The Different Attribution Models and How To Choose One

For eCommerce brands, defining their Attribution model is key to optimise the several sources of their online sales but, what is an attribution model?

As defined by Google,

An attribution model is a rule or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths (the route that you would choose to convert into sales).

Deciding on which ones to use proves to be a difficult decision across organisations due to the lack of knowledge about the outcome from using each of them. Ahead of the next #LEARNbyVERB that will cover an introduction to PPC, our team of experts briefly explain and suggest the diverse range of models that can be used from a brand perspective, and which ones can be the right ones to use:

A concept image of attribution models

Isolated attribution

This method assigns the credit to a single channel. Businesses continue to measure a channel’s performance in isolation, without considering the true effect on performance.

Choosing this method can be dangerous as there is a lack of information about the rest of the customer’s journey. Customers usually engage with a brand from several marketing channels as opposed to a single one and likewise, they also can convert a sale to a different channel.

First click

The First click attribution assigns 100% of the credit to the customers first interaction. This first click conveys into investing at the start of the customer’s journey.

This attribution model is easy to put into place due to its low-level of complexity and investment. However, its focus is mainly on raising brand awareness instead of understanding the behaviour further down the marketing funnel and therefore, having a clear understanding of where the sale is coming from.

Last Click

The Last click method assigns 100% of the revenue credit to the customers final interaction with a brand. The advantages of this are that it offers a valuable insight on the channels that convert the most and it’s easy and affordable to implement.

On the negative side, the last click makes engagement at the beginning of the marketing funnel disregarded despite being important. This method only focuses on the final touchpoint a customer makes and fails to acknowledge the rest of the journey.

Linear Attribution

This model considers all touchpoints of a customer’s journey. It offers more information about a customer’s journey than single touch attribution and assigns credit evenly across multiple channels. Although, being linear it assumes that every interaction is weighted the same and therefore, it may invest in channels that actually don’t require that much focus. It is more complex and expensive to implement than the methods mentioned above.

U-shaped or Position-based

In this case, every touchpoint in a customer’s journey is credited, but not evenly. This model splits the journey up into three sections: Beginning (first), middle and end (last). First and last would normally receive 40% of credit each, leaving the remaining 20% to the touch points in the middle.

Your brand can assign different levels of priority to each section by choosing different weightings. This model allows businesses to better understand their sales cycles. It is much more sophisticated and expensive but also reliable.

Time Decay

This model attributes credit to the touch points closest in time to the sale. In this case, channels like your website or email would receive the most credit when other channels at the beginning of the marketing funnel, for instance, social media, would receive significantly less credit.

Cross-Channel, data-driven approach.

This would be our preferred option and most professional recommendation. This model attributes credits to each channel based on existing data. Therefore, subjective opinions on where to assign credit would not count as valid if the previous data reflects a different behaviour. The advantages are obvious; the weight is distributed across every touchpoint, it quantifies the impact of each channel and combines data sources. On the cons side, it is the hardest model to implement but easier to use.