Once your data is uploaded, the next step is Field Mapping. This is the translation layer where you teach MMM Pilot what your data actually means. To the computer, “Facebook Spend” and “Temperature” are just lists of numbers. You need to assign them specific roles so the model knows which one is a cost and which one is just context.
The AI Co-Pilot: Automated Categorization
You don’t have to start from scratch. MMM Pilot uses AI-powered automatic field categorization to scan your CSV and guess the role of each column.
- How it works: The system looks at column names (like “date” or “revenue”) and data patterns to automatically suggest the correct category.
- Your Job: You act as the Pilot. Review the AI’s suggestions and confirm they are correct before proceeding.
The 4 Key Roles
Every column in your dataset must fall into one of these specific categories for the Robyn framework to process it:
1. Dependent Variable This is the single metric you are trying to predict (e.g., Revenue or Conversions).
- Rule: You can only have one dependent variable per model run.
2. Date This defines your timeline. The system needs to identify the column that contains your YYYY-MM-DD data.
3. Paid Media (Spend & Exposure) These are the marketing channels you pay for. MMM Pilot supports paired media variables, which allows you to link a “Spend” column (Cost) with an “Exposure” column (Impressions or Clicks).
- Why Pair Them? Linking spend to impressions helps the model understand price fluctuations (like CPM changes) rather than just total cost, leading to more accurate results.
4. Organic & Context These are variables that influence sales but don’t involve direct media spending.
- Organic: Unpaid traffic, email newsletter sends, or social media followers.
- Context: External factors like competitors’ price changes, economic indexes, or holidays.
Why Correct Mapping Matters
If you categorize a “Paid Media” channel as “Organic,” the model won’t know to calculate a Return on Ad Spend (ROAS) for it. Proper categorization ensures that the Model Artifacts (the results and charts) correctly separate your paid efforts from your baseline performance.