Faragiri

Learn Ranking How search works

A Skipton skip hire page sat at 3 on a phone in town and 19 on a desktop forty miles away, on the same afternoon, for the same five words.

4Sections
4Min read
4Depth

A page does not have a ranking, it has a position for one query at one moment. Skipton Skips has one page about skip sizes, and on the afternoon of 14 July 2026 somebody checked it from five queries in two places. Skip hire skipton was 3 on a phone in town and 19 on a desktop in Leeds. The page did not change between the two checks; it was the same stored copy, forty miles apart.

That is not a quirk to work around. It is what ranking is. The order is computed when somebody types, from candidates the index can supply, and the inputs to that computation include where the person is, what they are holding and what else has been asked recently. Nothing is stored per page that could be called a score.

Five queries, ten positions

The five checks were skip hire skipton at 3 and 19, skip sizes explained at 8 and 7, how big is a 6 yard skip at 14 and 12, skip permit north yorkshire at 6 and 22, and cheap skip hire at 41 and 44. Two of them barely move and two of them move enormously, and the split is not random.

Ranking

The same page, five queries, two places, one afternoon

The two queries that ask for a supplier move by sixteen places between a phone in town and a desktop forty miles away, and the two that ask for a fact barely move at all.

Position in results, 14 July 2026, checked by hand. A lower number is nearer the top.

Phone, in SkiptonDesktop, in Leeds

  1. skip hire skipton319A request for a supplier, answered from where the person is standing.
  2. skip permit north yorkshire622Same shape, same movement, and the county in the query does not stop it.
  3. skip sizes explained87A request for a fact. The answer is the same in both cities.
  4. how big is a 6 yard skip1412Also a fact, and also nearly still.
  5. cheap skip hire4144Far enough down that neither number means anything.

Ten readings taken by one person in one afternoon. A second afternoon would give ten different numbers, which is the point rather than a defect in the method.

A query with local intent is answered from where the person is standing. Skip hire skipton and skip permit north yorkshire are both requests for a supplier, so the distance between the searcher and the yard is an input, and a desktop in Leeds is asking a different question with the same words. How big is a 6 yard skip is a request for a fact, so it barely moves: the answer is the same in both cities.

Where this stalls: the average position

Search Console reports an average position and it is the most quoted number in this pillar. An average position is an average of things that are not the same measurement, because it mixes the 3 and the 19 above, and it also mixes queries the page half-matches with the one query it was written for.

Reporting

Why the summary position hides the real traffic

Averaging across places and queries invents a rank nobody saw, while sorting by position alone puts low-demand terms ahead of busy ones.

  • 14.6Reported average positionBlends positions 3 and 19 with half-matched queries into a figure no searcher saw.
  • 40Impressions at position 3Sits at the top when sorted by rank, even though demand is tiny.
  • 900Impressions at position 11Worth more than the rank-3 query, yet buried unless you sort by impressions.

Use the query filter and device tab in Search Console to split the 14.6 back into separate rows.

Skipton's page shows an average of 14.6, which describes no event that has ever happened to anybody. Read positions by query and by device, never as one number, and the report will do it: the query filter and the device tab are two clicks and they turn one meaningless figure into five useful ones.

The second half of that habit is to stop reading position on its own. Position 3 with 40 impressions is worth less than position 11 with 900, and a report sorted by position puts the 40 at the top of the page every time. Sort by impressions and read the position beside it.

Before any ordering happens

The candidate set is decided before any ordering happens, and most pages that never appear never lost a ranking contest. They were not in the set. A page has to be in the store, and the store has to be able to match it to the words typed, before anything is put in order at all.

Eligibility

What has to hold before any page is ordered

Pages missing from results for terms like grab lorry hire did not lose a ranking race; they never entered the candidate set.

When the second link breaks, the fix is adding the right language to the page rather than chasing rank signals.

That is why a page can be excellent and invisible for a query it does not contain the language of. Skipton's page never appears for grab lorry hire, not because it ranks badly for it but because the phrase is not on it and no stored entry connects the two. The fix is a writing decision rather than a ranking one.

What can be watched, and what cannot be bought

Search Console's performance report holds the positions and it is free. It replaced the keyword data Google removed from analytics in 2011, and the number to know about it is that its position is averaged over every impression in the range, including ones from places the business does not serve.

Measurement

What tools can track versus what decides visibility

Reports and rank trackers record positions from the surface, while indexing and answer citation operate underneath where checks cannot reach.

  • Search Console averagesFree since replacing analytics keyword data in 2011, averaged over all impressions.
  • Paid rank trackers (2026)Bought for fixed-place consistency over time rather than hidden accuracy.
  • Monthly manual five-query logMatches paid consistency for a small yard at zero cost.
Measured in reports vs decided by the engine
  • Whether a crawled page is storedDecides candidate eligibility before any ordering happens.
  • Citation in generated answersPays out differently from ten blue links when an answer is synthesised.
  • Per-query position swingsWhy a Leeds desktop showed 19 instead of 9 remains opaque.

Watch trends with consistent checks, but fix candidate eligibility and content substance underneath.

Rank trackers, the products that check a fixed list of queries from a fixed place every day, are all paid as of 2026, and the honest reason to consider one is not accuracy but consistency: they check the same query from the same place at the same time, which is the only way to see a change rather than a difference. At a yard this size a note of five queries checked by hand once a month does the same job for nothing.

What decides whether a page is a candidate at all is being stored, and why a crawled page may not be, and increasingly the order is not the whole result, because being cited when the answer is generated pays out differently. Both sit under the mechanism, before any tactic makes sense, and nobody at Skipton can say why the Leeds desktop showed 19 rather than 9.

Alongside this

Above this