How to Measure Your SEO Rankings (and Why Average Position Misleads)
Table of contents
Search Console's average position is not your position: it is the average of your site's topmost result in every search where you appeared. Understanding that is what separates a useful report from an argument with your client.
What Search Console's Average Position Actually Measures
Before reading any chart, it pays to know what sits behind the number, because almost every discrepancy that later looks mysterious comes straight out of that definition.
The Official Definition: Your Site's Topmost Result in Each Search
Google's help documentation is literal: average position is the average position of the topmost result from your site. If your site shows three URLs for the same search, only the best one counts, not an average of all three. It is not the position of your page or of your keyword either: it is an impression-weighted average across every context where your site was shown. A query with ten impressions and one with a thousand do not carry the same weight.
Aggregating by Property Is Not the Same as Aggregating by Page
The report aggregates data in two ways, and both are correct. The chart aggregates by property: the position of the topmost result from the whole site. The table aggregates by the dimension in each row, so filtering by URL shows that page's position.
Google's own example is clear: one search returns three URLs from the same site at positions 1, 2 and 3. Aggregated by property, average position is 1, there is one impression and CTR is 100%. Aggregated by page, average position is 2 (the average of 1, 2 and 3), there is one impression per URL and CTR is 33%. The documentation warns that CTR and average position are generally higher when aggregating by property if multiple pages from the same site appear.
No Impression, No Position
An impression means a user saw a link to your page or site. If a query never produced an impression, there is no position to average and the cell stays empty. Not every report measures the same thing either: Discover reports clicks and impressions but no position.
Why Average Position Doesn't Match What You See in Google
There Is No "Objective" Position
Google documents it plainly: relevancy is determined by hundreds of factors, which may include the user's location, language and device. Its own example is that searching for bicycle repair shops shows different results to a user in Paris than to a user in Hong Kong. Ahrefs puts it just as clearly: there is no truly objective position in any search engine, and you should expect discrepancies from volatility, time lag and personalization.
In practice, opening an incognito window from your own home does not show you the position: it shows you one snapshot of one context out of millions.
Anonymized Queries and the 1,000-Row Table Limit
Google omits rare (anonymized) queries from the table to protect searcher privacy, but those queries still count in the chart totals unless you apply a query filter. The table also displays a maximum of 1,000 rows, so the long tail stays out of the table and inside the total. That is why numbers do not add up between views, and it is neither your mistake nor the tool's. The consequence is uncomfortable: if your report rests only on the filtered table, you are measuring a subset.
SERP Features, AI Blocks and Ads
The more ads, local packs, people-also-ask boxes, videos or generated answers sit above the organic results, the lower your organic result lands even though your position has not changed. The same position number performs very differently depending on layout: position explains, it does not decide. The article on what CTR is in SEO and how to measure it goes deeper.
Preliminary Data, a Two to Three Day Lag and Pacific Time
The most recent part of the chart is preliminary, shown with a dotted line, and can change within hours. Google states that data is usually available in two to three days and that the report tracks daily data in Pacific Time. And the tool gets things wrong too: a logging error between May 13, 2025 and April 27, 2026 affected impressions, CTR and average position without touching clicks.
How to Measure Your SEO Rankings Without Fooling Yourself
Segment Before You Read the Number
Filter by query, page, country, device and search type: text web, multimodal web, image, video or news. Mixing search types is the fastest way to read a meaningless number. Always compare the same universe: same property, same search type and same filter.
Use 28- and 90-Day Windows, Never Day by Day
The 24-hour view exists, but it is the noisiest. Below the top 5, positions swing easily: results at 10 to 20 can differ by five positions from one update to the next with nobody touching anything. Practical rule: 28 days to decide and 90 days to read a trend, comparing equivalent periods.
Look at Clicks and Impressions First
Position is a diagnosis, not the outcome. Two sites can share the same average position while one gets triple the clicks. And you can move up in position while losing clicks if the SERP layout changed, so CTR by query and by page tells you whether the move was worth anything.
Keep a Log of Measurement Changes
Write down when you switched tools, properties, filters or countries. Half of the position drops reported as problems are changes of measuring tape. For the broader diagnostic tree, the article on diagnosing a traffic drop in Google covers the remaining causes.
Rank Trackers: What They Measure and When They're Worth It
What a Rank Tracker Does (and Why It Isn't Search Console)
A rank tracker runs a controlled query per keyword, location, language and device, and returns one position for today in that context. Search Console does not pick contexts: it averages what actually happened.
| Source | What it measures | What it's good for |
|---|---|---|
| Search Console | Real impression-weighted average across all contexts | Diagnosing clicks, CTR and visibility |
| Rank tracker | One position today for a fixed context | Comparing keywords and competitors |
| Manual incognito check | One personal, personalized snapshot | Curiosity, never reporting |
Frequency and Sampling
The default update frequency for a serious tracker is weekly: the top 5 is stable while 10 to 20 is highly volatile. Checking daily does not add precision; it adds noise and worse decisions.
When You Don't Need a Tracker
For a local business or a small site, Search Console plus a list of 20 to 50 intent-driven keywords and a close reading of clicks and impressions is usually enough. A tracker earns its keep with multiple countries, multiple branches, competitor tracking or reporting to a client.
DIY Alternative: API Plus a Spreadsheet or BigQuery
The API returns the position field as the average position in search results, enough to build your own series and alerts without a subscription. The limit is the report's data window of roughly 16 months: for permanent archives, the bulk export to BigQuery, as explained in exporting Search Console to BigQuery.
A Five-Step Rank Tracking Method
- Pick keywords by intent, not volume. Between 20 and 50 terms: brand, service or product, and long tail with buying intent.
- Fix location, language and device, then never change them. Context is the unit of measurement.
- Save a 28-day baseline before touching anything. Without a baseline there is no comparison, only impressions.
- Review weekly, decide monthly, document every change.
- Cross-check with clicks, conversions and revenue. If position rises and the business does not, position was not the problem.
Four Mistakes That Invalidate Any Ranking Report
Comparing Different Universes
Same query set, same property, same search type. Change any one of the three variables and the comparison becomes an anecdote.
Celebrating a #1 With No Demand
Ranking first for a keyword with ten searches a month does not pay invoices. Position matters in proportion to volume and intent.
Mixing Property Position With URL Position
The property's 1 gets reported as if it were the performance of a single page, and then nobody understands why that page brings no traffic.
Swapping Tools Mid-Series
If you migrate to a new tracker or property, overlap both sources for at least four weeks and note the cut-off date. Without that overlap, the drop you see belongs to the instrument.
Average position is there to explain why clicks change, not to chase a number. Measure with the same method, in the same window and the same context, and decide with clicks, conversions and revenue. For the full framework of metrics, the article on measuring the success of your SEO strategy with key KPIs is the logical next step.