My average position in Search Console got worse while my reach doubled

Last week I posted here about my side project https://tarot-magic.com ranking at positions 6-9 in Google with almost zero clicks. I promised to come back with numbers after a two-week test. This is the follow-up, and the most useful thing I learned is not about snippets at all. It is about how misleading the "average position" metric is on a small site.

What the dashboard showed

For my Ukrainian card pages, comparing Sep 7-20 with Sep 21 - Oct 4:

• average position went from 11.2 to 21
• impressions went from 231 to 463
• clicks went from 0 to 2

If you only look at the position line in the chart, it looks like the site collapsed. It did not.

What was actually happening

I exported the page list and sorted it. Some pages that already ranked in the top 10 held their spot: Two of Swords at 6.5, Ten of Cups at 6.9, Temperance at 8. Others fell, and I will get to them below. What changed is that Google started showing pages that had zero impressions before, like Three of Wands and Five of Wands, and they appeared at positions 20-35.

More pages entering the results at low positions is good news. But each new page pulls the average down, so the metric moves in the opposite direction from reality.

The honest way to read it is to split by position bucket. Top-10 pages only: 157 impressions and 2 clicks now, 197 impressions and 0 clicks before. Everything else is new reach that did not exist two weeks ago.

The real drops

Not everything was noise. Six of Wands fell from 7.4 to 37.5, and it was my biggest Ukrainian page in the top 10 with 87 impressions. A few more slid from 6-7 to 30-39, for example Ten of Swords and Nine of Wands, though each of them had only 4-14 impressions before. On a sample this small, one page like Six of Wands can wipe out a CTR experiment on its own.

What I do now

When I check Search Console on this site, I no longer look at the site-wide average position. I split it in two instead. First, impressions and clicks only for pages that were already in the top 10 in the previous period. Second, the number of pages that got their first impressions. On top of that I check by hand any page that moved more than 10 positions.

It is more work, but on a site with a few hundred impressions a week the average is mostly a measure of how many new pages Google is testing.

Next check is mid-October. If you track a small site, I am curious which view you use instead of the average.