Free MMM data preparation template

What the spreadsheet contains

The Weekly data sheet includes week start, net revenue, paid-search spend and impressions, paid-social spend and impressions, a promotion flag and a price index. A formula totals the two paid-spend columns. The Data dictionary explains definitions, source handling and checks.

This is a preparation template. It does not fit an MMM or estimate channel ROI. The eight synthetic weeks demonstrate the format and are not enough to train a useful model.

Start with one row per week

1. Choose your outcome and document its definition. Use a business outcome such as consistently defined revenue from your own records.
2. Align sources to the same week boundary, reporting time zone, market and currency. The sample uses Monday-start weeks.
3. Keep spend and exposure fields separate. Include all relevant channels, adding columns and extending the total-spend formula.
4. Record promotions, price changes and other plausible demand drivers. Check that controls are appropriate for the causal question.
5. Preserve the raw exports separately. Reconcile totals, investigate duplicate weeks and distinguish verified zero activity from missing data.

Example: spend totals are not modelled returns

The first fictional week contains $10,000 search spend and $7,000 social spend, totaling $17,000. Its $80,000 revenue is a business outcome. Dividing revenue by spend would not identify the incremental return of either channel.

How much history do I need?

History requirements depend on variation, geography, channels and model complexity. Google's Meridian data guidance gives rules of thumb of two years of weekly data for geo models and three years for national models. These are starting points, not guarantees of a reliable result.

Can I load this directly into Meridian or Robyn?

Map the fields to the selected model's required schema first. A geo model can need geography and population; a reach-and-frequency model needs additional inputs. Validate dimensions, units and time coverage before fitting. The workbook is not a ready-to-run model configuration.

Use the Meridian and Robyn comparison to understand the workflow choices, then follow the MMM output guide. For the full method, read the MMM guide.

How do I extend the example?

Replace all sample rows, add each additional week in order and copy the Total paid spend formula down. Add new channel columns explicitly. Reconcile the extended data with the source exports before analysis. Keep extraction dates and changes to metric definitions with those source records.