What Actually Makes MMM Predictable
Some people imply that marketing is too unpredictable. Some even go as far as to claim that is like predicting movement of atoms (Brownian motion).
I disagree.
Marketing and MMM are by an large predictable.
📌 What Actually Makes MMM Predictable?
At its core, MMM works because human behavior is not random.
As someone rightly said “We are suckers of Habit”.
📌 User Habit:
People don’t wake up every day and reinvent their choices. They predominantly go with:
– Same brands
– Same stores
– Same time of purchase
This creates structure in demand, which MMM can capture.
Habit overall reduces randomness.
📌 Seasonality
A seasonal pattern is a fluctuation which occurs at regular time intervals. These time intervals are predictable.
These again make MMM predictable.
But too much seasonality can do harm too. We built a MMM model for a tax filling company in USA. The pattern was 90% of the users always filled their taxes in the eleventh hour (March – April).
No matter how much media spends you do in Dec – Feb, one simply could not move the needle much. If things are this predictable, MMM will have it way too easy and the suggestions obvious.
📌 Festivals and Cultural Events
They are calendar anchored demand explosions. Examples include: Thanksgiving, Diwali, Black Friday, Christmas, Ramadan, Chinese New Year etc.
If one specifies when these events occur in the model, the model can capture it effectively. We use a special technique called RBF to capture seasonality and Festivals effect.
📌 Paycheck Effect
We have observed that for certain D2C brands, the sales sometimes peaks in the beginning of the month and kind of plateaus at the end of the month.
Assuming people don’t rely on credit cards too much. This goes to show that people (Especially salaried), spend more in the initial days post getting their paycheck.
📌 Distribution Effect
We have built a lot of MMM models for CPG brands. One pattern we have seen is – distribution is king. A mediocre brand with deep distribution will always beat a ‘better’ brand with lesser distribution.
We have even seen a competitor benefiting from ads of a rival, simply because people wanted the item in that category but simply could not find the store to buy it.
This phenomenon is applicable to the digital world too.
But distribution makes MMM predictable.
📌 Brand Equity
For established CPG brands (>40 years), base demand is always predictable.
📌 Competition
As strange as it may sound, competition brings in predictability to the category as a whole.
If there a monopoly, that brand can jack up prices or bring it down based on its wishes. But if there is duopoly or multiple rivals, the price set by each of them doesn’t waver too much. This is because of game theory.
Marketing and MMM is not unpredictable. One just has to understand marketing dynamics and the structure patterns. Most importantly know how to embed these information into the MMM model.
