Attribution setup

Attribution settings are split across three pages, all under Insights → Settings → Attribution:

Page	What it covers
Core settings	Paths table, column mapping, model selection, label splits, scheduling
Optional tables	Spend data (for ROAS/CPA), performance data (for revenue coverage), and the credit output table
Run history	Log of all past runs with status, duration, and trigger type

Core settings

Paths table

This is the BigQuery table that contains the customer journey touchpoints. Each row should represent a single marketing touchpoint within a journey.

Field	Description
Dataset	BigQuery dataset containing the paths table
Table	Table name
Channel column	Column that identifies the marketing channel (e.g. `channel`, `source_group`)
Journey ID column	Column uniquely identifying each customer journey (default: `journey_id`)
Converted column	Boolean or 0/1 column indicating whether the journey converted (default: `converted`)
Revenue column	Revenue value per journey (default: `revenue`)
Date column	Column used as the time window filter — typically the journey start date (default: `journey_start_date`)

After saving, Interact validates all column names against your BigQuery table schema and highlights any that cannot be found.

Path processing

Setting	Description
Junk channels	Comma-separated channel values to exclude before processing (e.g. `Direct, Other, (none)`). These are removed from journey paths before the model runs.
Passthrough channels	Channels that don’t interrupt the journey logic (e.g. brand search that always appears). Kept in the path but treated differently in Markov transitions.
Deduplicate consecutive	Remove back-to-back touches from the same channel within a journey. Recommended for most setups.
Max journey length	Maximum number of touchpoints per journey. Longer journeys are truncated. Default: 15.
Min journey length	Minimum number of touchpoints required. Shorter journeys are excluded. Default: 1.

Attribution models

Enable one or both models:

Model	When to use
Markov (recommended)	Fast, reliable, good for operational use. Always enable this.
Shapley	More thorough game-theory approach. Enable when you need to compare or validate Markov results. Adds processing time.

Label splits

Label splits allow you to run attribution separately for each value of a segmentation dimension — for example, running separate models for NL and DE rather than blending all markets together. Enable Label splits and define one or more split columns. Each split column maps to:

A column in the paths table (required)
Optionally, a corresponding column in the spend table for filtering spend data per segment

When label splits are enabled, the dashboard shows a multi-select filter so you can view results by specific segment combinations.

Use label splits for any dimension where channels differ meaningfully across segments — market, brand, or product line are common examples.

Scheduling

Automatically run attribution on a schedule so results are always fresh.

Setting	Description
Enable scheduling	Toggle automatic runs on or off
Frequency	Daily or weekly
Weekday	For weekly runs, which day to run
Time	Run time in HH:MM (server time)
Advanced (cron)	Custom cron expression for fine-grained control

When scheduling is enabled, Interact queues runs for the standard preset windows (90d, 180d, 365d) at the scheduled time.

Optional tables

Spend data

Connect a BigQuery table containing channel spend by date to unlock ROAS and CPA metrics in the attribution dashboard.

Field	Description
Dataset / Table	Location of the spend table
Channel column	Column identifying the marketing channel — should match values in the paths table
Spend column	Column containing the spend amount (default: `spend`)
Date column	Column with the spend date (default: `date`)
Label column	(Optional) Column for label-based filtering when label splits are enabled

Performance data

Connect a business-system revenue table (e.g. from your e-commerce platform or CRM) to display coverage — what percentage of actual revenue is captured in your attribution paths.

Field	Description
Dataset / Table	Location of the performance/revenue table
Revenue column	Column with the revenue figure (default: `revenue`)
Date column	Date column for filtering to the selected window

Credit output table

If your pipeline pre-computes per-touchpoint attribution credit and writes it to BigQuery, configure this table to power the Journey Attribution page.

Field	Description
Dataset	Dataset containing the credit table (default: `src_attribution_models`)
Table	Credit table name (default: `fact_attribution_credit__user_id`)
Channel column	Channel column in the credit table (e.g. `channel`, `source_group`, `channel_group`)

The credit table is written by a separate pipeline process and is not produced by the interactive attribution runs. It enables the per-journey detail view but is not required for the main channel attribution dashboard.

Run history

The Runs page shows all past attribution runs with:

Status — Success, Failed, or Queued
Window — Date range that was processed
Label — Segment label (if label splits are used)
Duration — How long the run took
Trigger — Manual, Scheduled, or Preview
Error — Failure message when status is Failed

Use Run all presets to queue fresh runs for the standard 90d, 180d, and 365d windows. This is useful after changing the paths table or model settings.

BigQuery permissions

The service account used by your BigQuery connector needs the following permissions:

Permission	Why
`bigquery.tables.getData`	Reading the paths, spend, and performance tables
`bigquery.jobs.create`	Running queries to build transition matrices
`bigquery.tables.create` + `bigquery.tables.updateData`	Writing channel weights output (when enabled)

The standard roles/bigquery.dataEditor + roles/bigquery.jobUser roles cover all of these.

​Core settings

​Paths table

​Path processing

​Attribution models

​Label splits

​Scheduling

​Optional tables

​Spend data

​Performance data

​Credit output table

​Run history

​BigQuery permissions

Core settings

Paths table

Path processing

Attribution models

Label splits

Scheduling

Optional tables

Spend data

Performance data

Credit output table

Run history

BigQuery permissions