> ## 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.

# Multi-touch attribution

> Understand how your marketing channels work together to drive conversions using Markov chain and Shapley value attribution models.

Multi-touch attribution (MTA) credits conversions to the marketing channels that appeared in the customer journey — not just the last touchpoint. Interact runs **Markov chain** and **Shapley value** attribution on your BigQuery customer journey data, and presents the results in a filterable, date-range-based dashboard.

## How it works

Attribution in Interact processes your customer journey paths table in BigQuery and runs statistical models to determine how much credit each channel deserves.

```
BigQuery paths table
(one row per touchpoint)
         ↓
Path pre-processing
(deduplication, junk channel filtering, length limits)
         ↓
Markov chain model    +    Shapley values model
(removal effects)          (coalition values)
         ↓
Attribution credit per channel
(revenue, conversions, credit share)
```

Results are cached per date window and label combination. Subsequent loads of the same window are instant — no recomputation needed.

## Attribution models

<CardGroup cols={2}>
  <Card title="Markov chain" icon="arrows-split-up-and-left">
    Uses transition probabilities between channels to calculate each channel's **removal effect** — how much conversion rate drops when that channel is removed from all paths. Channels with high removal effects are structurally important to the journey.
  </Card>

  <Card title="Shapley values" icon="game-board">
    A game-theory approach that fairly distributes credit across all possible channel orderings. More computationally intensive, but captures synergies between channels. Enable in settings when deeper analysis is needed.
  </Card>
</CardGroup>

**Last-touch attribution** is also always available as a reference baseline.

## Key concepts

| Concept            | Description                                                                                                                          |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------------------ |
| **Removal effect** | How much conversion rate falls when a channel is removed from all journeys. The primary Markov metric.                               |
| **Credit share**   | Fraction of total conversion credit attributed to a channel (0–1).                                                                   |
| **Date window**    | The time range over which journeys are analysed. Only journeys starting in this window are included.                                 |
| **Label splits**   | Segment attribution by dimension (e.g. country, brand). Run the model separately per segment.                                        |
| **Junk channels**  | Channels excluded from path analysis (e.g. "Direct", "Other"). They are removed from journeys before processing.                     |
| **Coverage**       | Percentage of journeys where at least one touchpoint matches a known channel. Low coverage suggests a data or channel mapping issue. |

## Pages

<CardGroup cols={2}>
  <Card title="Channel Attribution" icon="chart-bar" href="/insights/attribution/channel-attribution">
    Main dashboard — removal effects, attributed revenue, and Sankey journey flows across a date window.
  </Card>

  <Card title="Journey Attribution" icon="route" href="/insights/attribution/journeys">
    Per-journey and per-touchpoint credit view, powered by a pre-computed credit table.
  </Card>

  <Card title="Setup" icon="gear" href="/insights/attribution/setup">
    Configure the paths table, column mapping, models, spend data, and scheduling.
  </Card>
</CardGroup>