On Friday morning I decided to rewrite one of my own pages. It is the page that explains moving a shop off Wix, and a paying client came in through it. Search Console had it at position 45 on roughly 1,600 impressions — shown to a lot of people and found by almost none of them. Page five. Rewrite it.
By the afternoon I had not touched a word of it, because the page is at 25 and climbing.
Both numbers came out of the same account on the same morning, and neither of them is wrong.
A statistician drowned crossing a river with an average depth of four feet.
The first number is the 90-day average, which is what Search Console shows when you open it, and the second is last week. Neither is lying; an average just describes a stretch of water rather than the bit you are standing in. That page ran between 36 and 54 every week through June and July. Then I changed something at the start of August, and the 4 weeks since read 54, 34, 25, 21. So the figure is 11 weeks of before and 4 weeks of after, folded together into one number that comes out at 45 and describes a page that no longer exists.
Averages smooth things, and that much is not news. The default view is 90 days, the decision taken from it is usually is this page working, and 90 days is longer than the gap between most changes a small site makes. If you did anything to your site in the last two months, the number you are looking at is mostly the before.
Mine went further than a stale average once. Between January and March this year my own site quietly shed 66 of its indexed pages, 160 down to 94, over 10 weeks, and nothing I was looking at was wrong. The averages held the whole time. I spent £4,785 on an SEO platform between 2020 and 2025, 8 of those months at more than £178, and then I cancelled it and read Google’s own figures instead. The collapse hid from both of them. The number that would have shown it was sitting in a coverage export I had never once opened.
I had the file open. That is the part that stays with me — not that I read the number wrong, but that I was one command from spending a morning rewriting a page that was busy fixing itself, and I would have called it a productive morning afterwards and had something to show for it.
Here is the check, and it genuinely takes about a minute.
Open Search Console, go to Performance, and find the date range at the top. It says 3 months. Change it to 28 days and write down what you see. Change it to 7 days and write that down too. Do it for the 3 or 4 pages that matter, not for the whole site.
Where the 3 numbers agree, the page is settled and the 90-day view is fine. Where they disagree by more than a few positions, the page is in motion, and the 90-day figure is the wrong one to make a decision from. A page moving up wants leaving alone. A page moving down wants looking at this week rather than in a quarter’s time, and the 3-month average will hide a slide for weeks before it admits to one.
Send a client the 3-month average every month and you have sent them the same number twelve times.
Direction is not the only thing an average hides. A single page can be two audiences wearing one number. I have a page whose short searches convert at 9% while its long ones convert at nothing, and the two of them average out into a figure that describes neither — one that looks like a failing page and is nothing of the sort. The same trick shows up when two of your own pages are competing for one search and the total looks healthy while both halves are losing.
The pattern underneath all three is the same. Search Console gives you a summary by default, the summary is where the decision gets made, and the summary is the one place the answer is not.
