Overview
Compass recommends budget changes based on what the MMM model estimates each channel is contributing to revenue. Evidence Links surfaces the measurement data behind those recommendations so you can see why Compass is suggesting a change, not just what it is suggesting.
During initial setup, select Generate Evidence Links from Next Recommended Actions to compare the current model run’s recommendations with eligible completed incrementality tests.
After Evidence Links are generated, the setup button is no longer displayed. When Compass matches an eligible test to a subcategory, an evidence badge appears beside that row in the Next Recommended Actions table.
The badge indicates whether the test supports or contradicts the direction of the MMM recommendation. Select the badge to open the Compass Confidence modal and review the supporting evidence.
The Evidence Badges
Two badge states are possible on an Actions row:
Badge | What It Means | When It Appears |
Verified | A completed incrementality test measured this subcategory, and its result aligns with the MMM’s read. | The subcategory has a completed, non-stale geolift or incrementality test whose result supports the model’s direction. |
Contradicted | A completed incrementality test measured this subcategory, but its result conflicts with the MMM’s direction. | The subcategory has a completed test whose measured iROAS points in a different direction than the model recommends. |
Rows without a badge have not been measured by a completed incrementality test. The MMM’s recommendation still stands for those rows, but there is no external test result to compare it against.
The Compass Confidence Modal
Clicking a Verified or Contradicted badge opens the Compass Confidence modal for that subcategory.
The modal shows:
The badge and a plain-language summary of what it means for this subcategory
The name of the incrementality test that measured it
A side-by-side comparison of the MMM efficiency (iROAS) and the Test iROAS
The date the test was completed
At the bottom of the modal, select Open Test Results to view the full results for that test.
Reading the Comparison
The modal shows two numbers side by side: MMM efficiency and Test iROAS. These measure the same thing from two different vantage points.
MMM efficiency
MMM efficiency is the model’s estimated incremental return on ad spend for this subcategory, derived from the marketing mix model run.
Test iROAS
Test iROAS is the incremental return on ad spend measured directly by the geolift or incrementality test.
When these two numbers are close, the model and the test are telling the same story. The Verified badge reflects that agreement.
When they diverge significantly in direction, the Contradicted badge reflects that conflict.
The modal also notes that the MMM reached its estimate independently: the test was not used as a prior in the model. This matters because it means the agreement—or disagreement—is a genuine external check, not a case of the model echoing data it already learned from.
What to Do With Each Badge
Verified subcategory
The model’s recommendation is backed by a real-world experiment. You can act on it with higher confidence.
The test date shown in the modal tells you how recent that validation is.
Contradicted subcategory
The model is recommending a direction that a test measured differently.
Before acting on the recommendation, review the full test results by selecting Open Test Results. This can help you understand the discrepancy.
Common reasons for a discrepancy include a test that covered a different:
Time period
Geographic area
Audience
No badge
No completed incrementality test overlaps with this subcategory.
The recommendation is based on the MMM alone. You can still act on it, but you do not have an independent experiment to cross-reference.
Note: A Contradicted badge does not mean the recommendation is wrong. It means the model and a test disagreed, and that disagreement is worth understanding before making a budget change.
Frequently Asked Questions
Why does a row have no badge even though we have run tests?
A test must cover the specific subcategory shown in the Actions table to generate a badge for it.
If your test targeted a broader campaign group, a different geographic area, or a different time window than what Compass maps to that subcategory, it will not appear as an evidence link for that row.
Can a Contradicted badge cause Compass to reverse its recommendation?
No. The badge is informational.
Compass still surfaces the recommendation based on the MMM’s output. The Contradicted badge is a signal to review before acting, not an override of the recommendation itself.
What does “the model reached that independently” mean?
It means the incrementality test result was not fed into the MMM as an input.
The model built its estimate from your media spend and revenue data alone, and the test independently measured the same channel. Agreement between the two is therefore a genuine external validation rather than the model repeating data it already used.
Where can I see the full test behind a badge?
Select the badge to open the Compass Confidence modal. Then select Open Test Results at the bottom.
This takes you to the full incrementality test results page for that specific test.
Can a subcategory have more than one test linked to it?
A subcategory will show the most recent eligible test that overlaps with it.
If you have run multiple tests on the same channel over time, only the most relevant one appears in the modal.