SEO performance measurement: why average position can mislead
Quick answer
Read SEO performance by relevant queries and pages, alongside clicks and impressions. Your site’s average position can fall simply because you’re appearing for more new searches. Check which rankings changed before deciding that an edit made things worse.
One average can hide two different stories
Suppose an established page receives 100 impressions at an average position of five. A new page then receives 900 impressions around position 40. A simplified impression-weighted average across those observations is 36.5. The headline number looks much worse, although the first page has not moved.
This is an illustrative calculation, not a reconstruction of how every Search Console aggregation works. Its purpose is to show the effect of changing composition. A site that expands into new topics can acquire many low-ranking impressions before it earns clicks.
Google’s Search Console explanation of performance data defines position using the highest relevant result in the applicable aggregation. It is not a permanent rank attached to an article. Country, device, query and search appearance all affect what is being averaged.
Create a stable comparison group
Separate brand searches, established relevant non-brand queries, new relevant queries and irrelevant legacy traffic. Compare each group over matching periods. If your publication changes focus, record which old topics are being retired from editorial promotion instead of treating their decline as a failure of the new strategy.
For a marketing technology publication, a query about CRM lead handoffs belongs in the core group. A query about the general purpose of software may not. High impressions alone are not a reason to build more content around an audience you do not intend to serve.
Keep a fixed cohort of important query-page pairs for diagnosis. Add a second report for newly discovered queries. The first tells you whether existing visibility is improving; the second tells you whether the site’s reach is broadening. Neither should replace the other.
| Observation | Possible explanation | Next check |
|---|---|---|
| More impressions, worse average position | New query coverage | Fixed query cohort |
| Fewer clicks, stable impressions | CTR or search-result changes | Query and device segments |
| URL loses traffic after migration | Canonical or redirect change | Old and new URL records |
| Large percentage change on few visits | Small-sample volatility | Absolute counts and a longer window |
Read the metrics together
Falling clicks with stable impressions can reflect a weaker result snippet, a change in intent, more competing search features or ordinary variation. Falling impressions and clicks together may suggest lower demand or lost visibility. A changing average position is one clue, not the verdict.
Compare country and device before rewriting a page. A rise in impressions from a new market can alter the totals substantially. Check the landing URL too: redirects, canonical changes and duplicate pages can move reported performance between URLs.
Our marketing analytics guide argues for reports built around decisions. An SEO review should end with a specific action, such as improving a comparison section, fixing an indexing issue or leaving a stable page alone while more data arrives.
Keep an edit log and an observation window
Record what changed, when it went live and which pages should benefit. Distinguish technical repairs from changes in topic or content. When several large edits happen together, it becomes difficult to attribute a later movement to any one of them.
Use settled data for comparisons and label incomplete periods. For low-traffic pages, a few days may contain too few impressions to support a decision. Choose the review window around traffic volume and the business question; there is no universal number of days that makes a result conclusive.
Avoid repeatedly changing titles in response to tiny movements. A sensible title should describe the page and its value accurately. If the query mix is wrong, improving the page’s actual scope and internal links may matter more than another variation of the headline.
Build a report you can act on
Keep four things visible: relevant searches, the visits they bring, technical problems and what those visits help the business achieve. Show the sample size and separate legacy traffic. Otherwise an impressive total can reward exactly the wrong work.
For relevant visibility, track impressions and position bands for the fixed cohort. For visits, show clicks and click-through rate. For availability, record indexing and canonical problems. For usefulness, inspect engaged visits and appropriate actions where measurement is available, without treating analytics omissions as proof that no one visited.
Use the right chart for each comparison and preserve the limitations of small datasets. A goal of reaching the first page for valuable searches is actionable. A goal of forcing the entire site’s average below ten can encourage deleting useful early-stage visibility from the report rather than improving the publication.
Related reading
GEO for marketing teams: make content useful enough to cite · UTM naming conventions: a campaign tracking system people will use · Marketing budget allocation: fund the next useful decision
Photo: Luke Chesser / Unsplash.
