If you haven’t used them before, the attribution models inside Google Analytics can be confusing, daunting or a mixture of both. So what the heck are they and why should you be using them?
It’s all about how you attribute the conversions on your website to your marketing channels.
Let's say someone clicks on the ad you’re running on Google, they browse around and then they leave your website. A day passes and they click through to your website from Twitter. Then another day passes before they click through from one of your email campaigns. This time they convert.
So the question is – which of these marketing channels should the conversion be attributed to?
Google? Twitter? Or your email?
We could argue, that without Google, they wouldn’t have visited our website in the first place and therefore they’d never convert. But we could also argue that clicking through from the email was vital for the conversion to take place. Or we could say that without Twitter they might not have connected with our brand and build the trust needed to convert.
They’re all compelling arguments, so which one is right?
It’s likely that we needed all of these touchpoints for the conversion to occur.
But that sucks! I want to stop spending on marketing that isn’t giving me results.
I totally get it. I do. That’s exactly what I want to do with my own marketing. I want to focus my efforts and budget on what’s working. So what’s the answer?
Yes, you guessed right, we’re going to use the attribution models inside Google Analytics. They allow you to adjust the way you give credit to your marketing channels based on where they sit within the path to conversion. Before I weigh in on which attribution models you should (and shouldn’t be using) we need to cover some basics, so here we go…
Which attribution model is used in the standard reports?
The first thing to point out is that the standard reports inside Google Analytics all use the same attribution model. This is the Last Non-Direct Click attribution model. What this means is that when you’re looking at the standard reports and you see a conversion (or conversion rate), for example in this report…
All of the credit for these conversions is given to the last known channel used to find your website.
So if we continue our example where someone came from Google, Twitter and then email before converting, then our reports would say that email provided 1 conversion, while Google and Twitter provided 0 conversions.
And to give one more example, if someone came from Google, then the following day remembered our URL and typed it into their browser (which would be a direct session) we’d see that Google provided 1 conversion and direct provided 0 conversions.
So just remember, the default reports will give 100% of the credit for the conversion to the last non-direct channel someone used to find your website.
Here’s what you need to begin using attribution...
In order to use the attribution reports, you’ll need to make sure you have goals or transactions reported into Google Analytics. If you don’t have goals already configured, then this is an absolute must (and not just for attribution), so take a moment to learn how to correctly configure goals.
You’ll also need to make sure that you’re using campaign tags for all of your inbound marketing and that you have Google AdWords linked to Google Analytics if you’re running any AdWords campaigns.
So do a quick double-check to make sure you’re covered. If you don’t have goals or ecommerce configured, that’s okay, but make sure you check them off your list You’ll then need to wait for accurate data in your reports before you can begin using the attribution models.
What are the default attribution models inside Google Analytics?
If you want to play with the attribution models inside Google Analytics, you’ll need to head to the Model Comparison Tool within the ‘Conversion’ reports under ‘Attribution’.
The Model Comparison Tool allows you to compare up to three attribution models to see how they impact your conversion figures.
Here we’re comparing the Last Non-Direct Click model to Loves Data's Model (more on that later). We can see that the Time Decay model reduces the credit being given to ‘Other’ (which is for custom marketing channels), and additional credit is given to ‘Direct’ and ‘Social Network’.
This allows us to see if we’re overvaluing or undervaluing particular touchpoints in comparison to what we see inside the standard reports.
Which attribution models can we compare?
There are seven default attribution models available within Google Analytics. You also have the option to create your own custom attribution models. (And if you’re using Google Analytics 360, then you have the option to use the Data-Driven model too.)
Watch this video for a rundown of the models (and I'll walk through them below too)...
This model gives all of the credit to the final touchpoint. It’s super important to flag that this isn’t the same as the attribution model used in the standard reports (which is Last Non-Direct Click). It’s a little bit different because it will give credit to people who come in directly to your website and convert.
Don’t get tripped up on this one – it’s fairly common for people to compare this model to the standard reports and since it’s a different model your data won’t be the same. Instead, you should use the next model...
Last Non-Direct Click
This is currently the attribution model that you’ll find in the standard reports.
If the final touchpoint was direct, then it will step back in the conversion path and look for a touchpoint that isn’t direct.
Last AdWords Click
This model will give all of the credit to the last touchpoint (just like the Last Interaction model), but if there is a paid click from Google AdWords in the conversion path, then AdWords will receive 100% of the credit for that particular conversion.
Here’s an example of a path someone could take before they convert…
They’ve gone from AdWords, direct and then email before converting. If we use the Last AdWords Click model, then all of the credit goes to AdWords.
And here’s one more example where someone didn’t have AdWords in their conversion path…
They went from Twitter, to email and email again. The Last AdWords Click model will give all of the credit to the very last touchpoint (since there is no AdWords touchpoint), so email receives the full credit for the conversion.
This model gives all of the credit to the first known touchpoint that lead to a conversion.
I’d recommend avoiding this attribution model for two reasons. The first reason is that you’re limited to the historical data available for the conversion (this is called the lookback window). But the real problem is that it completely ignores the importance of the channel that closes (or completes) the conversion. So I’d skip this one.
Next up, we have the Linear attribution model. This is a multi (or mixed) attribution model because it splits the conversion value across more than one touchpoint. The Linear model is simple to understand because it evenly divides the value across all of the touchpoints.
If you had a conversion worth $10 and there were 5 touchpoints in the conversion path, then each touchpoint would be given a value of $2.
The Time Decay model is one of my preferred default attribution models inside Google Analytics. Greater credit is given to the touchpoints that are closest to the conversion action being completed, but credit is still given to the preceding touchpoints.
I like to think of the value cascading backwards. It all starts to get a bit complicated, but the model is based on exponential decay which means the more recent the interaction, the more value it receives. Let's keep things simple and look at this example…
This shows us how the Time Decay model is working, with each earlier touchpoint receiving half the credit of the one prior.
Finally, we have the Position Based model where the majority of the credit is split between the first and last touchpoints. Here you can see how the model works…
The final touch point receives 40% of the credit, the middle touchpoints share 20% of the credit (divided evenly) and the first touchpoint receives 40% of the credit.
What attribution model should you use?
When you head into the Model Comparison Tool, I recommend starting with the Time Decay and Position Based models. These will provide you with attribution that isn’t tied to a single marketing touchpoint. You can then compare these to the Last Non-Direct Click model (from the standard reports).
Now that you’re using the tool, you can begin to look at the difference between the model you’ve applied and the Last Non-Direct Click model. This will allow you to see the change in conversions as a percentage.
Look for channels that you’re undervaluing and overvaluing. These will be flagged with red and green arrows. This can then be used as the starting point for testing how you prioritize and assign budgets to your marketing channels.
Let me run through the Model Comparison Tool with you in this video...
You can take things to the next level by creating a custom attribution model inside Google Analytics. If you want to fast-track your custom model, then you can begin with Loves Data’s Model which I’ve based on Time Decay model.
The model has an adjusted half-life of 14 days, instead of the default of 7 days, so that more credit is given to the most recent touchpoints. The default lookback window is set to 70 days which will include just over two months of historical data. Credit for impressions has been reduced by 50% and credit is increased for people who engage with content (based on page depth).
Sharing the credit for conversions between your marketing touchpoints allows you to understand how your channels are interacting and where you should focus your efforts.
You’ll find the Multi-Channel Funnels reports are helpful in understanding how people are finding and engaging with your website. I encourage your to support your use of the Model Comparison Tool with the Top Conversion Paths report – it visualizes the paths people take…
You’ll find attribution is a combination of art and science, so don’t feel like you need to get it right from the very start. Begin by looking for channels that are undervalued and overvalued and go from there.