DEV Community

ricco020
ricco020

Posted on

One Bing query was 62% of my impressions and never got a click

I run a small site about self-hosted WireGuard, and I compute my own CTR curve from the Bing Webmaster API instead of borrowing an industry one. Today the curve said something absurd: queries phrased as questions, at positions 6 to 8, convert at 0.037%. Non-question queries in the same band convert at 6.3%. That is a factor of 170.

It is not a behaviour of question searchers. It is one query.

The numbers

GetQueryStats for the whole property, 12 June to 30 October 2026:

  • 85,371 impressions, 1,467 clicks, so a site CTR of 1.72%
  • one query, what is a vpn split tunneling, accounts for 52,680 impressions and 0 clicks
  • that is 61.7% of every impression the site received in Bing

Remove that single row and the site CTR becomes 4.49%. Same pages, same titles, same month.

The weekly rows show its shape:

week starting impressions clicks avg position
2026-08-14 183 0 9
2026-08-21 2,481 0 8
2026-08-28 6,327 0 8
2026-09-04 10,333 0 8
2026-09-11 19,517 0 7
2026-09-18 7,562 0 6
2026-09-25 6,277 0 6

Then nothing. Seven weeks, a climb to almost 20,000 impressions a week, never one click, then gone.

Why this matters more than it looks

GetQueryPageStats says Bing showed one page for it, our split tunneling explainer, for 49,356 of those impressions. If you look at that page's report you see a page with a catastrophic CTR at position 7, and the obvious move is to rewrite its title.

Its title has been rewritten three times since June. None of that could move this row, because whatever produced those impressions was not a person choosing between ten blue links. I do not know what it was (a rank tracker, an answer box that satisfied the question, an automated surface), and I am not going to pretend I do. What I can say is that 52,680 impressions with zero clicks at position 6 is not a distribution humans produce.

The rule I took from it

  1. Never build a CTR curve without looking at the top contributor of each cell. One row can own a band. Ours owned 97% of the "question, position 6-8" cell.
  2. Report site CTR with and without the largest query. If the two numbers differ by more than 2x, the average describes one query, not your site.
  3. A title rewrite is only a fix when the impressions are human. Check the weekly shape first: a smooth climb to five figures with zero clicks, then a cliff, is a signature, not an audience.
  4. Bing rows are per query per date bucket, not per query. Summing them without grouping by query is how you miss that a "cell" is a single query repeated seven times.

The page may or may not need work; this row cannot tell me either way. The thing that was broken was my instrument.

Top comments (0)