How does the MMM model attribute sales to each channel?

Learn how Morpheus uses Bayesian models to accurately determine the contribution of each marketing channel to overall sales performance.

Morpheus uses a Bayesian statistical model to estimate the contribution of each marketing channel to overall sales. The model starts with prior assumptions about how much each channel might influence sales, based on historical data or industry benchmarks. These assumptions are updated with actual data over time.

Through a process called Bayesian inference (a way of making statistical inferences in which the statistician assigns subjective probabilities to the distributions that could generate the data), the model calculates the probability of different sales outcomes given the media spend and performance of each channel. The model looks at correlations between marketing activities (like ad spend, impressions, or clicks) and sales, adjusting for factors such as seasonality and external events.

Let's see a real-world example:


For example, imagine we have data from channels like Google Ads, TV, Radio, Facebook, and external factors like Weather and Brand Interest, along with Total Sales. The model will evaluate how each channel correlates with the fluctuations in sales over time.

  • If Google Ads spend increases while sales also rise, the model calculates the likelihood that Google Ads contributed to that increase.
  • If TV campaigns run at the same time but sales don't increase as much, the model may infer a lower contribution from TV.
  • Weather and Brand Interest act as control variables to help the model distinguish between factors that can influence sales but are not directly tied to media spend.

The model iteratively refines its understanding of how each channel affects sales by updating the probabilities

In this way, the model doesn't just say that "Google Ads caused 30% of sales" but rather assigns a probability distribution to represent how likely each channel contributed to sales.

This probabilistic approach allows the model to reflect the inherent uncertainty in the data, making it adaptable as more data becomes available.

To get started with Morpheus, contact our team via website chat or email, and we’ll guide you through the process.

Request access to Morpheus