Ranking third. Almost no clicks. What is your average hiding?
By Plain Reckoner · 1,327 words · about 6 min · Updated
An average position of 3 can coexist with only three clicks: in our hypothetical example, 90 impressions at position 1 and 10 at position 21 produce that average. Separate query intent, pages, countries and devices before treating a strong headline position as broad search visibility.
- How it was measured
- A complete, hypothetical 100-impression example with impression-weighted arithmetic, not observed publisher performance. Reporting definitions and diagnostic guidance checked against primary Google documentation on 2026-10-21. No universal CTR curve, search-volume estimate or promised traffic increase is used.
- Last checked
Your Search Console average position looks excellent. Your traffic does not. Before rewriting every title or buying another SEO tool, ask a more useful question: which searches produced that average?
An impression means your result was shown in search. A click means someone followed it. Click-through rate, or CTR, is the share of impressions that led to clicks.
A position is not an audience size. A small number of searches for your own brand can look impressive while the questions your future readers ask remain almost invisible. Equally, a relevant page might appear often but fail to earn clicks. Those situations need different work.
This guide separates them with one reproducible example, a diagnostic table and a practical investigation. It does not tell you what click-through rate every website should achieve.
How a position of 3 can hide a position of 21
The following is an invented teaching example, not Plain Reckoner traffic, a benchmark or a study. Assume these two groups account for every impression in the example and use the same reporting scope.
Swipe or scroll the table horizontally to see every column.
| Query group | Impressions | Average position | Clicks |
|---|---|---|---|
| Searches for the site's own brand | 90 | 1 | 3 |
| Searches for the problem its guide solves | 10 | 21 | 0 |
| Combined | 100 | 3 | 3 |
The combined position is (90 × 1 + 10 × 21) ÷ 100 = 3. The overall click-through rate is 3 ÷ 100 × 100 = 3%.
Nothing in that calculation puts the guide third for its useful nonbrand searches. Those impressions averaged position 21. The strong headline comes mostly from the brand group. Nor does the table prove that 3% is good or bad: its purpose is to show what a blended average can conceal.
Google's definitions of impressions, position and clicks explain that position averages the recorded impressions and uses the topmost relevant result for the reporting scope. It is not a permanent rank for every keyword. Read the report's scope before comparing a whole property with a specific page.
You can reproduce the example from our illustrative CSV. It contains the two invented groups above, not a hidden export of real searches.
Find the bottleneck before choosing the fix
Use this as an investigation map, not an automatic diagnosis.
| What you observe | Question to answer next | A useful first action |
|---|---|---|
| Good position, few impressions | Does this visibility mainly come from narrow or branded searches? | Compare visible brand and nonbrand query groups |
| Relevant impressions, few clicks | Does the result offer the answer the searcher wants? | Inspect the actual result, title and landing-page opening |
| More impressions, worse average position | Are new queries broadening the mix? | Compare the same established queries and pages across periods |
| One country performs differently | Are the query, language or device mixes different? | Compare like-for-like segments before changing content |
| Clicks look healthy, visits do not match | Are the tools counting different events? | Reconcile dates, channels and measurement definitions |
Start with the page that already answers a useful question. A site-wide average can conceal a promising article as easily as it can conceal a weak one.
1. Keep the comparison stable
Choose a complete period and a comparable earlier period. Record the search type, date range and filters alongside your export. An incomplete week is a poor comparison with a complete month. A page filtered to one country is a different scope from an unfiltered property.
Keep an untouched original export. Make notes in a separate sheet: what changed, when, which page was affected and which metric you expect to move. This turns the next check into a comparison rather than a memory test.
Avoid changing several pages, titles and navigation patterns at once simply because a tiny total moved. With small counts, a handful of visits can produce a large percentage change without showing a durable pattern.
2. Separate people who know you from people who need you
Look at the query list and distinguish searches for your name from searches for a problem you solve. Review likely brand spellings rather than assuming one exact phrase catches them all. Treat this classification as a working interpretation you can audit.
There is an important reporting limit: Google's dimensions and data-grouping guidance explains that some queries are withheld for privacy and tables can omit data. Filtering queries also affects which totals include anonymized searches.
Do not force visible query rows to reconcile with the unfiltered headline. Our example is deliberately complete so its arithmetic can be checked. Your real export may not be. Label any brand-share estimate with its scope and missing-data limit instead of presenting it as an exact split of all searches.
3. Inspect the search result before blaming the headline
Pick a relevant query that actually generated impressions. Read the page it leads to. Does the title describe the answer? Does the opening deliver it? Would a reader understand why your method, calculator or example is useful?
Then inspect a current result for context. Results vary, so one manual search is not a replacement for your report. It can still reveal an intent mismatch: your detailed methodology may be appearing for a query that wants a quick definition, or a generic introduction may be competing where readers want worked calculations.
Google's Search performance walkthrough recommends examining relevant high-impression, low-click queries, the displayed result and differences by country or device. That is more informative than chasing a universal position-to-CTR chart.
Choose one clear improvement. Replace a vague title with the specific question the page answers. Add the missing example. Move the useful tool nearer the explanation. Keep the promise accurate; a stronger headline creates a worse experience if the answer never arrives.
4. Separate wider discovery from an actual loss
A worse average can accompany broader visibility. If your page begins appearing for additional questions at lower positions, the mix changes even when its established queries hold steady.
Check whether the same important page-query combinations lost impressions or clicks. Record newly appearing queries separately. Otherwise you risk deleting a useful expansion because the blended number looks less flattering.
When a real decline is present, use Google's search-traffic drop investigation to consider technical access, demand and ranking changes. It explicitly warns that changes need time and do not guarantee gains. A clever interpretation of an average is not a substitute for checking whether the page is reachable and useful.
5. Do not equate a search click with a recorded session
Search Console describes search interactions; on-site analytics describes activity after arrival. Their totals need not match. Google's guide to comparing Search Console and Analytics discusses these differences.
Before attributing a discrepancy to lost visitors, match the period, destination and relevant traffic channel. Check what your analytics setup can actually observe. Keep clicks, sessions, users and pageviews labelled separately when moving figures between reports.
This matters for money too. A search-position average cannot tell you your ad revenue. Our traffic-up, revenue-down guide separates traffic and earnings per pageview. Use matching dates and units in the free RPM comparison before drawing a financial conclusion.
Leave the investigation with one testable decision
Write a short record: the page, the query group, the complete comparison periods, the observed counts, the missing-data caveat and one proposed change. Decide what evidence would make you keep or undo that change.
You can check your site's public signals without an account. The scan does not access private Search Console reports or prove why Google showed a result. If the reports still do not make sense, request a free manual review and describe the specific mismatch. Do not send passwords or private access tokens.
The goal is to earn more useful visits by answering a real question well. An attractive average is only one clue.
Tracked as SEARCH-POSITION-EXAMPLE. If a number here looks wrong, tell us and we will check it against the source.