18 Sep 2020

A guide to multi-touch attribution modelling

By Hillel Chemel, Head of Paid Media and Insights, Ole!Connect

Attribution

Ever since I started working in digital marketing I have been hearing people talk about attribution modelling. When I first started out the only way I knew to track conversions was by last-click attribution. It's the easiest way to attribute a sale or lead to be honest… it's also the laziest way. I became obsessed with finding out a better solution. There are many opinions on this topic but here are mine:

There is no ‘one way’ to do it:

There are multiple free attribution models you get on Google Analytics as well as one which I will get into on the paid version of GA. Having tested all the models and coming up with different results each time it was clear that each of the models needed to be tested in order to get the clearest picture. A time decay model is used when there is a longer sales cycle, however, so is the position-based attribution model. So, it's about testing which works for you which is my next point.

Testing

You can't just switch to using one of these models and let that dictate how your media budget is determined. Thorough testing is needed before making any big decision. Word to the wise, make sure you test off a big enough sample and that you take seasonality into account. Work with a data analyst/scientist to get the best most accurate results. However, why do it when a machine can do it for you… 

Data-Driven Model

According to Google “If you have enough information to use data-driven attribution, it can paint the clearest possible picture of success in your account.”. In my experience, it's pretty accurate and I’ll explain how it works:

  • Data-driven attribution takes all the sessions on your site into account.

  • By comparing the paths of customers who convert to the paths of customers who don't, Google identifies patterns among those interaction points that lead to conversions.

  • There may be certain interactions that have a higher probability of leading to a conversion, so the model then gives more credit to those valuable interactions on the customer's journey.

  • This means that you are seeing the truest value of that channel and its role in driving conversions.

If I have left you confused on what to do next then welcome to the world of digital marketing. Unfortunately, you are going to have to go down the road of testing and trying until you get it right, but when you get it right it will be sweeter. Good luck out there.