Decision Sciences Manager Jack Palmer contributed to this post.
Multi-touch attribution models measure consumer response to advertisements across the full user journey by assigning credit to every marketing touchpoint in the user’s path. Unlike traditional first- and last-click models, which only assign credit to one touchpoint in the path, multi-touch attribution allows for an omnichannel view that enables marketers to evaluate the true value of each marketing touchpoint at every stage of the conversion funnel, down to the tactical level.
Why should you use a multi-touch attribution model?
The consumer journey and buying landscape have become increasingly complex, and a user’s path to purchase is no longer direct enough to measure success based on one single click or action. There are more ways than ever to reach the consumer, especially when utilizing an omni-channel approach. Multi-touch attribution models make measuring all touchpoints in the conversion funnel possible and provide a holistic or true view into how each marketing tactic is influencing conversion.
What are the different multi-touch attribution models?
The linear model values each touchpoint as equal by evenly distributing credit across all touchpoints in the user’s path to conversion. The linear attribution model is the least sophisticated of the multi-channel attribution models as it only looks at frequency and does not take key factors, such as recency, into consideration. By considering each touchpoint as equal, you run the risk of overspending in underperforming or duplicate marketing channels. While the linear model is the least recommended multi-touch model, if utilized it is best suited for B2B environments with long sales cycles.
The time decay model gives more credit to the interactions that occur closest to the time of conversion. In this model, the earlier touchpoints in the user path receive less credit and conversion credit increases for every interaction after the first touchpoint. This channel tends to undervalue upper and mid-funnel tactics, which play an integral role in introducing and nurturing the consumer through the funnel. This model tends to overvalue channels that inherently occur prior to conversion, such as direct or email promotions.
The position-based model is a hybrid of the simple first- and last-click models, where the majority of credit is given to the first and last click in the user’s path to conversion. With this method, 40% of conversion credit is assigned to the first and last interaction, while the remaining 20% is distributed evenly among middle interactions. The position-based model undervalues mid-funnel tactics and channels by assuming that the first and last touchpoints are the most influential.
The user-defined model allows the marketer to customize how interactions should be valued. To create a model specific to your business needs can be beneficial; but this approach requires the organization to have a great understanding of their business and industry. This model utilizes a Google Analytics model as a baseline (time decay, position based) based on an analysis of which touchpoints are driving conversions for the business. The custom model will also take typical latency patterns and site engagement into consideration.
Unlike the subjective rule-based models, the algorithmic or data-driven models are more sophisticated machine learning models designed to accurately represent the consumer journey. These models measure response to advertising and campaigns at all points in the purchase funnel by collecting user-level data across multiple channels and devices. The output is a set of weights that are custom to each user’s activity based on the observed influence each touchpoint had on the consumer’s decision to convert. The algorithmic/data-driven models are regarded as more accurate as they address the user’s location, day of week, purchase history, recency, and frequency; these models are also continually retrained based on user-level results, ensuring accuracy in the outputs.
Determining the Right Model
What attribution window are we solving for? That is the key question of cohort analysis. Cohort analysis allows us to determine what the optimal conversion window should be by analyzing the historical behavior of different cohorts. It also helps illustrate what role each channel plays within the funnel by showing the latency to purchase from different purchases.
What user behavior is valued more than others? A micro-conversion is a signal of intent that can be utilized to assign weight to interactions. For example, this might be a person who downloads a white paper. When creating a Google Analytics custom model, these micro-conversion events are key to creating an accurate attribution view as they can be utilized to signal touchpoints that have a high likelihood of leading to conversion.
How influential is a touchpoint in the user journey? Incrementality analysis allows us to model the influence of touchpoints in converting users. For example, Google brand keyword campaigns may be reporting a high number of conversions based on a last-click attribution model, but when an incremental analysis is performed, those campaigns might have little actual incremental impact. In that case, we would adjust the model to give those campaigns less credit.
We hope this gives you a clearer picture of multi-touch attribution methods. We’d love to help! If you have any questions or would like to know how our Decision Sciences team can help you utilize attribution strategy to gain key insights into campaign effectiveness, drop us a line.