Lesson 5 of 7 in Measure What Happened
- 1Define the measurement window before starting
- 2Choose a metric that matches the task
- 3Connect published URLs or provider identifiers where required
- 4Record manual results safely
- 5Read integration-backed results
- 6Separate task-linked metrics from workspace metrics
- 7Interpret met, partial, missed, and inconclusive results
Why this matters
For integration-backed tasks, you do not fetch the number; the app does. That is a convenience and a risk: it is easy to treat "the app will handle it" as "the result is settled." It is not. The baseline is fixed early, the window still has to close, and provider reads can fail. Reading integration-backed results correctly means knowing which state the result is in (pending, final, or failed and visible) and acting accordingly.
When to use it
After marking an integration-backed task done with its binding saved (see Connect published URLs or provider identifiers where required), at any point until the window closes and after.
Before you start
- The task is done and, where the card asks for one, bound to the exact published page or post.
- The relevant integration shows Connected (Workspace → Integrations).
Do this
- 1Open the task and confirm it shows "Task completed. Measurement pending." This is the pending state: the app knows what to read and the window is running. Do nothing except wait.
- 2Note what the result is measured against: the baseline the app captured when you marked the task done, or zero from the moment you saved a published URL. The final result is read as movement from that baseline. The app compares; you interpret.
- 3While the window runs, the card may show "Live tracking" with the current value and the date it finalizes; that value is interim. When the window closes, the card shows "Final result" with the value, the target, the verdict, and when it was last updated.
- 4If a provider read failed, it stays visible; the app does not silently drop the result. Read the failure state, then go to Workspace → Integrations, run Sync Now on the provider, and check whether the read recovers on retry.
- 5If the window closes without a usable read, the card shows "Measurement ended without a usable result." Note what was attempted and why it failed in your review, and fix the binding or connection before the next task. A visible failure is evidence; a silently missing number is not.
- 6Never treat a pending or interim reading as final. The verdict comes after the window closes (see Interpret met, partial, missed, and inconclusive results).
What you should see
These states are all legitimate: pending ("Task completed. Measurement pending.", with the date the final result is due), interim ("Live tracking", with the date it finalizes), final ("Final result", with the value, target, and verdict), and failed but visible ("Measurement temporarily unavailable." while the app retries the provider read, or "Measurement ended without a usable result." if the window closed without one). What you should never see is a number presented as final before the window closes.
How to judge it
You can point at the task and say: pending (waiting on the window), final (window closed, number recorded), or failed (read failed, retry attempted, failure noted). It is always one of the three, never "probably fine."
What it does not mean
Baseline-to-current movement is observed movement, not proven causation. "Pageviews rose after the publish" is honest. "The publish caused the rise" needs more evidence than one window on one page. Keep the language to observed after, coincided with, based on.
Try it in Signal & Science
Open the Marketing Plan and check whether any integration-backed task is still pending on its window.
Open in Signal & Science