Our Own Record #4 | 226 articles and 9,103 monthly users, with a six-week lag before the traffic moved

The series closes with the numbers of the one business, out of four, that is still moving: a web media site started on May 15, 2026, publishing explainers on personal finance in Japan.

Unlike the three that stopped, this one has externally confirmable numbers, Google Analytics 4 measurements and article metadata in the repository. Everything below is taken from those, unrounded, with no convenient windows cut out.

The last 30 days, and the 30 before them

Daily, mid-July hovered around 100 users a day. From late July onward it moved into a 400–650 band, peaking around 660. The line didn’t climb a staircase. It stepped up once in late July and then went sideways at the new height.

When, and how many, articles went in

Today 226 are live, with 26 more scheduled on future dates (the build includes them automatically when the date arrives).

The inputs were lumpy: the first 21 articles went out at once on May 27, then 54 more in the five days of June 8–12. July fell to a thin drip of about one a day, August spiked again.

Six weeks sat between input and outcome

Overlay the two tables and the biggest lesson of these three months appears.

The heaviest article input was early June. Traffic took off in late July. About six weeks apart.

In the 30 days from mid-June to mid-July, active users were 1,457, and 113 articles were already live by then. By article count, hardly different from August. Yet one-sixth of the people came. The same inventory produced a different quantity of traffic six weeks later.

We changed almost nothing in that period. July’s only activity was the one-a-day drip. So it is natural to read most of the increase as the time lag for June’s articles to be evaluated by search engines.

Which means: had we judged by the numbers of the week we shipped, we would have killed this business too by the end of June. The two apps in Part 3 died in June and July. In the same weeks, this site also looked like nothing was happening.

Each visitor reads 1.2 pages

Writing only the good numbers would be a lie, so here is the bad one.

About 11,000 page views against 9,103 active users: roughly 1.2 pages per person. People arrive from search, read the one page they came for, and leave. Almost no one browses.

For an article-based reference site this is a common shape, but put differently, 9,103 is closer to “used 9,103 times, once each.” How many became readers, this number cannot say.

About the build

The construction is plain: a hand-rolled static site generator with no external libraries, where adding one article directory adds one page, and the output sits on Cloudflare Pages. No server, no database.

That plainness helped in one specific way: we built it so that writing articles was essentially the only thing there was to do, so effort had nowhere to scatter. On the app side there was always something that wasn’t an article, a feature to add, a language to add, a store listing to fix.

There is also a mechanism where a future publish date simply waits. Write in batches on days you can, publish one a day. July’s thin supply ran on this.

Things whose effect we don’t yet know

We adopted the rule of shipping English/Chinese/Korean translations in early August, not enough time has passed to measure. Having written in Part 3 that 17 languages was “increasing supply without checking demand,” whether this site’s multilingual bet repeats that mistake is something the next 30 days of numbers will show.

On revenue there is nothing to publish yet. Ad and affiliate plumbing exists, but one affiliate network still shows zero partnered brands. There is literally no creative to place. The day you build the pipe and the day money flows through it are different days. That’s all we can say so far.

9,103 is small

Finally, the context. 9,103 monthly users is small for a media site. Among cases on this site are Own The Yard, 600 articles into a “backyard” niche, sold for ~$250K after 3 years, and Investor Junkie, nine years of never outsourcing an article, 80% search traffic, sold for $5.8M. The gap is orders of magnitude.

Against the side-hustle blog that went from ¥42,000 to ¥1M a month in a year, we haven’t even entered the revenue stage. Three months is that kind of age.

Nor is the current number guaranteed to persist. Our archive includes cases where a search update cut profit by 75%, and where 80% of articles lost rankings. What we have is a three-month waypoint, not a trend.

What this article cannot prove

As stated at the start, we haven’t named this site or its URL, so readers cannot verify these numbers from outside. The evidence lives in a GA4 console and a private repository.

The condition this site imposes on every published case is one our own article fails to meet. When we can publish the name, we will replace this with a verifiable version. Until then, these numbers must be read as unverified.

We started four, three broke, and the one that remains produced these numbers. The series pauses here, but the stopped ones and the moving one will both be different numbers in 30 days. When they are, we will publish again.

In this series

• Part 1 — We started four, and three broke
• Part 2 — The two-person SaaS broke on the cap table
• Part 3 — 277 commits and 17 languages, and it still stopped
• Part 4 (this article) — 9,103 monthly users in 3 months: a content site’s real numbers