This article is related to Attribution Modelling in Google Analytics. If you are brand new to attribution modelling then read this article first: Google Analytics Attribution Modeling – Beginners Guide

What is the Data-Driven Attribution model?

The data-driven attribution (DDA) model is an algorithmic attribution model.

What that means, it uses algorithms (predictive algorithms) to find and analyze statistically significant data from multiple data sources (Doubleclick Campaign Manager, Google Adwords, Google Search Console, YouTube, etc) and then assigns conversion credit to four most influential touchpoints in a conversion path, in the last 90 days prior to conversion.

These touchpoints are reported by the ‘Model Explorer Tool‘:

To learn more about Model Explorer (or Data-driven attribution model explorer), read this article: Understanding Model Explorer Tool in Google Analytics

In the context of Google Analytics, DDA model act as a baseline model.

What that means, you can use DDA model to create a new custom attribution model.

This new custom DDA model will distribute conversion credit to various touchpoints in a conversion path according to the DDA model and before the custom credit rules are applied.

The DDA model analyses both conversion and non-conversion path data. Conversion path data is the data from users who converted on your website. Non-conversion path data is the data from users who did not convert on your website.

This model refreshes (i.e. creates) once a week by updating the conversion credit weighting of marketing channels in the default channel grouping.

Why is the Data-Driven Attribution Model incrementally better?

Just like the ‘Proportional Multi-Touch Attribution Model‘, the data-driven attribution model assign credit for conversions to different marketing channels/touchpoints in proportion to their contribution to conversions.

The marketing channel/touchpoint which assists the most gets the maximum credit for conversion regardless of it being the first touch, last touch or middle touch.

All other channels/touches would get credit in proportion to their contribution to the conversion.

Since the assignment of credit for conversions is decided on the basis of the most recent conversion data and not on the basis of the position of the touchpoints, the attribution is data-driven and hence the model is known as data-driven attribution model.

The really cool thing about the data-driven model is that it automatically assign credit for conversions to different touchpoints based on your most recent conversion data.

What that means there is no need to assign arbitrary credit for conversions to different channels/touchpoints any more.

Note: The data-driven attribution model is valid only for a particular time period as this model automatically changes with the change in the conversion data.

 

Data-Driven Attribution Model eligibility checklist

Not every business is eligible to use an algorithmic attribution model like the DDA model.

There are not just one but several stringent requirements that need to be met as well as maintain before a business can use and benefit from data-driven attribution:

Requirement #1: GA 360 Access

The primary technical requirement to be eligible to use the DDA model is access to a Google Analytics 360/Premium account. If you do not have access to GA 360, then you can not use the DDA model.

Requirement #2: Immediate access to a high volume of high-quality data

Your DDA model is only as good as the data you feed to it.

If you feed it garbage, it will produce garbage. Garbage in garbage out. 

This is where most organizations fail miserably to benefit from data-driven attribution. They can afford to purchase GA 360. They can even afford to hire best analysts but cannot, in most cases, are able to build and maintain a high volume of high-quality data from multiple data sources.

As a result, the data-driven attribution insight they get is most likely to be flawed, misleading and sometimes downright dangerous to use.

As such, a traditional MMM model is not suitable for carrying out digital marketing mix modelling aka attribution modelling.

Requirement #3: Alignment of Goals and KPIs

Your chosen goals and KPIs must align across marketing channels and organizations.

If you measure success differently for different marketing channels then there will be no alignment of goals and KPIs and data-driven attribution model won’t work.

For example, if the main goal of your Facebook campaigns is to increase website sales then the main goal of your Twitter campaigns should also be to increase website sales.

So if you choose ‘ increasing twitter followers’ as the main goal of your twitter campaigns then it means there is no alignment between Facebook and Twitter marketing channels’ goals.

Requirement #4: Conversion tracking setup

Goal conversion tracking and enhanced ecommerce tracking set up in the GA Premium view for which you want to generate the DDA model. Without conversion data, Google Analytics will not be able to generate the DDA model for you, whether or not you are eligible to use the DDA model.

Requirement #5: Meet the minimum conversion threshold

The GA Premium view for which you want to generate the DDA model must meet the minimum conversion threshold for setting up a DDA model. Just because you have got conversion data does not automatically make you eligible for data-driven attribution analysis in GA.

To learn more about the minimum conversion threshold, read this article: How to set up data-driven attribution model in Google Analytics.

Requirement #6: Maintain a minimum conversion threshold

The GA Premium view for which you want to generate the DDA model must ‘maintain’ the minimum conversion threshold for setting up a DDA model.

Just because your GA view met the minimum conversion threshold requirement once, does not automatically make you eligible for ongoing data-driven attribution analysis in GA.

Requirement #7: Meet and maintain minimum conversion threshold for each conversion type

Your chosen GA view must meet and maintain the minimum conversion threshold for each conversion type.

Just because Google Analytics can generate a DDA model for your reporting view does not necessarily mean that it will also generate a DDA model for all individual conversions tracked through that view.

A DDA model is generated separately for each conversion type. Therefore, there is always a possibility that a DDA model is generated for some particular conversions and not for others.

Whenever you are using the DDA model and the model is not generated for a particular conversion then Google Analytics will display a warning message at the top of the attribution reports (i.e. model comparison, ROI analysis, model explorer reports) which reads:

“For one or more of your selected conversions, a data-driven model couldn’t be generated due to insufficient data at this time”:

 

Using Data-Driven Attribution Model to valuate organic search channel

Navigate to ‘Model comparison tool’ and compare ‘last interaction model’ with ‘last non-direct click’ and ‘Data-Driven’ model as shown below:

I selected the ‘last interaction’ model because it is the default model used in multi-channel funnel reports in GA.

I selected the ‘last non-direct click’ model because it is the default model used in non-multi-channel funnel reports in GA.

I selected ‘data-driven’ attribution model over the ‘Time decay’ model for two main reasons:

#1 Data-driven attribution model can analyze the data not only from my GA account but also from all those Google Accounts (like Doubleclick Campaign Manager, Google Adwords, etc) which are linked to my GA account.

#2 Data-driven attribution model assigns credit for conversions algorithmically, which I trust much more than manual conversion credits and/or the credits assigned via the time decay model.

Now look at the column named ‘% change in conversion (from last interaction)’ for ‘organic search’:

Form this report we can conclude that the % of change in conversion for organic search from the last interaction model to ‘data-driven’ attribution model is 22.66%.

What that means, if you use ‘data-driven attribution model (instead of last interaction model) to distribute credit for conversions to organic search, then the ‘organic search’ deserve 22.66% more credit for conversions.

In other words, ‘organic search’ is undervalued by 22.66% under the last-click attribution model (when this model is compared with the ‘data-driven’ attribution model).

So we can now conclude with confidence, that in this particular case, organic search is undervalued and is undervalued by (21.61 + 22.66)/2 = 22.135%

Note: You can download the data-driven attribution model into excel by clicking on the ‘Download the full model’ button at the top right of the ‘Model explorer tool’.