OpenChat Graph: 20 Months of AdSense Revenue Published in Full. A ¥20K Plateau, and the Month Pageviews Jumped 2.4x While Revenue Stood Still

Indie-developer ad revenue tends to be narrated as either a success story or a daydream. The developer of OpenChat Graph (openchat-review.me), a statistics site for LINE OpenChat, publishes the opposite kind of record on Qiita: 20 months of actual monthly AdSense figures since launch in May 2024. Month one brought ¥1,725, the latest month (December 2025) ¥20,242, and the cumulative total is about ¥287,000. Normal traffic runs 60,000 to 80,000 pageviews a month.

Twenty thousand yen a month is among the smallest figures in this site’s ad-model cases: a fifth of the tool app that reached ¥100K a month, and less than a tenth of the 3.5M-pageview media site earning ¥200-300K. We cover it anyway because the resolution more than makes up for the amount. Per-slot ad rates, tactics that worked and tactics that did not, and a measured data point that hits the ad model where it hurts, a month when pageviews jumped 2.4x and revenue barely moved, are all free to read.

Twenty months: reaching the plateau within half a year

The early months wander in the low thousands, including a June 2024 at ¥1,910. Around the fourth month revenue settled near ¥15,000, entered a band of roughly ¥15,000-20,000, and has not left that band in over a year. Even the best of the 20 months managed ¥20,680 (March 2025), so the first plateau doubled as the ceiling. This is not a growth narrative but a plateau-shaped curve that tops out early and stabilizes. Dividing the latest month’s revenue by pageviews gives about ¥0.25 per view. An RPM (revenue per 1,000 views) around ¥250 is half again below the ¥0.5 per view of the ad-supported media site at 3.5M pageviews, the low ad rate of a niche statistics site showing through directly.

Every slot’s rate is on the table

The latest month’s ¥20,242 breaks down as in-page ads ¥7,793 (38%), rewarded ads ¥6,703 (33%), interstitials ¥4,003 (20%), anchor ads ¥1,564 (8%), and ¥179 other. The decomposition matters because the developer compares RPM per slot and prunes with measurements in hand.

What worked was rewarded advertising (the offerwall type that trades features for watching a video). After its introduction in February 2025, monthly revenue averaged ¥17,003 against ¥15,627 before, an increase of about 10%. Its RPM of ¥500-800 towers over the other slots. What did not work was the side-rail slot, deleted at an RPM of just ¥20-35. Even the staple interstitial slid from an RPM of ¥280 to ¥197, a 30% decline, recording how a “set it and forget it” slot slowly starves.

Ad optimization is usually imagined as adding slots. This case did the reverse: cut the low-rate slots and lean into the one whose rate differs by an order of magnitude. Revenue moved 10% without growing total impressions.

The month pageviews jumped 2.4x, revenue stood still

The most valuable data point in the 20 months is not the peak-revenue month but June 2025. Pageviews surged from the usual 60-80K to 193,661, more than 2.4x. Revenue stayed at ¥18,627, in line with the months on either side.

The developer’s analysis is clean: the surge was low-engagement traffic that bounces before ads are viewed or interacted with, so it contributes almost nothing. A visitor who opens a page and leaves increments the pageview counter but is gone before an ad loads and registers as seen. An ad-supported site’s revenue is set by the volume of engaged pageviews, not pageviews in total. Against the indie developer’s hope that going viral will spike AdSense, counter-evidence this direct is rare.

The same structure cuts the other way too. As the case of app ad revenue sliding downhill over eight years shows, ad models sit permanently exposed to rate swings unrelated to one’s own effort. If even pageviews, the one variable you can move yourself, fail to translate into revenue, the predictability of ad income is doubly low.

How to weigh ¥20,000 a month

A cumulative ¥287,000 over 20 months is not a figure to discuss in hourly wages once development and maintenance time enter the picture. But the P&L of this case has two properties peculiar to the ad model. The revenue, for one, has a stock character: it keeps arriving even if the site is left alone. And if running costs such as servers are small, even ¥20,000 a month can be high-margin (running costs are not disclosed here, so profit cannot be computed).

Sixty to eighty thousand pageviews is a traffic success by indie-web-service standards. Even so, ads alone yield a measured answer of ¥20,000 a month. Placed beside the app that took three years to climb from ¥100 to ¥100K a month, the lesson reads: clearing tens of thousands of yen on the ad model takes either pageview volume or a high-RPM genre. OpenChat Graph sits on the low-RPM side as a niche statistics site, and squeezes rate out of slot optimization within that constraint.

Conditions under which these measurements transfer

• Choosing the ad model is sound only if you accept that revenue does not scale with pageviews and can live with a plateau in the tens of thousands of yen
• Slot optimization moves revenue only for sites that can reconcile order-of-magnitude-higher slots like rewarded ads with their content
• A plan that counts on traffic spikes holds only when inflow is high-engagement search or navigational traffic, since viral pageviews convert to almost nothing

The limits: the RPMs and per-slot results here are specific to the LINE OpenChat statistics genre and its audience, and do not port to other genres as-is. With running costs and development hours undisclosed, the business-level P&L cannot be assessed from outside. What can be verified is the shape of the ad revenue model: how fast the plateau arrives, how the rates distribute, and how nonlinear the relationship with pageviews is.