There are many complaints around marketing mix modelling outputs, such as:
- It can’t explain small spends
- There is a big piece of base
- It is costly to split by segment
- It is not always easy to action
Here are a few ideas to help get the most out of the process:
1. Provide additional insights or benchmarks to use
The modeller has to use their judgement to best explain the marketing trends.
Typically this is done in the context of time series, when spending patterns are fairly regular there often is not enough variance to truly know the impact. i.e. if you are measuring the impact of CRM – would you really turn it off for a month to get a base line?
Instead what you can do is use findings from other data sources to help feed into this base line, which could include results from structured tests, research or other data sources.
The modeller should be able to include these to get a better model.
2. Provide the context over periods like COVID
COVID didn’t stop after the lockdowns were lifted. People’s behaviour changed, and they changed for a while.
If you are for example a gardening company, the impact of lockdowns may have led you to recruit higher volumes of a new profile of customer – less likely to return. This may impact your reorder rates and seasonality impacts over longer periods than you expect.
It is important to provide that context of what has happened at each stage over the period you are looking at.
Too many models get left at the bottom of draws, largely this is because they can be hard for a channel owner to action.
It is important to explain what changes you are hoping to make, and what metrics you need to drive that and work with the right owner to implement change within the marketing plan.
If you need help with Marketing mix modelling then get in touch today