How to try out attribution models with Google Analytics
We often chat with companies who wonder how effective Last Click really is for them and whether another marketing attribution model might give them more bang for their buck. Unfortunately changing back and forth from different models tends to muddy your results, making it unclear how to interpret the metrics and leaving stakeholders to scratch their heads and start the guesswork.
Luckily there is a way for you to both have your cake and eat it!
We recommend using Google Analytics’ Model Comparison tool to play around with different attribution models without any deep and meaningful commitments; giving you insight into whether your business would actually derive value from making that kind of change. So once you’ve figured out which attribution models work best for your business, you can stick with it – ensuring a consistent and reliable way to measure your performance.
In our opinion, Last Click Non-Direct is often more than “good enough”, and using more sophisticated models don’t necessarily always add any value. Different models can be useful for measuring the performance of different channels – for instance, First Click is often more appropriate for paid social channels, as these tend to affect customers early in the journey. But allow us to list the basic attribution models that you should be aware of for this blog:
- Multi Touch – Gives credit for a transaction to multiple traffic channels throughout a customer’s journey.
- First Click – Gives credit for a transaction to the first click in a customer’s journey.
- Last Click – Gives credit for a transaction to the last click on a customer’s journey.
- Last Click Non-Direct: Gives credit for a transaction to the last click on a customer’s journey, unless the last click is a direct channel – in which case it credits the penultimate channel instead.
Fun fact: Although you may think that your default attribution model in GA is Last Click, it is usually (and a bit sneakily) set to Last Click Non-Direct!
This is actually a good thing!
When Google doesn’t know where traffic has come from, it categorises that traffic as Direct. Using traditional Last Click means that GA only considers your final click, even when it’s from an unknown traffic source! But If you don’t know where your traffic came from, you cannot accurately attribute credit. Luckily with the Last Click Non-Direct model set up, Google skips this unknown traffic and gives credit to the penultimate touchpoint instead!
Now that we have that cleared up. Let’s dive into a few ways to thoughtfully use GA for more effective attribution.
Step 1: Choose a meaningful timeframe to track
It may be beneficial for you to alter the tracking timeframe in GA. This is one of many weird and wonderful settings that isn’t terribly well documented or explained! The default tracking timeframe is 6 months, which means that anything that is Direct traffic gets automatically replaced with the penultimate interaction within a 6 month window.
Follow this pathway:
Admin > Tracking Info > Session Settings > Campaign timeout setting
Step 2: Evaluate Your Path Length Interactions
Now that you’ve established your preferred tracking timeframe you can look at the amount of interactions it takes for your customers to convert.
This is broken down in the Path Length table, which shows you the conversions that have taken place in the last few months. The Path Length in Interactions column specifically shows how many interactions it took them to convert. If a high number of your conversions take place in 3 or more interactions, then you are a good candidate for multi-touch attribution.
If however, the majority of your conversions happen in just 1 or 2 interactions, it’s worth knowing that many models will give you the same results for a lot of those transactions. This means it will be less impactful to change attribution models – no need for multi-touch attribution here!
Follow this pathway:
Conversions> Multi Channel Funnels > Path Length
Step 3: See how long it takes for customers to transact
Instead of considering the amount of interactions, you may care more about tracking the length of time (in days) it takes from a customer’s first interaction with your site to their final conversion.This is useful if you have a high amount of customers that interact & transact within the same day, as it allows you to optimise your customer journey to be accordingly persistent in encouraging them along your funnel.
Follow this pathway:
Conversions > Multi Channel Funnels > Timelag
Step 4: Implement an exclusion list for referrals
Google can often miscategorise channels, particularly when it comes to referrals and payment processors. This can happen when someone goes to transact and gets sent to a third party payment platform and then back to your website – Google sometimes doesn’t recognise the payment processor as such and credits it as a referral. This would cause those using Last Click to misattribute the transaction to the payment processor and any other information to be lost.
The upshot of this is that your traffic KPIs get distorted and become unreliable, and bad data leads to bad decisions!
In order to avoid this issue, simply implement an exclusion list. A good place to start is by adding the domains of your payment processors as well as your own domain to the referral exclusion list in GA
Google’s guide to implementing an exclusion list can be found here.
Step 5: Use the model comparison tool
Now that you’ve got all your ducks in a row, let’s get this party started! With the model comparison tool you can start playing around with different attribution models to see what might make a difference for your business.
Try looking at how the transactions spread out differently across channels of the two models you are comparing. So if you go from Last Click to Last Click Non-Direct for example you would lose loads of traffic previously categorised as Direct, Social traffic would go way up!
It’s also good to be aware that sometimes social channels won’t come out as well in Last Click because they are impression based and often rely on a person seeing an ad, before coming back and buying something later in the week. Therefore, if a huge amount of your spend is on social channels it may be wise to track First Click!
Step 6: Set up custom dimensions
Whilst you can play around with different attribution models using the model comparison tool, you won’t be able to actually implement a new model without first setting up your custom dimensions to track user journeys. This is a prerequisite for doing anything other than Last Click.
The custom dimensions that are key for attribution are
- Client ID
- Session ID
This can be done without too much hassle if your website uses shopify or google tag manager (we’ll even do it for you if you want!). If not, you’ll likely need a web developer to add the code to your website.
Before you can implement a new attribution model, you will need these custom dimensions in place and gathering data for the length of your typical sales cycle. This is because you will only be able to operate the model using this new data and therefore you need enough of it to make attribution meaningful.
If you have a longer sales cycle or higher AOV products that take weeks or even months for your visitors to convert, you will need to let this run for at least that length of time. But if you have a shorter sales cycle – days for example, you can quickly gather data across your users’ full lifecycles. If you’re not sure which category your business is in then see step 3 to determine how long it takes your visitors to convert!