> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getinteract.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Media Mix Modeling

> Understand how your marketing channels contribute to business outcomes using Meridian-trained MMM models.

Media Mix Modeling (MMM) is a statistical approach to understanding which marketing channels are driving results — and how much. Instead of relying on last-click attribution, MMM looks at the relationship between historical spend and outcomes across all channels simultaneously, accounting for seasonality, baseline effects, and diminishing returns.

Interact uses [Google Meridian](https://developers.google.com/meridian) as its MMM engine. Your Python pipeline trains the model; Interact displays the results.

## What you can do

<CardGroup cols={2}>
  <Card title="Data Book" icon="table" href="/insights/mmm/data-book">
    Inspect the prepared dataset that was used to train the model — column definitions, time range, and channel spend series.
  </Card>

  <Card title="Model Results" icon="chart-bar" href="/insights/mmm/model-results">
    View channel contributions, response curves, predictive accuracy, and model health diagnostics for the latest (or any previous) model run.
  </Card>
</CardGroup>

## How it works

```
Your data warehouse (BigQuery)
    └─ raw marketing spend + KPI data

Python pipeline (Meridian)
    └─ reads prepared data from BigQuery
    └─ trains Meridian model
    └─ exports results to GCS  →  gs://turntwo-mmm/{label}/results/model_{timestamp}.json

Turntwo Interact
    └─ Data Book page  →  queries the prepared dataset from BigQuery directly
    └─ Model Results page  →  reads the latest artifact from GCS on page load
```

The app does not run Meridian — it only displays what your pipeline produces. This means you control the model training schedule, the data preparation logic, and the granularity of the output.

## Multi-market support

You can configure **multiple data books** — one per market or label (e.g. NL, DE, UK). Each data book has its own BigQuery table and its own GCS results folder. The UI shows a dropdown to switch between them.

## Before you start

You'll need:

* A BigQuery connector set up in **Settings → Connectors**
* A `turntwo-mmm` GCS bucket in your client's GCP project, with the correct IAM role on the service account
* At least one Data Book configured in **Insights → Settings → MMM**
* A trained Meridian model with results exported to the correct GCS path

<Card title="Set up MMM" icon="gear" href="/insights/mmm/setup">
  Step-by-step: bucket creation, IAM roles, and settings configuration.
</Card>