Attribution modelling is a framework for analysing the various marketing touchpoints, channels and tactics and determining which receive the credit for a specific conversion. Using attribution models helps marketers to better understand which parts of their marketing efforts are having the most effect on conversions, therefore enabling better data-driven decision making which can transform organisations.

There are a number of different attribution models, with each attributing conversions to marketing channels differently. Let us introduce you to three of them…

Last-click attribution

Without proper tracking, Google Analytics uses last-click attribution to attribute conversions. With this model, all (100%) credit goes to the last touchpoint before a customer converted, therefore not taking into account awareness and nurture tactics which may have influenced the customer’s decision to purchase.

For example, applying this last-click attribution model to an education setting, the vast majority of conversions (applications) are attributed to remarketing and conversion tactics. No credit is given to those awareness and nurture tactics on paid search and display:

Linear attribution

On the other hand, a linear model gives credit to each touchpoint along the user’s journey equally, allowing a more even spread of attribution. For example, if there are 4 touchpoints, then each touchpoint would be given 25% of the credit, 5 touchpoints and each touchpoint would be given 20%, and so on…

However, should all touchpoints be given the same value? For example, does first entering the site via Google Search actually influence the likelihood of a conversion?

Markov Model

The Markov Model is arguably the most advanced attribution model that uses Markov chains to calculate the probability of a set sequence of events happening, essentially analysing the likelihood of each touchpoint influencing conversion.

In an education setting, we can calculate the sequence of touchpoints for all applicants and then calculate the effect of a certain touchpoint on the probability of an application.

Let’s observe the effect of removing the Facebook advertising touchpoint from an applicant journey:

By comparing the two images, you can see that having a Facebook advertising touchpoint doubled the probability of a conversion – the Markov model assigns a score to that channel based on this. We can use this process of removal to gain an understanding of how different media touchpoints affect conversions, allowing the importance and performance of a channel to come through, and data-driven decision making to be made.

A real Markov model would look like this:

Let’s stick with the simplified version!

Why is this important for you?

When reporting on return on advertising spend, certain tactics can receive an overinflated amount of credit in a last click attribution model, whilst others seem like they have contributed few conversions. You could therefore channel more money into the tactic that, on the face of it, is performing best, but that doesn’t give you the full picture.

A more complicated attribution model gives a more accurate return on advertising spend (ROAS) for each media channel, and this where the Markov models comes into its own. The model gives an idea of which media channels actually drove conversions and how they did it. You can then use this information to decide on future campaign strategy and media budget.

If you’re interested in exploring data-driven attribution that not only calculates ROAS, but analyses how paid media channels work together to drive conversions, get in touch with one of our data experts today.