AdSense pays ¥386 per 1,000 pageviews. What 225 blogs sold on Rakko M&A say about the distance between PV and revenue

This site has a rule: deals that sold for ¥1,000,000 or less do not get an article, because they are too small, and the rule still looks right to us. What we also knew is what falls outside it. The sold-listings archive of Rakko M&A, a Japanese marketplace for buying and selling websites, is full of exactly those deals. We once wrote up 29 of them and retired every one under that rule. A listing that cannot carry an article on its own can still form a distribution when you have 240 of them. So this time we counted someone else’s ledger ourselves.

The set is every Rakko M&A listing that contains the word “adsense,” is classified as web media (blogs and affiliate sites), and is marked as sold. That came to 240 listings across 12 pages as of September 26, 2026. Each listing carries the sale price plus three figures the seller declares: monthly revenue, monthly profit and monthly pageviews. We kept the 225 where both pageviews and revenue were at least 1.

There was only one number we wanted at the start. How much revenue does 1,000 pageviews produce? Everyone who writes a blog carries some version of that conversion in their head, and everyone’s version is a little different. The median across 225 listings is ¥386. What stopped us was not the median. It was that the sites with the most pageviews earned the least per pageview.

1,000 pageviews is ¥386, and half the listings sit between ¥200 and ¥800

Divide monthly revenue by monthly pageviews and multiply by 1,000. Following ad-industry convention, we call that RPM. Here is the distribution across 225 listings.

The median is ¥386 and the quartiles are ¥235 and ¥778. Fifty-six percent of listings fall in the ¥200 to ¥800 band. The bottom tenth is under ¥108 and the top tenth is above ¥1,436, a 13-fold spread from bottom to top.

A 13-fold spread invites the conclusion that pageviews cannot predict revenue. That conclusion is wrong. Regressing revenue on pageviews in log-log terms gives a slope of 0.95 and a correlation of 0.86. Ten times the pageviews brings roughly nine times the revenue, and that relationship holds across almost the entire range of the 225 listings. Pageviews are the strongest single variable behind revenue, and the 13-fold spread is the scatter around that line. We will get to what causes the scatter.

Our reading, stated up front: ¥386 is lower than the rate most bloggers carry in their heads. Among the cases on this site, the indie developer who took 14 years to reach ¥1,000,000 a month got there on 4.22 million pageviews, which is ¥237 per 1,000. The stay-at-home blogger’s second year produced ¥203,000 on 76,658 pageviews, an RPM of ¥2,648. The first month of Trivipedia’s AI-driven relaunch came to 10,971 pageviews at an RPM of ¥164.81. Three cases, all “ad revenue,” a full order of magnitude apart. The 225 Rakko M&A listings sit between them, toward the low end.

Past 300K pageviews, a pageview gets cheaper

Here is the median RPM by pageview band.

Up to 300K pageviews, the median barely moves. It stays between ¥390 and ¥430 whether a site has 10K or 200K. Then it falls to ¥258 above 300K and to ¥179 for the six sites above 1M. Six is too few to treat ¥179 as settled, but the 16 listings at ¥258 point the same way. The more pageviews a site piles up, the cheaper each one sells for.

The reason is subject matter. The six sites above 1M pageviews cover manga spoilers, TV dramas and celebrities, fortune-telling, “overseas reactions” and serialized personal stories driven by social media. They pull huge traffic from search and social in a short time, and advertisers pay little per click for it. At the other end, the under-10K band’s RPM of ¥417 is propped up by affiliate sites for home security, beauty, cars and finance, where each pageview is worth more and there are far fewer of them. The next section checks this split against what the sellers themselves declared.

AdSense-only listings earn ¥476, AdSense plus affiliate ¥557, and the 142 under the old label ¥348

A search for “adsense” does not guarantee that AdSense is the only revenue source. Each listing’s detail page has a “revenue model” field, and splitting on it gives this.

As the table shows, the field changed shape partway through. Every listing sold through 2023 carries the single word “advertising.” The breakdown into CPC/CPM and affiliate appears only on listings from 2024 onward, and 94% of 2025 listings use the new labels. The first row is therefore 142 sites where we cannot tell whether anything besides AdSense was involved, and the remaining rows are the 83 where we can.

Within what can be told, the 41 sites running on click-based ads alone had a median RPM of ¥476, with quartiles at ¥254 and ¥786. That is above the overall ¥386. The 36 sites combining affiliate links came in higher at ¥557, but their median pageviews are only 14K, a size where a handful of affiliate conversions can swing the rate. The four affiliate-only sites had a median of ¥4,903, which is a different business altogether. The top tenth of RPM was where the mixed and affiliate-only listings gathered.

Splitting by category sharpens the picture. The 148 sites whose category includes entertainment, celebrities and trends, games, manga and anime, overseas reactions or sports had a median RPM of ¥370. The other 77 had ¥464. But the first group’s median pageviews are 46K against 13K for the second, a 3.6-fold difference in size. Of the 22 sites at 300K pageviews or more, 86% were in the entertainment group. Of the 57 under 10K, 42% were. The subjects that gather pageviews and the subjects that make a pageview valuable barely overlap.

