Did Your SEO Change Work? Measure It in Search Console

01How do you tell whether an SEO change worked?

Compare equal windows before and after the day the change went live, for that exact page in Search Console, leaving the change day out and waiting a few days for reporting to catch up. Call it a real change only if clicks moved by a meaningful percentage and an absolute number, with enough impressions in both windows.

That is the whole method. The rest of this guide is the detail that stops it going wrong, because the usual advice online is about something else. We looked at what ranks for these queries. Page monitoring tools such as the ones in seoClarity's roundup tell you that a page changed, not whether the change paid off. Broad measurement guides such as Siteimprove's list KPIs for a whole programme. We found no ranking page that gives a dated, single-page before and after procedure, which is the job here.

One caveat belongs at the top. This method shows what changed between two periods. It cannot prove your edit caused it. The sections on position and confounders explain how to make the evidence stronger, and where it stays weak.

02Which date counts as the change date?

The day the edit went live on the public page, not the day it was drafted or the day you asked Google to recrawl. Write it down when you ship, because you will not remember it in a month, and if several people edit the page the real date is lost.

Leave that day out of both windows. The edit went live part way through it, so the day holds a mix of old and new page and belongs to neither side. Note also that Search Console's daily data follows Pacific Time, so if you publish late in the evening in another time zone, your change day in the report may be the day before or after your own calendar says.

03How long should the before and after windows be?

Equal length, and a whole number of weeks. Search traffic has a weekly rhythm, so two windows of 21 days each contain every weekday the same number of times. Google's own help for comparing periods suggests weekly or monthly granularity to smooth out day of week effects.

Take the before window as the days immediately ending the day before the change, and the after window as the days immediately starting the day after. Then wait. Google says data is typically available after two to three days, and the newest figures may be preliminary and can change within hours. So with a 21 day after window, look at it about 23 days after the change, not 22.

Why 21 and not 7? A short after window mostly measures how fast Google recrawled the page. A very long one picks up other changes. Three weeks is a stated rule of thumb, not a finding: long enough for a recrawl and for a few hundred impressions on a modest page, short enough that little else changes. For a page with very high volume you can go shorter, as long as both windows keep a whole number of weeks.

04How do you compare before and after in Search Console?

No tool is needed. Following Google's help on advanced filtering and comparison:

1. Open Performance, then Search results, and make sure clicks, impressions, CTR and position are all switched on.
2. Click + Add filter, choose Pages, and enter the page. "URLs containing" is a text match, so a page whose address is a prefix of other pages will also match them. Use "Custom (regex)" with an anchored pattern when that is a risk.
3. Edit the date filter, choose Compare, and enter the two periods: the 21 days ending the day before the change and the 21 days starting the day after it. Apply, and the table gains a Difference column.
4. Read the totals first. Then open the Queries tab to see which queries moved, the Dates tab to see whether the shift is a step or a drift, and the Devices tab if you suspect a mobile only effect.

Write the four numbers for each window on paper or in a sheet before you interpret them. It keeps you from reading the chart you hoped to see.

05How big a move counts as a real change?

Use two tests together, plus a floor on data. A percentage alone is fooled by small numbers: going from 3 clicks to 4 is a 33 percent gain and means nothing. An absolute number alone is fooled by big pages: 10 extra clicks on a page earning 2,000 is noise.

These are our rules of thumb, not research findings. They are the same thresholds the Linkeddit Search Watchlist applies, chosen to be conservative so that a result you would act on is not a coin flip. Tighten them for a high traffic page and treat anything below the floor as not enough data rather than as no effect. Anything between the thresholds is flat: neither proof it worked nor proof it failed.

06Why is average position a bad verdict?

Because it averages over whichever queries showed the page. Google defines the table value as the average position for the URL or grouping dimension in that row, and for a page that is every query that displayed it, weighted however they happened to appear.

That makes it move for reasons unrelated to your edit. Say your rewrite makes the page eligible for a dozen new, long tail queries where it shows at position 25. Clicks rise, impressions rise, and the average position gets worse. Read naively, a success looks like a failure. The reverse also happens: lose the queries where you ranked poorly and the average improves while clicks fall.

So report position, but let clicks decide, and use the query list to see whether the position change came from the queries you targeted or from a change in the mix. If you set out to win one query, check that query's own clicks and position as a separate line.

07What else could explain the change?

Plenty, and a before and after test cannot rule them out. Check these before crediting the edit.

• Seasonality. If demand for the topic rises or falls at this time of year, both windows are affected. If you have the data, run the same two windows a year earlier and see whether the same shape appears.
• Other edits. A new internal link, a title change by a colleague or a sitewide template update in the same weeks all land in your window. Keep a change log for the site, not just for the page.
• Algorithm updates and result page changes. A shift on the day a Google update rolled out, or a new answer box appearing for your main query, can move clicks with no help from your page.
• A control page. Run the same windows on a similar page you did not touch. If it moved the same way, the cause is probably shared. If it stayed flat while the edited page moved, the evidence is stronger. This is a technique for making a comparison fairer, not a guarantee.
• One-off spikes. Look at the Dates tab. A single day of unusual impressions can drag a whole window, and a step that begins before your change date points away from the edit.

08How do you confirm Google saw the change?

Before you credit an edit, check that Google has looked at the new version. Open the URL Inspection tool for the page. By Google's description, it shows the last time the page was crawled, and its default results reflect the most recently indexed version, not the live page. If the last crawl is earlier than your change date, the index still holds the old page and nothing you see in the report can be the edit.

Compare the user-declared canonical with the Google-selected canonical too. Google notes there is no guarantee it will choose your preferred canonical, and if it picked a different URL, your clicks may be recorded against that other page, which makes a per page comparison read as a drop. You can request indexing from the same screen, but Google says doing so does not guarantee the page will appear in the index, and that indexing typically takes a day or so but can take much longer.

09What does a worked example look like?

Clicks moved 30 percent and 72 clicks, clearing both thresholds, and both windows are far above 100 impressions. Most of the gain came from clicks, not impressions, which points at the title and snippet rather than ranking. The page's last crawl was 3 September, after the change, so the index held the new version. Verdict: it probably helped. The two caveats still apply: check the Dates tab for a spike, and check a control page before telling anyone it worked.

10Can you automate this?

You can, and doing it by hand for every edit gets tedious. Linkeddit's Search Watchlist is one way to automate it, and we make it, so weigh that. You add a page and the day you changed it. It waits 23 days (21 plus the 2 day reporting delay), then compares the 21 days before the change with the 21 days after, change day excluded, using your own synced Search Console data.

• The verdict is Improved or Declined only when clicks moved by at least 15 percent and at least 5 clicks, Flat otherwise, and Not enough data under 100 impressions in either window.
• Position is reported but is not part of the verdict, for the reason above.
• It lists up to five top query moves and shows Google's index status from URL inspection beside the numbers.
• It is free to run: no checks are used.

It is part of the $60 per month Compete plan. It does not remove the caveats: it reports an observed difference, not proof of cause. The help page Watching a Page Change in Search covers setup, and the rank tracker page shows where it sits.

Frequently asked questions