How to Know If the Recovery Actually Worked: Segment, Measure, Verify

Most recovery projects are judged on the wrong number. Total organic traffic is the last thing to move and the least informative.

Verification is a measurement problem, not a patience problem. Set it up before the work starts and the answer arrives on its own.

When can you measure at all?

Not until the reassessment has happened. That date depends entirely on what demoted you.

Cause	When reassessment happens	Measure from
Technical fault	On recrawl	2 to 4 weeks after the fix
Manual action	After approval	2 to 4 weeks after the message clears
Spam policy	Continuously	4 to 8 weeks after cleanup
Core quality demotion	Next core update only	28 days after that rollout completes

Measuring before that date tells you nothing. Rankings move in both directions during a rollout, so week two is not a result.

Segment before you look

This is the step that changes most verdicts.

If sixty percent of your site was never affected, its traffic dilutes the signal from the forty percent that was. A real recovery in the affected group can look like nothing in the total.

• Split by page group. The template or folder that lost traffic, versus everything else.
• Measure each separately. Against its own pre-drop baseline, not the site total.
• Compare the two. Affected group up while the rest is flat is the clearest possible success signal.

The page scope analyzer does this split from a comparison export.

The four metrics, in order of usefulness

Read them in this sequence. Each one answers a different question.

1. Average position, affected pages only. This is the ranking signal. It moves first and it moves cleanest.
2. Impressions, affected pages only. Are you being shown more often than you were?
3. Distinct queries ranking in the top ten. Breadth returning, which precedes volume returning.
4. Clicks. Last, because it is affected by SERP features you do not control.

Position improving while clicks stay flat is still a successful recovery. It means you won the ranking back into a results page that now gives fewer clicks.

What a successful recovery looks like

Rarely a vertical line. Expect a staircase.

• Your strongest pages move first. They are crawled most often.
• The long tail follows, sometimes weeks later.
• Movement happens during the rollout, not evenly after it.
• It plateaus below the old line. Sixty to ninety percent is the normal ceiling.

Why the ceiling exists is covered in my recovery stalled at partial.

Four false readings to avoid

Each of these has cost somebody a correct decision.

• Measuring against a peak. If your pre-drop figure was your best month ever, you are comparing to an outlier. Use a normal twenty-eight day period.
• Ignoring seasonality. A recovery completing in your soft season looks like failure. Compare year on year.
• Measuring mid-rollout. Week two gains routinely disappear by week four.
• Counting new content. Pages published during the project are new traffic, not recovery. Exclude them.

The last one inflates results quietly and is the easiest to do by accident.

If nothing moved at all

Work through these four before concluding the work failed.

• Has a reassessment actually happened? If no core update has run since you finished, there is nothing to conclude yet.
• Was the work crawled? Check the last crawl date on your changed pages. Uncrawled work is invisible.
• Was the diagnosis right? A perfect fix for the wrong cause produces nothing.
• Was the change deep enough? Partial cleanups reach the same verdict as no cleanup.

The third is the most common and the least comfortable. Two reassessments passing with no movement usually points at the diagnosis rather than the execution.

Set this up before you start

Five things, recorded in writing, before any work begins.

• The baseline period and its figures, agreed by everyone involved.
• The affected page group, defined as a URL pattern.
• The metrics you will judge on, and in what order.
• The earliest honest measurement date, based on the cause.
• What would count as failure. Decide this before you have a result to defend.

The last one is the discipline that separates a test from a justification. Presenting all five is covered in how to explain a traffic drop to your boss.

Freeze the site while you wait

This is the hardest instruction to follow and the most important.

Continuing to change things during the wait destroys attribution. If the site recovers, you will not know which change did it. If it does not, you will not know which change prevented it.

Ordinary publishing is fine. Structural changes are not. Keep templates, URLs, navigation and indexing settings still until the measurement date passes.

Setting the baseline itself, and the three ways it quietly gets gamed, is covered in how to set a recovery baseline.