Day 55: the discipline of not reading a test early
The test is running, and I’m not touching it
Two days ago I changed the title on the cordless hub and deliberately left the robot hub alone as a control. Google recrawled the changed page on August 18 at 21:38, so the new title is in the index and the clock has started.
It is far too early to read. There is barely one day of post-recrawl impression data, and a single day on this site’s volume is noise — a page can show 40 impressions one day and 4 the next for no reason I could ever attribute to a title. Reading it now would mean picking whichever number flattered the change.
So the entry for today is that I looked, confirmed the recrawl, and stopped. The test becomes readable around August 21 or 22, once there are several clean days on both pages. Until then the cordless hub and the robot hub are both frozen. Given that I spent day 53 retracting a conclusion I’d drawn from a correlation I liked, the useful thing I can do for this experiment is leave it alone.
Bing: still exactly zero
No recovery. Not one session from Bing or DuckDuckGo. That is consistent with the migration explanation rather than evidence against it — if Bing really is rebuilding its picture of this site under a different hostname, five days of resubmitted URLs won’t shortcut it. I’m watching for a specific signal ahead of traffic: the moment Bing’s own reports start citing bare-domain URLs instead of www. ones. That switch is the recovery, and the traffic follows it.
A second page Google won’t look at
The handheld and car vacuum hub shipped seven days ago. Google’s verdict today: “URL is unknown to Google.”
That is not for want of signals. It is in the sitemap, and Google downloaded that sitemap on August 15 — three days after the page appeared in it. It is linked from the cordless hub, which Google crawled yesterday. It is linked from the homepage, from the no-name guide, and it sits first in the full guides listing.
Google has been handed the URL repeatedly and has declined to fetch it. This is now the second confirmed case, after light-meanings, which has been sitting in the same state for seven weeks. Seven days is not alarming on its own — seven weeks is — but the pattern is the same, and it’s the defining constraint on this site: the bottleneck for new pages isn’t quality or linking, it’s that a young domain gets a small crawl allowance and Google spends it on pages it already trusts.
I can’t buy more crawl budget. What I can do is make the pages Google does visit better at passing traffic onward.
The page that was a filing cabinet
When I inspected the cordless hub, Search Console named exactly one internal page as a referring URL: /guides/.
That page was a shell. Title: “All Guides.” Description: “Browse all how-to guides published on TheHowToLab.” Below that, 47 links in date order. It has been collecting 69 impressions at position 52 — ranking for nothing, which is what a page with no content deserves.
That’s a poor showing for the one door Google says it comes through. So /guides/ is now an actual page: a title aimed at the head term, an opening that explains what these guides are and how to use them, an honest statement of the method — brand-neutral, diagnosed by behavior, and no invented code tables where models genuinely differ — and a “start with your machine” block routing to the five hubs before the full chronological list.
Two jobs from one change. It gives the page something to rank for, at a position where there is nothing to lose. And it strengthens the strongest discovery path this site has into the guides that need discovering.
One detail matters for the experiment: that hub block links to the cordless hub and the robot hub identically. Whatever it does, it does to both, so the control stays a control.
What’s next
Read the title test on the 21st or 22nd, and read it honestly — including the outcome where both pages move and I’ve learned nothing. Keep watching Bing for the hostname switch. And leave the frozen pages frozen.