Average position gets worse while Search Console impressions grow
Separate a change in Search Console query mix from a genuine ranking decline, compare consistent groups, and assess useful clicks and outcomes with clear limits.
By Crawlito · 6 min read
In short
- An aggregate average can worsen when new lower-position impressions enter the mix.
- Compare consistent query and page groups before claiming that existing rankings fell.
- Do not average row averages blindly or reconstruct hidden query data from visible rows.
On this page
- Record the scope of the position metric
- Use a simple example to understand the mixture
- Build comparable query and page groups
- Check other composition changes before naming a cause
- Avoid misleading calculations from exported averages
- Review whether the new exposure produces useful visits
- Report the finding with the right level of certainty
- Frequently asked questions
If Search Console's average position gets worse while impressions grow, check whether the site is appearing for a different mix of queries before concluding that existing rankings fell. New impressions at lower positions can change the aggregate even when an established query stays stable. Compare consistent groups and review useful clicks and outcomes alongside the position metric.
This does not mean every decline is harmless. The aggregate is a starting signal; the question is whether comparable pages and queries changed, or whether the composition of the measured impressions changed.
Record the scope of the position metric
Save the property, search type, date ranges, filters, and grouping used in the report. A site-level view and a page-filtered view can represent different aggregations.
Google's Search Console data explanation describes how property and page aggregation affect metrics, including position.
Do not treat average position as a fixed rank that every person sees for every search. The report summarizes observations under its documented rules.
Keep clicks and impressions visible while inspecting position. A change in the amount and kind of exposure is central to this diagnosis.
Also note recent content launches or changes in query coverage. A new section can introduce many impressions without immediately occupying the same positions as an established branded page.
The Search Console query totals guide explains why visible query rows do not necessarily provide a complete reconstruction of the report.
Use a simple example to understand the mixture
Consider a deliberately simplified hypothetical dataset with two query groups at the same aggregation level.
In the first period, an established group receives 10 impressions at an average position of 2. In the second, it still receives 10 impressions at position 2, while a new group receives 90 impressions at position 20.
First period: 10 impressions, average position 2
Second period:
Established group: 10 impressions at position 2
New group: 90 impressions at position 20
Illustrative weighted average:
(10 * 2 + 90 * 20) / 100 = 18.2The aggregate looks worse even though the established group's position did not change. The example illustrates composition; it is not a claim about your site's performance.
Do not apply this arithmetic blindly to downloaded rows from different aggregation levels or incomplete query tables. The source's reporting rules and missing rows still matter.
The useful lesson is to inspect what entered the mixture before attributing the entire change to a loss in existing positions.
Build comparable query and page groups
Choose a stable set of important pages or query themes and compare them across complete periods using the same filters.
Keep branded and nonbranded questions separate when that distinction helps the business question. Their position and intent patterns can differ substantially.
Use Google's performance-report use cases for the supported comparison workflow.
If using regular expressions to define groups, test the expressions against known matching and nonmatching examples. The regex filter guide explains how an accepted expression can still select the wrong URLs.
Record the exact group definition so a later reviewer can reproduce it. A group that changes silently between periods cannot support a clean comparison.
Avoid choosing only the queries that improved. Include a consistent, defensible set and label coverage limitations.
A stable group helps answer whether a particular established area changed; it does not automatically explain every new impression outside that group.
Check other composition changes before naming a cause
Compare relevant device, country, page, and search-appearance segments when the available report supports them. A different distribution across those dimensions can also affect the aggregate.
Keep one analytical question at a time. Adding many filters until the numbers look familiar can create a selection bias rather than an explanation.
Review whether a new page began receiving impressions for broad questions that it only partially answers. That can be an editorial opportunity without proving that a technical defect caused the position change.
Inspect the public page where the content question is concrete. Does the opening answer the intended query, and does the page provide the promised evidence or next step?
Do not infer an algorithm update from a date coincidence alone. A credible cause needs supporting evidence beyond a changing summary statistic.
If the established comparison group actually declined, continue investigating its content, availability, competing intent, and relevant site changes rather than dismissing the finding as mixture.
Avoid misleading calculations from exported averages
Do not calculate a simple mean of row-level average positions and present it as the property's reported average. Rows can carry very different impression counts.
Even impression weighting requires compatible scope and sufficient data. Missing query rows, property-versus-page aggregation, and filter behavior can prevent a visible table from reproducing the headline metric.
Keep exported dimensions, metrics, filters, and dates with the calculation. A number without those definitions is difficult to review.
If you calculate a custom statistic for a known subset, label it as that subset's statistic. Do not rename it “the site's average position” unless the underlying scope supports the claim.
Preserve the original platform values alongside any derived analysis. This helps distinguish a real change from an aggregation error introduced in a spreadsheet.
A transparent comparison of a stable subset is more useful than an exact-looking overall estimate built from incomplete rows.
Review whether the new exposure produces useful visits
Look at clicks and the relevance of the new queries before deciding what action to take. More impressions for unrelated questions may not help the app.
For pages that attract appropriate visitors, examine accepted outcomes using the available measurement design. The conversion measurement guide distinguishes completed actions from superficial interactions.
Do not join individual Search Console queries to individual users unless the actual data and supported methodology establish that relationship. Aggregate search reports do not reveal every visitor's query-to-conversion journey.
Keep attribution limitations visible when comparing articles. A guide can contribute to a later action that a simple last-touch report does not fully describe.
Avoid rewriting a useful page solely to improve one aggregate position number. Prefer a concrete diagnosis: the answer misses an important condition, the next step fails, or the page is attracting a different intent than intended.
Measure changes using a consistent prospective period after the definition is settled.
Report the finding with the right level of certainty
A defensible conclusion might say that established query groups remained broadly stable while new lower-position impressions changed the overall average. State the groups and periods that support it.
If the data is incomplete or mixed, say that the aggregate alone cannot establish whether existing positions declined. That is a useful limit, not a failed analysis.
For a client report, show the headline metric alongside a small number of relevant comparable groups. Include impressions and clicks so the reader can understand the composition.
Do not claim that worse average position is always good because it can accompany broader exposure. New impressions still need relevance, and real declines can occur at the same time.
Keep the decision tied to a specific action or observation: inspect a newly visible topic, improve an incomplete answer, repair a failing page, or continue monitoring a stable group.
The goal is to avoid reacting to a summary number as though it explains its own cause.
Frequently asked questions
Can average position worsen without an established query falling?
Yes. A different mix of impressions can change the aggregate. Compare consistent groups to test whether that explains your report.
Can I average the position values in my exported rows?
A simple mean ignores different impression counts, and even weighted calculations need compatible and sufficiently complete data. Label any custom subset calculation clearly.
Should I ignore the decline if impressions increased?
No. Inspect comparable groups, relevance, clicks, and available outcome evidence. Increased exposure and real losses can occur together.
Use the free Crawlito check on a page whose useful search traffic changed, then assess its technical findings alongside the properly scoped performance comparison.
Filed under Strategy and audits. Topics: google, search-console, analytics