DEV Community

Just a Side Project
Just a Side Project

Posted on Originally published at justasideproject.blogspot.com

Debug Log #12: The Impressions Number That Doesn't Add Up

Two routine status checks, two days apart, and the same Search Console summary query returned the exact same number both times: 2 impressions, 0 clicks. Nothing unusual about a flat number on a near-zero-traffic blog -- except that the page-level breakdown for the same date range adds up to 4, not 2.

Check date Total impressions (summary query) Sum of per-page impressions
2026-11-09 2 4 (1 + 1 + 2)
2026-11-11 2 4 (1 + 1 + 2)

The per-page rows, identical both times:

Page Clicks Impressions
Homepage (/) 0 1
I Checked Google's Own Billing Console... 0 1
My Service Account Key Got Blocked... 0 2

Where the math doesn't close

1 + 1 + 2 is 4. The top-line summary query says 2. Same site, same 28-day window, same account, run twice with a two-day gap in between and no new activity in between to explain a change either way. That's not a rounding difference -- it's double the reported total.

What I checked before writing this up

The query code itself is the same two-call pattern used throughout this series: one call with no dimensions for the summary, one call with dimensions: ["page"] for the breakdown. Same site URL, same date range computed fresh each run. I don't have a confirmed explanation. The most boring one is that Search Console's own documentation warns the API can return slightly different totals across dimension breakdowns due to internal sampling and deduplication on very low-volume properties -- which would mean this is expected behavior on a blog this small, not a bug in anything I wrote. I can't confirm that from here without more data points than two impressions a month produces.

Why I'm posting a debug log with no fix

Debug Log #10 set the precedent for this: write down what's actually observed, say plainly what isn't confirmed, and don't manufacture a tidy explanation to fill the gap. A discrepancy that's reproduced identically twice is worth recording even when the dataset is too small to debug further. If it's still there next time impressions tick up, that'll be the moment to actually chase it.

Related reading

Top comments (0)