Getting Started With Causal Attribution
Go From Connected Data to Confident Budget Decisions
Before you can act on causal results, your workspace needs three things in place: connected advertising data, promoted measurement inputs, and an active scenario to measure against.
The five steps below take you through each one in order, from connecting your platforms to opening Attribution and reviewing your first results. Work through them once, and your workspace is ready to show what is driving growth and whether spend is on track.
1. Account for every dollar of spend
Your results are only as complete as the data behind them.
- Connect your platforms
- Go to Integrations and connect each advertising platform your workspace uses.
- Confirm every connection is live
Once connected, confirm that every connection is active and receiving current campaign data. A paused or stale connection leaves spend out of your totals, which shifts both your iROAS and your pacing against plan.
For example, if your TikTok connection stopped syncing two weeks ago, that spend won't appear in your results, and your overall efficiency will look better than it really is. Checking this now means the numbers you review later reflect your full media investment.
2. Make your results incremental
This step is what separates causal results from platform-reported credit.
Build and validate your measurement
Create and validate the Marketing Mix Models and experiments you want to use for incremental measurement.
Promote what you trust
Promote the assets that should contribute to Attribution and model calibration (the process of tuning your models with experiment evidence so they stay accurate over time).
Only promoted assets inform your results, so promote the ones you trust most. The stronger your measurement inputs, the more reliable your incremental revenue, incrementality factor (the share of platform-reported revenue that is truly incremental), and recommendations will be.
3. Set the plan you measure against
Pacing and recommendations need a benchmark to work from. Prepare the planning scenario that provides your planned spend and the benchmark used by causal recommendations.
What the active scenario enables
The Attribution Overview compares your actual activity against this active scenario, so you can see at a glance whether you are on pace, running ahead of plan, or leaving budget unspent.
Without an active scenario
You can still review performance, but pacing and Scale, Maintain, and Reduce recommendations will not have a plan to measure against.
4. Confirm you're viewing current data
Go to Attribution and select the date range you want to analyze.
Before you interpret anything, check the last updated indicator at the top of the page. It tells you when your data was last refreshed, so you know whether recent campaign changes are reflected in what you're seeing. For example, if you paused a campaign yesterday but the data was last refreshed three days ago, that change won't show up yet.
5. Start broad, then drill down
Get the big picture
Use the Overview for a quick read on:
- Headline spend
- Incremental revenue
- iROAS
- Pacing against plan
- Which channels are contributing the most real growth
Find the detail behind it
When a result raises a question, move to the Breakdown to inspect performance at channel, tactic (the strategy or ad type within a channel, such as prospecting or retargeting), campaign, and ad set level. This is where you find the specific activity behind a channel-level number and decide exactly what to adjust.
Keep results accurate
A few habits will keep your results accurate and your decisions sound:
- Keep platform integrations active and current so no spend is missing from your totals.
- Use comparable date ranges when reviewing changes, so you are measuring real shifts, not seasonal ones. For example, compare this November with last November rather than with October.
- Review platform-reported and incremental metrics together to see where platforms are over-crediting themselves.
- Confirm the active scenario before acting on pacing recommendations so every adjustment is measured against the right plan.
- Validate important changes with MMM and experiment evidence before making significant budget moves.
Not seeing any data yet? If your workspace shows no attribution data, check that your data pipeline is running, your measurement assets are promoted, and your permissions are configured correctly. Once these are in place, your results will begin to appear.
Frequently Asked Questions
What do I need in place before using Attribution?
Three things: connected advertising data, promoted measurement inputs (Marketing Mix Models and experiments), and an active scenario to measure against.
Why does a paused integration affect my results?
A paused or stale connection leaves spend out of your totals, which shifts your iROAS and your pacing against plan.
Why do I need to promote models and experiments?
Only promoted assets inform your Attribution results. Promoting your most trusted models and experiments is what makes your results incremental rather than based on platform-reported credit.
Can I use Attribution without an active scenario?
Yes, you can still review performance. However, pacing and Scale, Maintain, and Reduce recommendations need an active scenario to measure against.
How do I know my data is up to date?
Check the last updated indicator at the top of the Attribution page. It shows when your data was last refreshed.
When should I use the Overview versus the Breakdown?
Use the Overview for a quick read on headline performance and pacing. Move to the Breakdown when you need to find the specific tactics, campaigns, or ad sets behind a result.
Why should I compare similar date ranges?
Comparable date ranges help you measure real changes in performance rather than seasonal shifts.
Why am I not seeing any attribution data?
Check that your data pipeline is running, your measurement assets are promoted, and your permissions are configured correctly.
Do I need to repeat these steps regularly?
The setup steps are done once, but keep integrations active, measurement assets promoted, and your active scenario current so your results stay accurate over time.
Updated 30 days ago