In September we cut from fourteen posts a month to four, as a deliberate test. In October we went back to fourteen. This is what Search Console recorded while we did.
It is not the result the October 2 update to our own playbook predicted. We said volume looked like it was driving reach. A month of real data says otherwise, at least so far.
🔬 How this was built
All figures come from the Search Console Performance report and the Page indexing report, both for the property https://buzzriding.github.io/. We pulled three custom date ranges on 6 November 2026 — 1–30 September, 1–31 October, and 1–5 November — plus the live indexing dashboard. The raw numbers and design note are committed alongside this post.
The month in numbers
- September (4 articles): 14 clicks, ~2,650 impressions, 0.5% CTR, average position 36.
- October (14 articles): 6 clicks, ~1,820 impressions, 0.3% CTR, average position 21.3.
- 1–5 November (partial, still processing): 1 click, 390 impressions. Too early and too incomplete to read as a trend.
- Indexed pages: 39 at the end of September, 42 in the most recent Pages report snapshot (dated 23 October).
Three of those four articles a month in September. Fourteen in October. Impressions and clicks both fell in the month with more than three times the output.
Finding 1: impressions fell 31%, clicks fell 57%, in our highest-output month since the audit
October's 1,820 impressions are 31% below September's ~2,650. October's 6 clicks are 57% below September's 14.
This is the opposite of what the 2 October note in our own blog production module argued from the June–September trend. That note said impressions per indexed page had been climbing and volume looked like the driver.
October is the first real test of that claim at scale, and the result went the other way. One month is not enough to overturn a four-month trend, but it is enough to stop treating the trend as settled.
Finding 2: average position improved while total reach shrank
Average position went from 36 in September to 21.3 in October — a real improvement, roughly fifteen positions higher on average.
That sounds like it should mean more traffic, not less. It does not, because average position is only computed across the queries that showed impressions at all.
Fourteen new articles likely replaced some of September's long-tail, barely-ranking impressions with fewer, tighter queries where the newer pages rank more respectably. Better average position on a smaller, more concentrated set of queries is consistent with both numbers moving the way they did.
We cannot prove that mechanism from the aggregate report alone. It is the explanation that fits without requiring either number to be wrong.
Finding 3: 11 new posts, 3 more indexed pages
Between 1 and 23 October we published 11 new articles. The Pages report, snapshotted 23 October, shows the indexed count rose from 39 to 42 over that stretch — a net gain of three.
In the Jul–Aug window, by contrast, indexed pages rose from 9 to 19 on 13 new articles — roughly one indexed page for every 1.4 published. October's ratio is closer to one indexed page for every four published.
Indexing conversion, not writing pace, looks like the actual constraint again — the same conclusion our August discovery-vs-indexing post reached with a much larger gap.
Finding 4: the indexing dashboard itself is stale, and that matters more than the headline number
The Pages report lists two URLs as "Crawled – currently not indexed," both last crawled 15 June 2026: the GEO generative-engine-optimization article and the newsletter signup page.
We ran URL Inspection on both, live, on 6 November. Both now return "URL is on Google" and "Page is indexed."
The aggregate dashboard — 42 indexed, 3 not indexed — is a snapshot from 23 October. At least two of those three "not indexed" pages were already indexed by the time we checked, two weeks later.
The practical takeaway: a site's indexed count is always older than the dashboard implies. Spot-check individual URLs with live inspection before concluding a page has a real indexing problem.
What this means for the September cap decision
The 2 October note said: "Re-check at the November and December metrics runs, and revert the cap if added volume stops adding indexed pages or impressions."
October is exactly that check, one month early. Added volume did not add impressions. It did not add clicks. It barely added indexed pages relative to what was published.
One month against a four-month trend is a lean, not a verdict — the same caveat the predecessor to this post insisted on for its own numbers. We are flagging it, not reversing the cadence cap on the strength of a single month.
How to run this check on your own site
This needs no tool beyond Search Console, which every verified property already has.
- Pull three custom date ranges — the month before a change, the month of the change, and the partial current month — from the Performance report's date picker.
- Read clicks, impressions, CTR and average position together. A position improvement with falling impressions is a real pattern, not a contradiction, once you check the query mix.
- Check the Pages report's "Last update" date. Treat anything older than a week as provisional.
- Live-inspect a sample of "not indexed" URLs before reporting the dashboard count as current.
What we are changing because of this
- Not reverting the cadence cap yet. One month's reversal of a four-month trend needs at least one more data point.
- Stop quoting the Pages report count without a date. "42 indexed" means nothing without "as of 23 October."
- Spot-check flagged URLs with live inspection before submitting indexing requests the dashboard says are needed — two of three turned out to be unnecessary this session.
- Re-run this exact comparison at the December metrics session, with a full month of November data instead of five partial days.
💡 Key takeaway
Publishing three times as much did not produce more impressions, more clicks, or proportionally more indexed pages this month. It produced a better average position on a smaller set of queries. Those are not the same win.
What this data does and does not tell you
It tells us the direct comparison between a 4-post month and the following 14-post month, on this one property, went against the volume argument.
It does not tell us why. We have a plausible mechanism for the position/impressions split, not a proven one. It does not include November, which is five days of partial, still-processing data.
It also cannot separate a real cadence effect from seasonality, algorithm changes, or the specific topics each month happened to cover. Four of September's articles and fourteen of October's are not a controlled experiment — they are two different sets of topics published at two different speeds.
The obvious caveat
Two months of data, one property, no control group. The October indexed-page count is a snapshot almost two weeks old by the time we checked it, which is itself one of this post's findings. The raw numbers are committed to the evidence folder so anyone can re-pull the same date ranges and check our arithmetic.