A Deep dive into Attribution Maths — Part 1
We have got you covered.
Each attribution model has its own approach, nuances and underlying maths. In this article, we’ll walk through some of the most popular ones, like First Touch, Last Touch, Linear, and more advanced models like Shapley, Markov chains and Media Mix models.
By understanding the math behind these methods, you will get a clearer picture of what’s moving the needle in your campaigns. You will also be able to make the best decisions for your organization. Whether you’re a marketer diving into data or just curious about how companies make these decisions, this comprehensive guide will break down each model’s fundamentals.
All attribution models can be classified into four large categories based on the methodology used to arrive at the results.
These are Heuristics , Algorithmic , Data Driven and Controlled Experimentation.
In this series of articles we will explain all these models starting with four Heuristics Models in this article.
Let’s dive into the math behind marketing!
First Touch Attribution
The first touch attribution model assigns all credit for a conversion to the first touchpoint in the customer journey. This model works well when initial exposure plays a critical role in conversion, such as in brand awareness campaigns. However, it lacks nuance, as it disregards all subsequent interactions that may also influence the purchase decision
The Maths
First touch attribution is straightforward, requiring no advanced math. In a dataset, you identify the initial interaction and assign it a conversion weight of 100%. This simplicity is ideal for campaigns focused on top-of-funnel activities.
Last Touch Attribution
The Logic
Last touch attribution gives all credit to the final touchpoint before a customer converts. This model is commonly used because of its simplicity and emphasis on closing tactics, like retargeting ads. However, like First Touch, it oversimplifies the customer journey by ignoring early engagement that might have led the customer to the final stage.
The Maths
This model requires isolating the final touchpoint in each customer’s journey and assigning it full conversion credit. Last touch works well for evaluating end-of-funnel tactics, such as retargeting ads or checkout offers.
Linear Attribution
The Logic
Linear attribution divides the conversion credit equally across all touch points in the customer’s journey. This model is used to maintain neutrality, acknowledging each step without weighting any one touchpoint more than another. Linear attribution works well for customer journeys where multiple touch points have an equal impact on conversion.
The Maths
For each conversion, assign an equal fraction of credit to each touchpoint. If there are four touch points, each one receives 25% of the credit.
Time Decay Attribution
The Logic
Time decay attribution assigns more credit to interactions closer to the conversion event. This model recognises that touch points that occur near the conversion are likely more influential. Time decay is useful in shorter sales cycles or when recent interactions have greater relevance to the final decision.
The Maths
Time decay usually uses an exponential decay function, where each touchpoint’s credit diminishes over time based on its distance from the conversion point. The formula can look like this:
where
t is the time span between the channel and conversion
λ is a decay factor. Higher values of λ mean steeper declines, giving more recent touch points higher credit One way to understand λ is in terms of half-life. If the credit of a touchpoint decreases to half its value after time - T.
then T is the half-life and λ = ln(2)/T
These four attribution models are examples of Heuristic Models. Heuristic Models use an established rule to attribute conversions among different channels. These rules are often based on sound logical reasoning and hence the results are usually a close approximation of how marketing actually works. However heuristics can not solve for the complex interplay between different channels.
This does not mean that Heuristics models are inferior or unusable.
The choice of an attribution model depends on a company’s objectives, customer journey complexity, and data availability. Often heuristics are the right choice. Simpler models like First and Last Touch suit single-channel or top/bottom-funnel analysis.
Their most important benefit is that the inherent simplicity of these models means that Growth teams can easily digest the results and take actions on them.
In our next article we will focus on Algorithmic Models. Algorithmic models like Markov and Shapley take attribution a step further by capturing the dynamic and collaborative nature of multiple touch points, providing nuanced insights that can guide more effective marketing strategies in multi channel setups.
