Your average position is an average of one query

Open the performance report on a shop that has been live a few months and there are four numbers across the top. Clicks, impressions, click-through rate, average position. The last one is the one people quote, because it sounds like a league table — a position is a place, and a place can be improved.

It is not a place. It is the mean of every position your site held, across every query it appeared for, weighted by how often each query was searched. That construction is fine. The problem is what it does on a site whose queries are not alike, and almost no shop's queries are alike, because one of them is your own name.

Why the brand term dominates the average

People who search your business name already know you exist. They are checking a return policy, finding the shop again, following up on a link somebody sent. There is very little competition for your name, so you rank first or second for it, and because those searchers are the ones most likely to have a reason to look you up, that query often carries a large share of your total impressions.

Now average it with the rest. A brand term at position one, carrying a seventh of your impressions, pulls the mean down hard. Every discovery query — the ones from people who have never heard of you, which are the only ones that can grow anything — is averaged with it. The headline comes out looking like a site hovering near page two.

What ours actually looks like

This site's own report, read this week, is a clean example and it is only fair to use our own numbers rather than a hypothetical. The headline: average position 11.1. Read as a league table that is page two, close, a near miss.

Split by query it is a different site. The brand term sits at position 1.5. Every other query on the property — four platform searches, one competitor comparison, one category word, one head term — sits between 49.5 and 93. There is no query anywhere in the middle. The 11.1 describes nothing that exists: it is one query at the top and seven in the fifth to ninth page of results, and the average lands in a gap where we have never appeared at all.

The decision this changes

This matters because the two readings point at completely different work.

• If you are genuinely at position 11, you are being shown to people and not clicked. The job is the title and the description — the thing a person reads before deciding. Rewriting those is the highest-value hour you can spend.
• If you are at position 60, nobody is reading your title. A result on page six is not a result that lost a click; it is a result that was never put in front of anyone. Rewriting it changes nothing, and you can rewrite it ten times and watch the number not move, because the number was never about the words.

The second case is upstream work: being crawled at all, being linked to from somewhere that already has an audience, having pages that exist for the thing people type. Slow, structural, nothing to do with copywriting. Choosing it over the title rewrite is only possible if you split the average first — which is the entire practical point of this post.

How to split it in about five minutes

In the performance report, open the Queries tab and add the Position column, which is off by default on some views. Sort by impressions. Then read it in two groups: queries containing your business name, and everything else. You want two separate medians, and you do not need to be careful about it — the gap is usually so large that eyeballing it is enough.

Two things to watch while you do it. A long tail of queries with one impression each is not a signal; one impression at position 80 means Google tried your page once, deep in a results page nobody reached, and you should not design anything around it. And if a query appears that has nothing to do with you — a random domain name, somebody else's brand — leave it in the count but ignore it as intent. Both are normal on a small property.

The habit underneath this generalises past search. A measurement answers exactly one question and it is easy to ask it a second question it cannot hear. The other common version is the number that moves while the number you care about does not — tidying up without spending less — and the shape is the same each time: something aggregate standing in for something specific.

For search specifically the rule is short enough to keep: never read an average position without knowing what share of the impressions came from your own name. Until you have split those out, the number on the dashboard is mostly telling you that people can find you when they already know who you are.