CTR Calculator: Expected CTR by Position

The top panel solves for one of three things. Choose CTR and it divides clicks by impressions: 40 clicks from 2,000 impressions is 2.00%, and the line under the answer shows the sum, “40 clicks ÷ 2,000 impressions × 100”. Choose Clicks and it multiplies impressions by a CTR: 2,000 impressions at 2.5% is 50 clicks. Choose Impressions needed and it divides clicks by a CTR and rounds up: 40 clicks at 2.5% needs 1,600 impressions.

The third sum is the one for planning. Say a page gets 8,000 impressions a month at a 3% CTR, which is 240 clicks, and you want 500. From the same impressions you would need a 6.25% CTR, more than the default table expects at position 5. At 3% you would need 16,667 impressions. The first asks for a better snippet, the second for more rankings.

Counts accept thousands separators (1,200), and the CTR takes a decimal point or comma (2.5 or 2,5). The CTR must be above 0 and at most 100, and clicks cannot exceed impressions.

An impression is counted when a link to your page appears on the page of results someone loaded. Google’s help page on impressions, position and clicks says an ordinary result does not need to be scrolled into view to count. Results inside carousels and expandable sections usually do. So a result at position 9 collects impressions from people who never scrolled that far, which is one reason CTR falls so steeply down the page.

Position is an average of the topmost position your site held each time it appeared. A row at 4.6 might be a page that sat at 3 on some days and 7 on others, or two of your pages taking turns.

The export holds at most 1,000 rows per table. On a bigger site, export in slices: filter to a folder, a country or a device, and export each slice. Queries that Google treats as too rare are anonymized and left out of the query table to protect searchers’ privacy, so the query rows usually add up to fewer clicks than the totals above the chart.

The CTR column in the export is rounded, so the tool ignores it and divides clicks by impressions itself. Rows with no impressions, or with more clicks than impressions, are skipped and counted in a note.

The tool reads the standard Performance export as it comes.

From a ZIP, the tool reads Queries.csv and Pages.csv, skips the other files, and opens on the queries; the File menu switches between them. Upload Dates.csv or Countries.csv on its own and it tells you so, rather than reading dates or countries as queries.

Exports from other locales work. Comma, semicolon and tab separators are all detected, and column names are recognized in English and several other languages, such as Klicks, Impresiones and Posizione. A five-column file with names it does not know is read in Search Console’s standard column order, and a note says so. If your Search Console runs in another language, the ZIP’s file names may not be Queries.csv and Pages.csv; pick the right file in the File menu, or upload that CSV on its own.

The file is read in this browser tab and never uploaded. The limit is 20 MB.

Each row is judged against an expected CTR for its position. By default that is a table of 20 numbers in the tool’s code. They are not taken from a study. The code marks them as assumptions, round numbers picked as a starting point with no source attached, and the page says so above the table.

Positions between whole numbers are interpolated in a straight line. An average position of 2.5 expects 13%, halfway between 16% and 10%. Position 4.6 expects 5.8%, and 10.5 expects 1.25%. Rows past position 20 are not compared at all.

Why not use a published curve? Because published curves disagree. CTR studies use different data, years, countries, devices and query mixes, and their figures for the top result differ by wide margins. The results page moves the numbers too: ads, AI Overviews, map packs, videos and People also ask boxes push the first ordinary result down, and which appear depends on the query. Google publishes no expected CTR by position. Any single curve, this one included, is a rough average that will be wrong for many of your queries.

So treat the defaults as a filter, not a verdict. Edit any value in the Expected CTR by position panel and the results update at once; Reset brings the defaults back. Your edits are not saved, so note them if you settle on your own table.

Set Compare against to “Your site’s median CTR at each position” and the tool builds a curve from your own export. For each whole position from 1 to 20, it takes every row whose average position rounds to that number and uses the median CTR of those rows. Each row counts once, however many impressions it has.

A position needs at least 5 rows before its median is used; rows at a position with fewer are not compared, and a note counts them. The Expected CTR panel then shows a “Your site” figure under each default value.

This mode does not interpolate. A row at 4.6 is compared with the median at position 5, and a row at 4.4 with the median at position 4.

Remember what a median means: about half the rows at each position sit below it by definition. In this mode the flagged rows are the weaker half of your own results, not proof of a fault, so lean on the missed-clicks order to pick the few worth your time.

