If you are investing in a two or more different search ad platforms, a handful of mobile sites, a few websites for display ads, a couple video providers and retargeting Facebook ads and so let’s assume there are a total of 15 different digital ad channels (which is not a lot). The number of potential paths, considering that anyone consumer could see none, one, two or all 15 – in fact they could see the same ad more than once – that would produce 437 quadrillion paths that could happen. So the idea of somehow boiling it down (and keep in mind this is a small example) to the most relevant paths is silly.
As the options for integrated marketing channels continue to rise, there is a growing list of reasons to rethink your current attribution model and begin planning for a new method of validation. The facts you must consider are:
- consumers using multiple devices to do different things, some directly related to revenue and many indirectly
- clicks don’t equal revenue
- you must measure incremental sales
- attribution cannot be the sole method of marketing success and evaluation
What does that mean? Marketing scientists have been taking on the attribution problem for more than three decades now. Marketing and Media mix models were largely invented in the consumer package good space and have been a tool, sometimes valuable, that has spread to many other B2C industries. They were born at a time that was much simpler – you had TV, radio, print, out of home, price and promotions to handle. The better models included exogenous effects like the economy and weather. With the significant transition to new digital media and quicken pace of traditional media becoming digital, a new stream of data is now available.
In parallel, data scientists have been working on digital attribution for maybe five to eight years. The notion of the data scientist has only recently emerged (and now apparently the sexiest job of the 21st century). The popular description being one who is well versed in statistical and mathematical modeling but understands a wide variety of data structures (and the really good ones can write decent code). The data scientist’s job, again, popularly described, is to sift through tremendous amounts of data looking for valuable insights. This is an obvious shift from the marketing scientist who tends to approach a problem with a hypothesis in mind and evaluates data (structured and unstructured) to support or reject the hypothesis.
Attribution in the Marketing Mix
There is a place for both marketing and data science in the attribution of marketing spend to performance. By itself, marketing mix models are large in scale, pulling aggregate data from multiple sources, with a lofty but important goal (simplifying here) of establishing how budgets drive sales, calculating return on investment and informing the planning and budgeting process. In doing so, it is required that these methods understand what sales are driven by marketing vs. non-marketing influences, what sales would have happened in the absence of marketing (that’s your incremental sales) and needs to account for seasonal effects and larger category trends.
Whereas attribution models tap into large streams of data from individual’s interacting with different media in near real-time speed. They have an equally lofty goal (simplifying again) of building a dynamic understanding of how individual consumers use various media that leads them through a path to ultimately purchasing the brand. An important part of the attribution process is mapping the different types of digital marketing tactics used to specific objectives or parts of the purchase funnel. Then utilizing math and logic to attribute the influence that each tactic has on that path. This can inform spend priorities, timing and targeting.
So there is clear value individually for both marketing mix optimization and digital attribution. But the best scenario is when they work together. Utilizing methods that can model consumer behavior from the bottom-up alongside individual consumer click stream data creates a virtuous closed loop that is superior for understand and optimizing marketing. When the granular attribution model can inform the more expansive marketing mix model, you get more integrated and comprehensive understanding of how marketing works.
Within this idea of an Integrated Marketing Optimization we can find the best of both worlds. And many of the world’s most recognizable brands including Sprint, Hyatt, Nestle, Urban Outfitter, DeVry and even the Cleveland Indians practice this method of optimization with great result.