We opened Google Search Console and saw a clean, worrying number: average position had fallen from 14.1 to 18.
The obvious conclusion was that rankings were slipping. That conclusion would have sent us in the wrong direction.
During the same 28-day period, search impressions increased 32%, clicks increased 16%, and homepage clicks increased 35%. Several commercial queries were also moving closer to page one. The site was not becoming less visible. It was becoming visible for more searches.
What the headline number hid
| Metric | Last 28 days | Previous 28 days |
|---|---|---|
| Clicks | 352 | 304 |
| Impressions | 25,293 | 19,100 |
| CTR | 1.4% | 1.6% |
| Average position | 18.0 | 14.1 |
The homepage was a useful control. Its impressions rose from 7,787 to 9,604, clicks rose from 212 to 286, and its average position improved slightly from 10.8 to 10.6. The page driving most of the business was moving in the right direction.
Why average position behaves this way
Search Console calculates position across impressions. When a site begins appearing for new queries at positions 20, 30, or 40, those impressions enter the calculation. The average can decline even when existing high-value rankings hold steady or improve.
Google describes average position as the topmost position occupied by a property, averaged across the queries where that property appeared. It is an aggregate, not a diagnosis.
How Google defines average positionA worse average can be the side effect of Google giving a site more opportunities to rank.
The query-level picture
This analysis came from PostgresGUI, a native PostgreSQL client for Mac. Its Mac-specific queries were already performing much better than the property average.
PostgresGUI is an Accellsoft-owned product, so the review could move directly from diagnosis to implementation. Our PostgresGUI case study covers the product, free tools, content structure, and search work built around these findings.
| Query | Position | CTR |
|---|---|---|
| postgresql gui mac | 4.7 | 9.1% |
| postgres gui mac | 6.2 | 7.5% |
| postgres gui | 11.1 | 1.1% |
| postgresql gui | 16.3 | 0.6% |
The useful problem was now specific. Mac-focused terms had a credible path into the top three. Broader terms remained on page two, where established products and high-authority community sites were stronger.
Geography revealed another gap. The United States generated 26,423 impressions at an average position of 19.6 and a CTR of 0.3%. Several international markets averaged positions around nine or ten. A single sitewide number had concealed a major market-level difference.
A seven-step check before changing anything
We now use this sequence before treating a position decline as an SEO problem.
Compare clicks and impressions first
If both are growing, a worse average position may simply mean Google is testing the site for a wider set of searches.
Separate important queries from the sitewide average
Track the terms that match the product and buying intent. A thousand unrelated impressions should not outweigh movement on the queries that matter.
Separate branded and non-branded searches
Branded searches usually rank well and can hide weaker discovery performance. Non-branded queries show whether new people can find you.
Inspect the pages receiving new impressions
A new guide ranking at position 35 can lower the property average even while an established product page moves from position six to four.
Break the report down by country and device
Competition and intent vary by market. Our United States performance was much weaker than several international markets.
Look for queries appearing at positions 20 to 50
These are often the source of a falling average and the evidence that Google is beginning to understand a broader content set.
Prioritize pages already sitting between positions 4 and 15
Moving a relevant page from six to three is usually more valuable than trying to rescue every low-volume query from position 40.
What the competitive review changed
Ranking data explained where we were. Reviewing the search results explained why.
Established competitors benefited from years of authority and links from trusted PostgreSQL resources. Newer competitors were publishing unusually thorough comparison pages with original screenshots, current pricing, feature matrices, testing methodology, and recommendations for specific workflows.
Our backlink count looked healthy at first glance, but many of the leading linking hosts were raw cloud-server IP addresses. The meaningful endorsements came from a much smaller set of sources, including Apple, Reddit, Hacker News, and a few directories. The number of links was less important than who was doing the linking.
What we are doing next
- Consolidating overlapping pages so one commercial page owns the main Mac PostgreSQL GUI intent.
- Rebuilding the existing comparison guide with original testing, screenshots, current pricing, and a clear methodology.
- Publishing reproducible product benchmarks instead of another generic list of tools.
- Seeking inclusion in trusted PostgreSQL and Mac developer directories.
- Improving titles and descriptions on pages already ranking between positions four and fifteen.
- Adding site-wide search so visitors can move between product pages, tools, and technical guides without relying on the main navigation alone.
The search implementation is intentionally small. We documented the index, ranking rules, URL state, accessibility, and tests in our Next.js site search article.
Measure the opportunity, not the comfort of the average
Average position is useful as an orientation signal. It is not a target worth optimizing in isolation.
We care more about non-branded clicks, top-three positions for the Mac-specific query cluster, United States desktop CTR, and qualified visits to commercial pages. Those measures describe the opportunity more accurately than one property-wide average.
A declining average should trigger investigation, not panic. Start with the queries and pages behind it. You may find that the metric that looks worse is recording the first stage of wider growth.