Two things distort a site curve. Branded queries earn very high CTR at position 1, which raises that median and flags the non-branded queries around it. Filter your brand out before you export: add a Query filter, choose Custom (regex), set it to exclude matches, and enter your brand name and its common misspellings. And the curve is built from every row in the file, including rows under the impressions minimum, so a handful of tiny rows can move the median at positions with little data.

For every row with enough impressions at position 20 or better, the tool compares the actual CTR with the expected one. If the actual CTR is lower, the row is flagged and its missed clicks are impressions × (expected CTR − actual CTR). That is the extra clicks the row would have earned over the export’s date range at the expected CTR.

Rows are sorted by missed clicks, highest first, with impressions breaking ties. That order matters more than the size of the CTR gap. A query at 0.5% instead of 2% on 300 impressions misses about 5 clicks. A query at 4% instead of 7% on 20,000 impressions misses 600. The second gap is smaller and worth far more.

The Minimum impressions field, 100 by default, keeps out rows too small to judge. At 40 impressions a single click moves CTR by 2.5 points. Raise the minimum on a large site to cut the noise; lower it on a small site if too few rows are left.

Four figures sum up the run: Rows read, Compared, Below expected and Missed clicks. A sentence under them accounts for rows not compared: unreadable, under the minimum, past position 20, or without a site median. The table lists the top 50 flagged rows; Show all lists the rest, and Download as CSV saves them all as below-expected-ctr.csv.

Missed clicks is an estimate built on whatever expected value you chose. Change the curve and it changes. Use it to rank rows against each other, not as a forecast of the traffic a rewrite will bring.

These values are illustrative, not real data. A plumbing company exported three months of queries and uploaded Queries.csv with the default curve and the 100-impression minimum. Six rows went in.

The summary reads 6 rows read, 4 compared, 3 below expected and 276 missed clicks. The three flagged rows need three different responses.

Only the first is a snippet job. Under 1% at an average of 4.6 on 4,200 impressions is low against almost any curve you might pick, and the row holds three quarters of the missed clicks. The query is a task, so the title should promise the fix: something like “Running Toilet? How to Fix the Flapper or Fill Valve Yourself”. Search the query first, though. If an answer box already lists the steps, expect a smaller gain.

The second is flagged, but its fix is ranking. At 14.2 the page is mostly on page two, where a better title has few readers. Improve the answer on the page and link to it from related posts.

The third is close enough to leave. 23.33% against 25.80% at position 1.3 is well within the error of a round-number curve.

The branded row, at 45%, would lift the position 1 site median for every other query, so filter brand out before you switch baselines.

A flag says a row earns fewer clicks than its position suggests, not why. Search the query in a private window, on the device that brings most impressions, before you change anything.

Titles and descriptions belong to pages, not queries. When Queries.csv shows a problem, switch the File menu to Pages.csv, or filter Search Console to the query and open its Pages tab, to find the URL that earns the impressions. Then read that page’s flagged queries together. They show the words people typed and what they expected to find, which is what the new title should answer.

Look at Pages.csv for patterns too. If most posts in one category sit below expected, a shared title template is the likely cause, and one change fixes them all.

Change one page at a time, or one template at a time, so you can tell what worked. Clear the page cache after saving.

Write down the date you published the change. Google has to recrawl the page before the new title can appear, and Search Console data usually runs a couple of days behind. Search the query yourself to see when the new title shows; count from that day, not the day you saved.

Compare equal periods. In the date picker, choose Compare, and set the weeks after the change against the same number of weeks before. Filter to the page and read the queries that were flagged.

Judge CTR only where position held steady. If the average moved from 4.6 to 3.2, CTR would rise anyway, and the comparison mixes two effects. As a rule of thumb, if position stayed within about half a place and CTR rose, the snippet did the work.

Give it enough impressions: a query with 300 a month needs several weeks before a few extra clicks mean anything. For seasonal queries, compare with the same weeks last year.

Then export a full period again and run it with the same curve and minimum, so missed clicks are comparable. A row that leaves the flagged list, or drops far down it, is the result you wanted.

A flag is a question, not an instruction. Leave the row alone in these cases.

Rows above the expected CTR need no attention either. A high CTR at a low position is a hint that the page deserves to rank higher, and is often a good candidate for a content update or more internal links.