One more set of numbers comes from the detail pages. The median listing had 219 articles, with quartiles at 106 and 452. Dividing pageviews by article count gives 128 pageviews per article per month, which works out to ¥74 of revenue per article per month and ¥1,500 of sale price per article. The median time from launch to sale was 15 months, with quartiles at 9 and 36. Half the AdSense blogs that sold were let go a little over a year after they started.

Sale prices ran 19 months of revenue, except that sites under ¥10,000 a month sold for 60 months

Dividing the sale price by monthly revenue gives the price as a number of months of revenue. The median across 225 listings was 19.4 months. The quartiles were 10.2 and 53.9, another wide spread, and this one is explained by the size of revenue.

The 94 listings under ¥10,000 a month sold for 60 months of revenue, five years’ worth. Above ¥100,000 a month the multiple drops to 9 or 10 months. Within a single marketplace, the multiple varies six-fold.

Reading this as “small sites are overpriced” misses the point. The median price for a site under ¥10,000 a month was ¥150,000, and the buyer was paying for something other than five years of ad income, namely the domain’s age, the article count and an AdSense account that has already passed review. What sets the price is not a multiple of revenue but what the buyer is buying the site as. Under ¥10,000 a month, it is a foundation to start from. Above ¥100,000, it is income, and the moment it becomes income the multiple converges on the 7.8-month median of the 20 Japanese exits we counted among this site’s own cases.

One more ratio: sale price divided by monthly pageviews. The median is ¥9.2 per monthly pageview. A site with 100K pageviews would price out at ¥920,000, which is close to the ¥890,000 median in the 100K to 300K band above.

Four in ten sites sold at less than half their best month

Alongside the latest monthly revenue, sellers declare the average and the peak over the period. The median ratio of latest to peak is 0.66, and 39% of listings sold with revenue at less than half their best month. Selling on the way up is the other six in ten, and even there, only 49% of listings had a latest month above their average, an even split.

Profit is nearly identical to revenue. The median ratio of monthly profit to monthly revenue is 0.956, and in 26% of listings the two figures are identical. That is what the books of a business with nothing but server and domain costs look like. For a buyer, the safer reading is not “95% margin” but “costs were not written down.” Outsourced writing and the seller’s own hours sit outside these numbers.

The median time from listing to sale was 18 days, with quartiles at 7 and 39. A quarter of listings closed within a week.

From 2021 to 2025, a pageview doubled in price and the multiple halved

Sorted by year of sale, the rate and the multiple move in opposite directions.

We left 2020 out because it has only three listings. Median RPM rose from ¥230 in 2021 to ¥494 in 2025, more than doubling. Over the same period the multiple fell from 21 months to 13, and the median sale price dropped from ¥420,000 to ¥190,000.

We expected to explain this with a change in subject matter. Counting proved otherwise. The share of entertainment-type sites wanders between 55% and 80% by year with no direction. What is certain is that listings got smaller. The median pageviews of sold listings fell from 63K in 2021 to 16K in 2025. But as the earlier section showed, RPM is nearly flat below 300K, so shrinking size does not explain a rising rate either. From here on we are guessing: either the price of a click itself went up, or sellers changed how they report revenue, and this ledger alone cannot separate the two. The falling multiple is easier to read. Buyers stopped pricing ad income as something that lasts, and our view is that this is what repeatedly watching search-algorithm updates knock the pageviews out of a purchased site does to them.

How we counted, and what these numbers cannot tell you

The set is Rakko M&A’s sold listings containing the search term “adsense,” nothing more. A blog whose seller never wrote “AdSense” in the listing is not here, and a site that runs mainly on affiliate income with AdSense as a sideline is here as long as the word appears. The top tenth of RPM is where those sites gathered, including a mobile-carrier affiliate site earning more than ¥40,000 per 1,000 pageviews. The median is robust to that contamination and the mean is not, which is why there is not a single average in this article.

Revenue, profit and pageviews are all seller-declared figures that Rakko M&A does not audit. Listings with a Google Analytics or Search Console connection badge pull some figures from live data, but not all listings have one. As a source, this ranks a step below the self-published revenue reports that this site’s case articles rely on. We are using the figures as a distribution, not as facts about any individual listing.

The population is sites that sold. Sites that were listed and never sold, and sites that were never listed, are absent. A site that sold is one whose numbers were tidy enough for a buyer to price, so these 225 lean toward the better-looking end of all AdSense blogs. It is the same structure as this site’s own cases, which skew toward numbers people wanted to publish. The premise that a published number is a number someone had a reason to publish holds here too.

No listing titles or listing text appear in this article. Using figures for a tally and reproducing another company’s listing copy are two different things, and we do not do the second.

Writers count pageviews, and subject matter sets their price

What stayed with us after counting was not the median but how flat the rate was. From 10K to 300K pageviews, 1,000 pageviews were worth around ¥400 and nothing moved that. Multiplying pageviews by ten did not make a pageview worth more, and past 300K it made each one worth less. A blog’s pageviews turned out not to be the kind of asset whose unit value rises as you accumulate it.

Meanwhile, the 13-fold spread came from subject matter. A pageview on home security, beauty or finance and a pageview on a celebrity roundup sit in the same AdSense dashboard at different prices. What the writer can move is pageviews. What moves the rate is the choice of what to write about, and that choice was made the day the first article went up. ¥386 is just what that choice looks like when someone counts it afterward.