fbpx Cookie Settings
AttributionData Science

Beyond Black Box Attribution, Part 1: Going Behind the Data Science Curtain

By January 12, 2016 No Comments

Marketing attribution is a relatively simple concept supported by relatively complex data science. The idea of attributing an action or conversion to a particular activity or source intuitively makes sense; it is, in fact, what pretty much every marketer worth his or her salt has been trying to do since the beginning of time. It’s the how that gets tricky.

In the early days of digital attribution, marketers focused on the last click. This wasn’t perfect, but understanding the point of conversion was helpful, and certainly an improvement on the best-guess days of old. As technology evolved and we began to apply more sophisticated models to marketing data, we began to account for all of the actions that led to that last click. We know that people don’t make decisions in a vacuum; each action is a result of a variety of exposures, interactions, and enticements.

This “multi-click” model – aka cross-channel attribution – is another step forward in understanding the how, what, and why of conversions. But it came with its own problems. Namely, no one understands how it works.

We call this black box attribution. An organization signs up with an attribution vendor, crosses its fingers, and hopes for the best. All of the “magic” seems to happen in a mysterious black box that, if you’re lucky, spits out some meaningful data.

We do things differently at Conversion Logic. We think it’s important to understand not only why attribution is important, but how it works. That doesn’t mean you need to be able to articulate the construction of each algorithm in detail. You should, however, know what makes your vendor’s model unique in order to get the most value from the solution.

For us, it boils down to two words: The Ensemble Method. The Ensemble Method is our not-so-secret sauce, an approach that uses multiple algorithms instead of just one. Instead of relying on a single model, which can be inaccurate and limiting, we blend individual algorithms together to achieve a prediction accuracy higher than any of the parts of the whole. It’s the wisdom of crowds, on an algorithmic scale.

You can read more about our methodology here. Forget black box attribution. Data science is a weird and wonderful thing, and we couldn’t be more excited to throw back the curtains and bring it into the light.

Leave a Reply