Attribution modelling in marketing
Over the last 12 months I seem to be having conversations about attribution modelling on a more regular basis, so the need to write about it seemed to be inevitable. The reason for the increase in discussion is simple, more people are understanding that a user does not just simply click and convert, they shop around, check out your competitors, research and will come to your site multiple times before actually converting.
This means that the advertiser needs to change their strategy to ensure they are measuring throughout the consideration process to maximise the overall ROI and ensure they are utilising the best data driven attribution methods.
Despite these increasingly frequent conversations, I still hear companies saying, “We use Google Analytics last click as our source of truth”. Now unfortunately, this in itself is incorrect and a common misconception. Google Analytics uses a native last non-direct click as its reporting model, not last click.
So, what does this mean exactly? Well, it means exactly what’s in the title, the credit for the conversion is given to the last click that is not a direct click (someone going directly to your website). In this model, a direct traffic conversion will only be recorded if a user comes straight to your site and converts without going through any other channel in the last 30 days.
In the example below the user comes to the site 5 times before they convert through a direct click. In analytics, the credit for this conversion would go to organic as this was the last click before a direct click. The second example would be credited to paid and the third email.
Organic >> Paid >> Display >> Organic >> Direct
Paid >> Social >> Paid
Organic >> Organic >> Paid >> Social >> Direct >> Email
So, I ask the question, why is this the most accurate representation for the performance of your online marketing? It’s certainly the easiest, but if you are looking to grow or have a long lead time then it is probably not the most appropriate. Your business’ individual aims should dictate which model you use.
In this article I will highlight and explain some conversion model examples alongside their different benefits and use cases.
First click attribution gives all the credit for the conversion to the first click in the conversion path. As a standard, this is set to a 30-day conversion window but can be adjusted up to 90 days or down to a single day depending on your business type and lead time.
This model is great for aggressive growth targets as you are focusing on users at the top of the purchase funnel and the channels that drive new traffic and brand awareness.
Last touch attribution models
Last touch attribution, which is different to native conversion modelling, gives all the credit to the channel at the end of the purchase cycle regardless of other interactions. In general, for larger business you will find that a lot of conversions are attributed to direct implying that your brand is responsible for all the revenue. This may not be the most accurate interpretation, since many users may not have come to your website direct had they not been exposed to your brand at other touchpoints (e.g., Facebook Ads).
This model is associated with conservative growth focusing only on the final engagement channels such as remarketing, brand and email. It capitalises on existing brand authority converting existing customers.
Linear attribution models
Linear attribution models credit the conversion evenly across all interactions through the conversion path.
This is the middle ground, it allows you to credit all the touch points, engage new customers while ensuring customers in the pipeline are converting, drive brand awareness but also capitalise on the existing brand presence.
Position based attribution
Position based attribution modelling gives 40% of the credit to the first touchpoint, 40% to the end of the funnel, and then divides the rest evenly across all other interactions.
The model is great for growth while also allowing for the final channel to be credited for reengaging. It is not kind on nurturing channels such as social but if you are looking to sustainably grow it is the model to pick.
Time decay attribution
Time decay attribution assigns cascading credit to clicks and gives the majority of credit to the last click leading to a conversion.
This model is skewed towards the last click and therefore is more associated with conservative growth. It does recognise that all interactions in the conversion path have value but deems the last channel to be of the greatest value.
Choosing the best data driven attribution model for your business
So, what is the best model for you? If you are looking for aggressive growth then first click is the charm, if you are looking for conservative growth and just maintaining your customer base then last click & last non-direct click is the way forward. If you would like somewhere in between one of the other models is for you, depending on how aggressive you would like to be. If none of these are relevant you can build your own and create a custom view for this.
Whatever model you choose it should be a considered choice that considers your marketing needs and objectives. Do not just use the native modelling because it’s there. In the long term, this is not be the best strategy to grow your business and you are likely to discount channels without really understanding how they impact your online performance.
If you’d like some help deciding the best way to track and attribute value to your marketing channels, you can read more about Alpha Digital’s analytics & custom tracking solutions.