MMM: Has the model applied an appropriate decay rate?
Is the advertising being well managed in the model?
Advertising effects fade, but they don’t disappear instantly.
Decay rates ( also known as adstock) tell the model how long and advertising channels impact persists. The application of decay rates can drastically change its ROI. Too short, long-term effects are lost. If too long, its ROI can be inflated.
The model is trying to match chanegs in sales to changes in advertising activity. When there is no advertising data, the model looks for other data to explain these changes, such as seasonality, prices, promotions, otherwise it will count them as part of the baseline (sales that would happen regardless).
A decay rate, ultimately decided by the modeller using a range of experience and statistics to determine the most likely rate, adds a tail to the media spend to help the model determine a match between sales and activity.
Outdoor advertising often builds memory over months. A 3-day decay assumption would drastically undervalue it where sales could be misattributed or be swallowed by the baseline.
Does my model include decay rates?
YES!
Decay rates were tested and make intuitive sense.
Good - delayed impact is being captured.
You can further explore this by asking how the decay rates were chosen and asking for a visual to help intuitively assess the rate.
NO!
Decay rates weren’t tested or feel unrealistic.
Longer-term impact may be misattributed or absorbed into baseline.
Ask to re-run the model with decay rates and run sensitivty tests with varying half-life tests. A typical brand-building channel will have half-life rates of 4-8 weeks.
Learn more about MMM: