After discussing all the factors that you should take into consideration when creating your own attribution model, and even going step-by-step to create one, we leave you with the final instalment of Attribution in E-Commerce from our Online Marketing Rockstars Masterclass. In this article, we go through some guiding questions that help determine the right attribution model for you.

After all – despite what everyone says about the ideal attribution model, only you can decide what works best for you – your business needs, KPIs and resources. So let’s get started!


Factors to consider when determining the right attribution model for you

Algorithmic or deterministic

Have you already been collecting a lot of data concerning customer interactions on all your marketing channels – accurately? If yes, and you are confident of making use of an algorithm to evaluate all the data, then an algorithmic attribution model would make sense.

If you prefer to set the rules, then a deterministic attribution model makes sense for you. This is particularly common, as many companies have a predefined strategy, e.g. to expand market presence through affiliate partners, that the team follows.

Retrospective or in real-time?

Already collect accurate data and prefer to analyse and identify patterns from them? Then you would work best with retrospective attribution. This way, you can compare the effects of varying attribution models on the data from the last 30 days.

If you need to make quick and agile changes – and have the resources to – then you would need real-time attribution. With this, you can apply a newly configured model in the very moment, so all new conversions take on this model. You can then process and share results and make more advanced decisions.

One or several winners?

If you have predominantly CPA costs, then an attribution model with just one winner would suit your business best. This means regardless of the number of touchpoints, your attribution model determines a single winner.

However, if you work with many CPC and CPM-based campaigns, these costs are incurred anyway, and can only be adjusted when optimizing your campaigns.

On the other hand, to evaluate the involvement of certain channels independently from the others, it would make sense to work with attribution models with several winners.


If you only wish to have one winner, or if you always point out the correct net sales of a conversion, you don’t need to split sales in your attribution.

The possibility to share revenue comes when you wish to only show the proportionate sales per channel involved.


If you wish to have only one winner, or if you assign costs to marketing partners based on individual agreements, you don’t need to share costs in your attribution.

Cost-sharing works if you would like to assign costs to the sales. This is only possible for CPA campaigns, because CPC and CPM costs arise before the attribution.


In a nutshell


The five commandments of attribution

No matter what attribution model you end up creating, thou shalt remember the five commandments of attribution:

  1. Attribution abstracts the customer’s decision-making process.
  2. Attribution supports the search for the best mix of advertising channels.
  3. Attribution is dependent on accurate data and reliable technology.
  4. Attribution must be permanently monitored and optimized.
  5. Attribution is the most important instrument for the analysis and optimization of advertising activities.

We hope you’ve enjoyed the Attribution in E-Commerce series. If you need help, feel free to reach out to us!


Here is part 5 of the masterclass by COO Robert Schneider. The masterclass was held in German.


By loading the video, you agree to YouTube's privacy policy.
Learn more

Load video

Catch up on the rest of the Attribution in E-Commerce Series: Part 1 | Part 2 | Part 3 | Part 4