Paste a CSV exported from the Search Console Performance report and this tool sorts the rows into those worth a second look. It flags rows with no clicks, rows whose click-through rate is well below other rows at the same position, and rows sitting just outside the top three.
It never connects to your account. The file you export is the only input, which keeps the data in the tab and means the tool sees exactly the date range, country and search type you chose.
What the file needs
One row per query or page, with a name column and an impressions column. Clicks, CTR and position are optional but make the output useful. The header is detected by looking for clicks or impressions, and the other columns are matched by name: query, top queries, page, landing page, url, ctr and position are all recognised.
If the first row does not look like a header, the tool reads the columns in the order name, clicks, impressions, CTR, position and warns you that it did so. That fallback is useful for a hand-made file and risky for anything else, because a reordered export will be read against the wrong headings without complaint.
Numbers may contain commas or a percent sign. CTR written as 1.11% and CTR written as 1.11 are treated the same, since the symbol is stripped before parsing.
The three reasons a row is flagged
No clicks. Any row with zero clicks is flagged, with severity 3. It had impressions, so it appeared in results, and nobody clicked.
Weak CTR for its position. The tool splits the export into four position bands, works out the average CTR of each band within your own file, and flags a row when its CTR is below 60% of that average. Severity 2. No external CTR curve is used, which is the honest approach: your site's normal rate is a better comparison than a published average built from other sites.
Close to page one. A row with an average position between 4 and 15 is flagged with severity 2. It is already visible, and a small improvement could move it into the top three. Rows are sorted by total severity, then by impressions, then alphabetically, and the table shows the first 15.
Two filters sit above the result: a minimum impressions figure, which defaults to 500, and a maximum position, which defaults to 20. Raising the minimum removes low-volume noise. Lowering the maximum hides rows that are too far down to act on.
A worked example
A ten-row query export produced seven flagged rows, 327 clicks and 24,100 impressions, an average CTR of 1.36%. Every flag came from the position rule rather than the CTR rule, because each row sat between position 4.2 and 12.7.
| Position band | Average CTR | What it suggests |
|---|---|---|
| 1 to 3 | 4% | Two rows, one of them converting well |
| 4 to 10 | 1.42% | Six rows, the bulk of the file |
| 11 to 20 | 0.27% | Two rows, much weaker |
| 21 and beyond | 0% | No rows in this export |
The band averages are the interesting part. The same page at position 5 and position 12 is not the same opportunity, and the export shows how steeply CTR fell between the bands on this site. A row at position 12.7 with 5,200 impressions and 18 clicks was flagged for position, not for its rate, because its rate was normal for that band.
A known quirk in the CTR column
If your export writes CTR as a percentage below 1%, with a percent sign, the table can show it multiplied by 100. In the example, a row with 18 clicks from 5,200 impressions is 0.35%, and the table showed 35%. Rows above 1% are shown as written.
Until that is fixed, read CTR from the clicks and impressions in the same row, or from Search Console itself. The flag reasons are based on clicks, impressions and position, so the list of rows to review is not affected by the display error; the percentage beside each row can be.
What the export cannot tell you
A low CTR has several possible causes: the query, the device, the snippet, the brand, the other results on the page, or a title that promised something the page did not deliver. The tool cannot see any of them. It knows only that the rate is low for the position.
Nor can it join queries to pages unless the export contains both dimensions. The Performance report exports one dimension at a time, so a queries file and a pages file are separate views of the same traffic. Comparing them by hand is possible; treating one as the other is not.
Finally, Search Console totals do not always add up the way you expect. Google's documentation notes that data grouped by queries is aggregated by property, while data grouped by pages is aggregated by page, and that chart totals can differ from table totals for that reason. A row-by-row sum will not always match the site total.
Turning the list into work
Export queries first and look for pages that appear for several related queries at once. Then export pages and check whether the same URL shows the same weakness. Two exports, read together, tell you whether the problem is the snippet or the page.
For a flagged row, open the query in the target market and look at the result as a searcher would. If the title is cut or the description is generic, fix the snippet with the snippet simulator. If the page does not answer the query, no title change will help, and the content optimiser is the better next step.