Why Search Console's average position misleads, and what to read instead

The average position shown for a page in Google Search Console is a blend of every search that page appeared for, including searches it was never meant to rank for. A page can look weak because of a long tail of irrelevant searches while the ones that matter are doing fine, and the reverse is also true. Before changing a page because of its average position, look at the individual searches behind the number. Here are three cases from our own client sites where the average pointed at the wrong problem.

Case 1: a page at 36.7 that was mostly healthy

On a property information site in September a development guide showed 3,504 impressions at an average position of 36.7. A second post showed 523 impressions at 60.0. We were asked which one needed work.

The guide had 178 separate searches behind it. The ones it was written to win, such as the cost of building a dual occupancy, sat between positions 15.7 and 22.7, with 810 impressions between them. The 36.7 came from about 95 definition style searches at positions 55 to 90, where the page showed up as a weak second result and never would win. Shedding that tail would make the number look better without adding a single click.

The second post was the real problem. Its best position on any search was 49.2. It had a tenth of the impressions, but nothing it ranked for was close to page one.

If we had gone by the page average alone, we would have rewritten the healthy page and left the sick one.

Case 2: a page at 20.0 that was a few scrapers

On our own site, a WordPress migration checklist showed 92 impressions at position 20.0 over 90 days. That reads like a page that is a small push from page one.

The searches said otherwise. Seventeen of the impressions came from scraper queries made up of the address of another site's WordPress feed, where we happened to appear at positions 7 to 11. The searches people actually typed for a migration checklist were at positions 53.6 and 68.7. Those few junk rows had pulled the average about 30 places better than the real demand.

An answer block written to lift a position 20 page would have been aimed at searches nobody makes.

Case 3: a homepage at 40.7 that was mostly out of area

On a pest control site the home page showed 3,727 impressions at position 40.7 with 14 clicks, and it looked like a collapse. We split its 112 named searches by place:

Take out the out-of-area group and the page sat at 28.3, not 40.7. That group will never produce clicks, because nobody in another city hires a pest controller from a different region.

Two things caused the out-of-area traffic, and it is worth checking your own site for both. The first is a name collision. The business name was close enough to a place name in another area that searches for pest control around that place pulled 739 impressions. The second is a region word that means somewhere else. Some Sydney suburbs sit on a railway line named after the client's region, so a site optimised around the region name matched searches in Sydney. Look up every unfamiliar suburb before assuming it is in your area.

What to read instead

For any page you think has a position problem:

1. In Search Console, open the Performance report, click the page, then switch to the Queries tab. The searches are listed with their own positions.
2. Pick out the searches the page is meant to win, usually the commercial ones.
3. Ignore rows that are URLs, odd strings, or places you do not serve.
4. Look at the impressions from the searches that sit between positions 11 and 30. That band is where a better page can realistically reach page one. A page with a lot of impressions there is worth working on; a page whose best search is at position 49 needs a different approach.

If the searches that matter are already in good positions, the right action is often to leave the page alone and note what its average is made of.

A caution on small numbers

Searches with a handful of impressions swing the average further than any other group. Our own site recorded between 8 and 10 clicks a month over recent months, so a single query that appears four times can move a page's average by several places. Before you act on a movement, look at how many impressions sit behind it. A change from position 20 to position 14 on 6 impressions is not a trend.

The same applies to weekly comparisons. On a small site, a jump from 1 click to 9 reads as an increase of 800 percent and means very little.

What this changes in practice

Reading a report at page level is quick and often wrong. Reading at search level takes longer and tells you which pages deserve the work. It is also why we ask for the searches behind a number before agreeing a page needs rewriting, whether the page is ours or a client's.

If you would like this done for your own site, it is part of what the SEO add-on covers, and our post on what a monthly SEO retainer should contain lists what to ask for in any report. For a first look, our free website audit is a starting point.