# Measurement windows

> The Action Center measures each finished action over a set number of tracking runs, not days. What each countdown means and why runs are the unit.

Source: https://ranqo.ai/docs/methodology/measurement-windows

When you move a recommendation to **Done** in the [Action Center](https://ranqo.ai/docs/guides/action-center), its measurement window starts. The window is counted in tracking runs, not days, because your numbers can only change when a run happens. This page explains how long each window is, what the countdowns on the page mean, and why a brand that skips a week is never marked down for it.

## Why runs, not days

A metric moves only when a run asks your prompts again. A countdown in days assumes a run lands on a schedule, and that is not always true:

- **Brands run on their own day of the week**, once every 7 days, so the same number of days holds a different number of runs for two brands started on different days.
- **A run can be missed.** A failed run is retried at the next daily check, a day later.
- **A paused brand produces no runs at all.** A day countdown would expire with nothing measured.

Counting runs means a window for one type of work always holds the same number of runs, whichever day your brand runs on and however long a pause lasts. If no run happens, the window waits.

Every run that completes after you move the card to Done counts, including a run that was already in progress when you pressed Done. That run may have asked your prompts before you acted, so it can count as the first run of the window without containing your work.

## The first reading

The first reading is due after 2 runs: one to see a change, and one to see it hold. Ranqo measured why across real brands' runs: a mention seen for the first time is still there in the next run only 54.2% of the time, while one seen in two runs in a row is still there 96.1% of the time. A single new mention is close to a coin flip, so it is not treated as a result. See [Confirmed wins](https://ranqo.ai/docs/methodology/confirmed-wins).

**Re-measure now** is available before that too. A reading taken after a single run uses that run alone, none of its answers can be confirmed, and it is the card's only reading.

## The verdict

A "no effect" verdict needs longer, because some work takes several runs to show up in AI answers. How long depends on the type of work:

| Recommendation type | Runs before a verdict |
| --- | --- |
| Create content | 4 |
| Build authority | 4 |
| Outreach | 4 |
| Optimize a page | 3 |
| Fix a technical issue | 3 |

In the drawer's **Lift** step, a reading larger than its [noise band](https://ranqo.ai/docs/methodology/noise-bands) shows as a number, with **early reading** under it while the card is short of its verdict window and **re-measured** after. A smaller reading shows as **No detectable change**, with *Too early to call* and the runs still to go before the verdict window, and with the metric's normal week-to-week swing after it. A reading that compared two runs of answers shows answers won instead (see [Confirmed wins](https://ranqo.ai/docs/methodology/confirmed-wins)).

These labels are worked out from the runs completed so far, not from the runs the reading was taken on. The reading itself stays as it was taken, so a reading taken after one or two runs is labelled as final once the card's verdict window has passed.

Technical fixes carry **No visibility change expected** in the drawer's **Lift** step until a reading is taken, and their expected outcome says they are tracked as done rather than measured for lift. They still run the window in the table above and show **Re-measure now**. A reading taken on one is shown and counted in **Verified Lift** like any other.

## Reading the countdowns

| Where | What it shows | Meaning |
| --- | --- | --- |
| A Done card's footer | **After N more runs** | Runs still to complete before the first reading is due. |
| A Done card's footer | **Re-measure ready** | The first reading is due. It stays until you take the reading. |
| A Done card's meta row | **run N/M** and a thin bar | Runs completed since you moved the card to Done, out of the verdict window for this type of work. The count stops at M. |
| The drawer's **Re-measure** step | **N more runs**, then **run available** | Runs still to complete before the first reading is due. **Re-measure now** sits under it throughout. |
| The **Awaiting Re-measure** card | *soonest in N runs (~Nd)*, or *window open now* | The nearest first reading across Done cards that have no reading yet. The days assume one run every 7 days, for planning. |

Once a reading is taken, the card's meta row shows **measured Nd ago** instead of the run counter, and the drawer's **Re-measure** step reads **done**.

## Taking a reading

Ranqo does not take readings on its own. Open a Done card and press **Re-measure now** in the loop. Ranqo recomputes the recommendation's metric from the latest completed run that finished after you moved the card to Done, and compares it with the baseline recorded when the recommendation was written.

- If no run has completed since then, a notice says so and nothing is recorded.
- If the metric has no data in the latest run, for example because you stopped tracking the engine or the category it measures, a notice says the lift cannot be measured. Ranqo never switches the recommendation to a different metric.

A card gets one reading. After it, **Re-measure now** is gone, and the reading stays as it was taken while later runs arrive. Moving the card out of Done and back keeps the old reading and does not bring the button back.

To take the reading at the end of the window, press **Re-measure now** once the run counter is full (**run M/M**). If a run was in progress when you moved the card to Done, wait for one more run after that.

The metric, its baseline and the answers it covers are fixed when the recommendation is written, from the latest completed run at that moment. A recommendation without a recorded baseline has no window to run.

## Related

- [Noise bands](https://ranqo.ai/docs/methodology/noise-bands): How big a change has to be before the Action Center drawer shows it as one.
- [Confirmed wins](https://ranqo.ai/docs/methodology/confirmed-wins): Counting the specific answers you won, which the overall number is too noisy to show.