Why Average Position in Search Console Misleads
You opened Search Console, saw an average position of 8.4, and either felt a little relieved or a little worried. Short answer: neither reaction is justified on its own, because average position is a blend across every query, page, device and day you have data for, weighted by impressions, which means it can rise while the specific pages you actually care about are falling, and it should only be trusted after you filter down to one query or one page at a time.
This post shows the mechanism plainly, with a real hillcane.co example labeled as one small site, and a simple illustration of two queries that shows how the site wide average can move without any single page changing rank at all. It ends with a practical routine: filter by page, filter by query, compare like with like, and treat the number as a direction, not a grade.
What Average Position Actually Is
Search Console Help defines position as an estimate of the average ranking of your site's URLs for a given query, based on the highest ranking URL for that query at the moment an impression happened. When you look at average position for your whole site, without filtering, you are looking at that same idea blended across every query and every page that generated an impression in the date range you selected, weighted by how many impressions each one had.
That weighting is the part that trips people up. A query with a thousand impressions at position 20 pulls the site wide average far more than a query with five impressions at position 1, simply because it contributes far more to the underlying calculation. A single site wide number cannot tell you which of those two situations is happening, or whether both are happening at once and partly canceling each other out.
None of this is a flaw in how Search Console works. It is simply what an average across a mixed, changing set of queries and pages will always do: compress a lot of different situations into one number that, read alone, tells you very little about any specific one of them.
It also helps to remember that the underlying set of queries and pages contributing to that average is not fixed from one week to the next. New pages start generating impressions as they get indexed. Old queries can drop off if a page stops appearing for them. Seasonal terms come and go. Every one of those changes shifts what is being averaged together, which means a moving site wide number can reflect a moving mix of contributors just as easily as it reflects any actual change in how well existing pages are ranking.
A Simple Illustration
Here is a labeled, illustrative example with made up round numbers, kept separate from any real data, to show the mechanism directly. Imagine a site with exactly two queries generating impressions in a given week. Query one, a branded search for the shop's own name, sits steadily at position 1 with 10 impressions. Query two, a new, unbranded long tail search the site just started ranking for, sits at position 60 with 100 impressions.
The site wide average position across those two queries, weighted by impressions, comes out far closer to 60 than to 1, simply because query two contributed ten times the impressions of query one. Now imagine the following week: query one stays exactly the same, position 1, 10 impressions, and nothing about it has changed at all. Query two, still brand new, climbs from position 60 to position 45, a real, meaningful improvement, and its impressions grow to 150 as it starts showing up more.
The site wide average position in week two will have improved, because query two moved up. But notice what a naive reading misses: neither query's improvement was uniform or guaranteed, and if query two had instead added a third, entirely new long tail query at position 80, the site wide average could have gotten worse in the same week that two individual pages were doing exactly what you wanted, ranking for new terms and slowly climbing. A new page appearing for a long tail search at a low position can drag the average down even as the site is genuinely improving, and that is precisely the situation this illustration is built to show plainly, with labeled, round numbers rather than anything drawn from real data.
Run the illustration one step further to see how quickly this compounds. Add a fourth week where a third new query appears, position 70, 200 impressions, while queries one and two hold their week three positions exactly. The site wide average, now blending three very different situations instead of two, moves again, not because anything about queries one or two changed, but purely because the mix being averaged grew larger and shifted. A site owner watching only the top line number across these four weeks would see it move multiple times without ever learning which query actually caused each shift, or whether any of those shifts reflected good news or simply a new page finding its first foothold.
A Real Case From Hillcane's Own Site
We can offer one real, small case rather than only an illustration, described honestly as one site's data. On hillcane.co, comparing the site wide average position for a given window against the position for individual top pages and top queries in that same window shows a meaningfully different picture depending on which level you look at, which is exactly the pattern this post is describing.
A branded query for the site's own name tends to sit at a strong, stable position, while a newly published page targeting a specific, narrow topic can sit much further down the results while it is still building any track record at all. Blended together into one site wide figure, those two very different situations produce a single number that represents neither of them accurately on its own.
We are stating this as one small site's pattern, dated to this post's writing, not as proof of how any other cafe's Search Console data will behave. The mechanism, an impressions weighted blend across very different queries, is the documented, general reason average position behaves this way; the specific numbers on any one site, including ours, are just one example of that mechanism in action.
This is also why we are deliberately not publishing a specific blended average position number for hillcane.co in this post. Any single number, pulled from a specific date range on a small, young site, would invite exactly the kind of misreading this entire post is arguing against, treating one snapshot figure as though it meant something conclusive about the site's overall health. The pattern, not the number, is the useful takeaway here.
Myth Versus Fact
Myth: a rising site wide average position means the site is doing better everywhere. Fact: it can mean one heavily weighted query improved while others stayed flat or worsened, with the improvement simply outweighing everything else in the blend.
Myth: a falling site wide average position means something is going wrong. Fact: it can mean a brand new page just started ranking for a new term at a low position, which is often a sign of real, early progress rather than a problem, since a page has to appear somewhere before it can climb.
Myth: average position tells you how a specific page is performing. Fact: it tells you nothing about any specific page until you filter down to that page directly, since the site wide figure blends every page together.
Myth: a lower average position number is always better than a higher one on its own. Fact: it depends entirely on what is being averaged. A site with ten branded queries all at position 1 and nothing else will show a better average position than a site actively ranking for fifty new, competitive terms still climbing from position 40, even though the second site may be doing genuinely more valuable work.
Myth: if the site wide average holds steady, nothing meaningful is happening. Fact: a steady blended number can hide real movement underneath it, older pages slipping slightly while newer pages climb, roughly canceling out in the average while both trends are worth knowing about individually.
