Configure and Interpret Testing Decisions

Understand rule profiles, decision labels, reasons, and human review in the Testing workspace.

Before you begin

Review the organization’s Testing access and current rule profile. Gather the commercial assumptions the team intends to use and confirm that costs, order evidence, and advertising connections are sufficiently complete. Rule changes can alter how many rows are classified, so note the current behavior first.

How it works

Testing rules evaluate supported row evidence and produce an action bucket and reason. Thresholds and rule conditions provide a repeatable first assessment, while setup-health checks identify evidence that may make the result unreliable. The output supports prioritization; it does not execute advertising, purchasing, or product decisions.

Step-by-step

Open Testing and review several representative product or campaign rows.
Note their current actions, reasons, break-even context, and setup-health status.
Open the available Testing rules settings with an authorized account.
Review the active profile and each threshold before changing it.
Use the rule preview when available to understand how proposed values affect classifications.
Save only rules that match the organization’s documented review approach.
Return to the grid and compare action labels and reasons across the same evidence window.
Investigate warnings before accepting a decision label.
Use review state to record human follow-up separately from the calculated action.
Revisit rules when product economics or operating goals materially change.

Check your result

Decision labels have understandable reasons, the rule profile reflects the intended organization thresholds, and reviewers can distinguish a calculated action from their own review state. Rows with incomplete evidence remain visibly qualified rather than appearing certain.

Common problems

Many rows change unexpectedly: compare the edited threshold with the previous value and preview on representative evidence.

A label conflicts with business judgment: read the decision reason and inputs, then record human review rather than forcing the data to match a conclusion.

No decision appears: required spend, sales, order, match, or economic evidence can be missing.

The settings panel is unavailable: rule configuration may require broader organization access than viewing Testing.

Permissions and data notes

Rule settings are shared organization behavior and should be changed by authorized users. Keep an internal record of why thresholds changed. Decision labels are analytical outputs, not autonomous business actions, financial advice, or guarantees of future performance.