketing efforts. These decisions can be strategic such as: deciding to invest in a new social media channel; discontinuing use of a non-performing affiliate partner; or reallocating budget to highly successful channels. But an attribution model can also play a major role in making daily life marketing decisions such as: which keywords to bid on during a specific campaign; who should receive an email promotion; or where to place that out of home billboard to attract the most attention.

In conclusion, Marketing Attribution continues to be an Achilles’ heel to many marketers. But, the good news is that approaching attribution with the right toolset and a framework for solving the attribution riddle is definitely the way to go. Throughout my latest research, I talked with companies like Barnes & Noble, LinkedIn, and the Gilt Groupe to learn how they’re using and applying Marketing Attribution models. I’ve also had the good fortune to demo some of the latest attribution tools from industry leading vendors like Teradata Aster and Visual IQ. Through this research, I learned that there is some truly innovative work going on with regard to attribution, but there is no single best way to do it. I’d love to hear how you’re solving for attribution. Please shoot me a note, tune into our webcast, or comment on how you’re re-examining attribution.