The Three Filters to Apply Before Believing the Number
First, filter by page. Look at average position for one specific page at a time, not the site as a whole, so you can see whether that particular page's position is moving and in which direction, rather than inferring anything about it from a blended figure.
Second, filter by query. For any page that matters to you, look at the individual queries driving its impressions, since a page can rank well for one term and poorly for another at the same time, and only a query level view separates those two realities.
Third, compare like with like. When you do look at a trend over time, compare the same page or the same query across two comparable windows, not a site wide average from one period against a site wide average from another period where the underlying mix of pages and queries has changed meaningfully, since a shifting mix alone can move the number regardless of what any individual page did.
It is worth writing down, somewhere you will see it again, which specific pages and queries you have decided matter most, so this three filter routine does not depend on remembering the right filters from memory every time. A short, standing list, this local page, this branded query, this new topic page being watched closely, turns the routine into something repeatable in a few minutes rather than something rebuilt from scratch on every visit to the report.
A Practical Routine
Treat position as a direction, not a grade, and build a short habit around that. When you open Search Console, resist reading the site wide average position card as a verdict on how things are going. Instead, pick the handful of pages or queries that matter to your business, a page for your most important local search term, a query for your shop's own name, and check their individual position trends specifically.
If you want a single site wide number for a general sense of direction, use it only as a rough signal to investigate further, never as a conclusion on its own. A meaningful shift in the site wide average, in either direction, is a prompt to go look at the page and query level data behind it, not a result to react to immediately without that closer look.
This routine costs a few extra minutes compared with glancing at one number, and it is worth those minutes, because the alternative, reacting to a blended average as though it describes any one page's real performance, leads owners to either panic over improvements or celebrate declines, both for the wrong reasons.
Build this into whatever regular check in you already do with your site's numbers, weekly, monthly, whatever fits your schedule, rather than treating it as a separate task. The goal is not to spend more total time in Search Console, it is to spend the same amount of time looking at the parts of the data that answer the question you care about, how is this specific page or this specific term doing, instead of a blended figure that answers a much vaguer question you probably were not asking in the first place.
Related Reading
More from Hillcane on this topic, plus the pages on the site that sit next to the work.
Frequently Asked Questions
What does average position measure?
Search Console Help defines position as an estimate of the average ranking of your site's URLs for a query, based on the highest ranking URL at the moment each impression happened. Viewed without filters, average position blends that figure across every query, page, device and day in your chosen date range, weighted by impressions. So a query with a thousand impressions at position 20 outweighs one with five impressions at position 1.
Why can average position rise while my important pages fall?
Because the site wide figure is an impressions weighted blend. A query with a large volume of impressions at a strong position can lift the average while the pages you care about most lose ground, and the single number never separates the two. Filter Search Console by page, then by query, to see which is happening. Changes in the mix, such as seasonal terms dropping off, also move the blend without any page changing rank.
Why can average position fall even when the site is improving?
A new page appearing for a long tail search at a low position adds a heavily weighted, low ranking query to the blend. In this post's labeled illustration with made up round numbers, one query sits at position 60 with 100 impressions beside a branded query at position 1 with 10 impressions, so the average lands near 60. Adding another new query at position 80 could worsen it during genuine progress.
What are the three filters I should apply before trusting the number?
First, filter by page, so you see one page's trend instead of a site wide blend. Second, filter by query, since a page can rank well for one term and poorly for another at the same time. Third, compare like with like: the same page or query across two comparable windows, not two periods with different mixes. Keep a short standing list, such as a key local page and your branded query.
Does a lower average position number always mean a site is doing better?
No. It depends on what is blended. A site with ten branded queries all at position 1 and nothing else will show a better average than a site ranking for fifty new, competitive terms still climbing from position 40, even though the second site may be doing more valuable work. The number rewards a narrow set of strong queries, so read it only after filtering by page and query.
Is Hillcane's own Search Console data typical of what other cafes see?
No. It is one small site's pattern, not a benchmark. On hillcane.co, a branded query for the site's own name tends to sit at a strong, stable position, while a newly published page on a narrow topic can sit much lower as it builds a track record. The impressions weighted mechanism described in Search Console Help is general; the specific numbers on any one site are not.
Should I ignore the site wide average position entirely?
Not entirely. A meaningful shift in the site wide figure, in either direction, is a useful prompt to open the page and query level data behind it, but never a conclusion on its own. Pick the handful that matter, such as the page for your most important local search term and the query for your shop's name, and check those trends in the same weekly or monthly check you already do.
How is the illustration of two queries meant to be used?
It shows the mechanism with labeled, made up round numbers, kept apart from real data. In week one, a branded query holds position 1 with 10 impressions while a new long tail query sits at position 60 with 100 impressions. In week two, the branded query stays flat and the new one climbs to 45 with 150 impressions. The site wide average moves even though one query never changed at all.
What should I compare when looking at a position trend over time?
The same specific page or the same specific query across two comparable time windows. Comparing a site wide average from one period against a site wide average from a different period, where the mix of pages and queries has shifted, can move the number for reasons that have nothing to do with any individual page's real performance.
Does this post claim any specific fix will improve my average position?
No. This post explains why average position misleads when read without filters and offers a routine for reading it accurately: filter by page, filter by query, and compare the same page or query across comparable windows. It predicts nothing about rankings and promises no change will move your position, site wide or for any single page. Treat the number as a direction to investigate, not a grade.
Work with Hillcane
Average position blends every query, page, device and day into one number, weighted by impressions. It can rise while the pages that matter fall. Filter before you believe it.
If you want help reading your own Search Console data page by page instead of guessing from one blended number, reach Hillcane at hillcane.co/contact or call (256) 384-2449.
Comments