MMM: Does your model account for diminishing returns?
Is the advertising being well managed in the model?
At lower levels of spend, advertising can deliver strong incremental gains because you’re reaching new audiences and increasing mental availability.
But as investment rises, each additional pound typically works a little less hard. You begin to reach the same people again, and while repeated exposure still has value, the uplift is usually smaller than the first impact. This is what we mean by diminishing returns. Each addtional pound still drives sales, but less efficiently than the pound before it.
A robust MMM should account for this reality. They might see patterns such as, “when you spend this much, sales tend to increase by this amount,” and, “when you spend more, the increase is smaller.” It should recognise that doubling spend doesn’t automatically double sales and that efficiency changes as scale increases.
If diminishing returns aren’t modelled properly, optimisation advice can become misleading. The model may assume every extra pound is equally effective, encouraging over-investment. Or it may apply poorly evidenced response curves that unfairly suppress certain channels.
For OOH, where spend often appears in larger bursts with limited variation, the shape of the response curve is especially important. If it’s based on weak or assumed data, ROI conclusions can quietly shift,influencing future budget decisions more than you might realise.
Has this model accounted for diminishing returns and are response curves based on real data variation?
YES!
Diminishing returns are modelled credibly.
Good - optimisation decisions will be more realistic.
To sense-check how it’s being applied, you can check the data used to build the diminishing return – is it reach or spend? Reach tends to offer a more realistic curve to work with due to differences in areas such as campaign formats, targeting, media costs and seasonality.
NO!
Curves are assumed or weakly evidenced.
Budget guidance may be distorted because it could be assuming each pound spent is equally effective.
This brings risk of over-investment and over-confidence in channels that have reached their potential. Work out why there are no curves (no data?) and seek to find a relevant source, or design future activity to add data to your model.
To understand more: