Your ranking is not a single number.

What varies, and by how much

Location, which is the big one. Not just country but city and often narrower. For anything with local intent the results reorder over a few miles, which is why a single citywide ranking number for a business with branches is close to meaningless. It is also why a London ranking is really dozens of neighbourhood rankings averaged into one figure.

Device. Mobile and desktop results differ in order and in what sits above the links, and the split matters because your traffic is rarely fifty-fifty.

Personal history and signed-in state. Modest for most queries, real for some, and enough that checking your own ranking in your own browser is the least reliable method available. You have been visiting your own site for months.

The page itself. The number of organic slots above the fold varies with how many features Google decides to show for that query on that day. Position three with an AI Overview, a map pack and four ads above it is a different position three from one with nothing above it.

The more honest figure

Search Console average position is closer to the truth, precisely because it is an average of real impressions from real people across all those varying conditions. It is also frequently distrusted for looking worse than the tracker, which usually means the tracker's chosen conditions were flattering rather than that Search Console is wrong.

Its own caveat is worth knowing: an average across a range hides the shape of that range. Sitting steadily at position eight and swinging between three and fifteen can produce the same average and are completely different situations, one stable and one contested. Look at impressions and clicks alongside it, because those reflect what actually happened rather than where you theoretically sat.

Why the single number causes bad decisions

It invites optimising the measurement. Pick the tracking location where you look best, and the report improves without the business changing. Chase a head term from position twelve to nine and celebrate, while the terms actually producing enquiries go unwatched because they were not on the list. Domain Rating and traffic fail the same way: a convenient number becomes the target and stops describing the thing it stood in for.

It also produces false alarms. A two-place drop is well inside normal variation between conditions and days, and treating it as an event burns hours diagnosing noise. Track the direction over weeks, not the digit today.

What to report instead

Rankings are a diagnostic, not an outcome. Keep them for what they are good at: telling you whether a page is gaining or losing ground on the queries you chose to compete for, with the conditions held constant so the comparison means something. Then report the things that pay: clicks and impressions on the pages that matter, and the enquiries or sales that follow. If a keyword moved up three places and nothing downstream changed, the movement was not the point, and that is a finding worth stating plainly rather than presenting as progress.