Wondering when to update your MMM model?
Here is a guide ๐
Predominantly, a Machine Learning model needs to be updated for two reasons:
1. Model Drift
2. Business Reasons
Let’s first breakdown scenarios where MMM models need updating.
โพ๏ธ Seasonality – Some brands in CPG/ FMCG space exhibit seasonality. For e.g., Packaged Juices may sell more during summer.
If your models’ last few data points were say of ‘Winter months’ then the same level of sales demand may not be applicable to the spring and summer months.
To factor this, MMM model needs to be updated.
โพ๏ธ Campaign Launch – Sometimes post model building, the brand might have launched a big campaign or perhaps the competitor might have launched a campaign. To capture this effect in the Model, the model may need to be updated.
โพ๏ธ Past Error Corrections – Often it comes to light that a campaign spend / TV Spends / Digital Spends etc. were either inflated or under stated due to some error.
Now the model has already been built but it does not mean that we keep the model as it is in light of this error.
In order for the model to capture the reality on the ground, the model needs to be updated.
โพ๏ธ Economic / Geo political and Covid factors
Economic and Geo political factors do affect the MMM models. The current Ukraine – Russia War is already causing high inflations. These changes will be reflected more in certain markets.
Similarly, Covid waves brings with it frequent lockdowns like situations and these too in return affect the economy.
To reflect the effect of these macroeconomic, geo- political and force majeure factors in the model, the model needs to be updated.
MMM models typically needs updating in the frequency of:
โช๏ธ Monthly
โช๏ธ Quarterly
โช๏ธ Half Yearly
How frequently the model needs to be updated depends on:
โพ๏ธ How quickly things have changed on the ground.
โพ๏ธ How quickly you want to capture the effect and react to these changes.
๐ Monthly – Generally companies in the non CPG / FMCG tend to prefer monthly level updates since the variables change that quickly and its effect on the outcome variable be it sales, CTR, lead conversion etc. could be reflected early. This is not to say CPG/ FMCG brands don’t / won’t require monthly updating.
๐ Quarterly – CPG/ FMCG brands typically opt for quarterly updates. This is because some of the initiatives of the brand do take a longer period of time to show effect on real ground.
Moreover, some of the data capture cycle too happens at a quarterly level. Having said this, if there is a frequent change in the market due to macroeconomic, geo political or pandemic, then perhaps the CPG /FMCG brands too can opt for monthly updates.
๐ Half Yearly – Updates this long drawn out is pretty rare. However, there are some CPG brands which may not see frequent changes in the market and hence opt for half yearly updates.
