Attribution modeling explained

Attribution Modeling explained:

Since we are a Data Agency, we get many requests about Attribution Modeling. We agree, it’s definitely possible to make a lot of fancy models using data. In many cases they even give a better indication of the value of a sale when it comes to touchpoints. However, we don’t think it’s really necessary. First of all, the value of a sale is usually determined by both the timing and combination of touchpoints.

Does that mean that you can decrease your spending on a ‘low value’ touchpoint? We would say no, as you will likely change the aforementioned combination of touchpoints. Secondly, it’s highly dependent on what you, or the chosen model, values the most. Therefore: whichever department gives us the biggest bottle of wine will be crowned the winner.

Where is the value in Attribution Modelling?

So, you might be wondering, what’s the value of attribution modeling? We always focus on the differences between models and make sure we understand them. When a ‘last click model’ indicates that email is way more valuable than a ‘first click model’ tells us, it shows that email is probably the end of the chain of touchpoints. You should therefore, in that case, focus on sealing the deal with email. Does it then make sense to start an email campaign to gain new prospects? Probably not. Would you need to invest more in retargeting and reminder emails to make sure your customer finalizes the purchase? For sure!

How to use GA4 attribution modeling

Data should not be used to make more advanced models; it should be used in understanding what is actually happening and how that translates to business. GA4 now has an interesting feature to understand attribution modeling even better: you can compare the different models next to each other. Do not use this to assign the highest value to your specific domain, instead look at the difference and understand your domain’s position in the journey of a customer and how you can leverage that position.

After all, when you create an attribution model you could also leverage your position and address the highest value to the department that gives you the biggest bottle of wine. It’s just a new truth you’re creating that is based on the point of view you have chosen.

Want to read more data stories? Follow The Data Story on LinkedIn to stay up to date on our blogs and more!

Need some help?

2022-03-10MRF (1055)

Jeroen Bakker

“Data moet boekdelen spreken en de beste ondersteuning of input geven voor bedrijfsvraagstukken. Door data visueel te maken wordt het voor iedereen herkenbaar.”

More Data stories

Data stories

What is Server-side tagging in Google Tag Manager?

You’ve probably heard about Server-side tagging and might be wondering “What is it exactly?” and “How is it any different than the current Google Tag Manager setup?”. This blog will...
Data Maturity Models
Data stories

Why Bother with Data Maturity Models?

In a world saturated with data, the ability to make the best use of it, is not always straight forward and it is a defining factor in the competitive landscape...
Data value
Data stories

Unlocking Data Value with MDMA

Recent research by Salesforce shows that a lot of companies still have issues getting value out of their data. In our opinion a big part of a (partial) solution to...
Help starting Dataform
Data stories

Getting started with Dataform: tips and tricks to make you more confident

Now that we have decided to switch from Dataform to DBT, how do we get started? Starting to use a new tool always comes with some hurdles. We definitely came...
Screenshot 2023-09-13 at 11.21.48
Data stories

Choosing the right path: following the Dataform trail

In the dense forest that data science can often be, finding a way through can be cumbersome. Although this forest is not made up of decision trees, making choices is...
Mike the headless chicken
Data stories

You should filter out useless traffic, here’s how to!

Useless traffic, be it bot or developer traffic, can be a major issue for data analysts. Certain traffic can be considered useless if it is not representative of actual user...