The Three Numbers Sharing One Word

AIR Media-Tech's study of 300 channels on its own network roster, last updated 1 July 2026, prints three prices for one thousand views of the same kind of video: $9 to $11.50, then $5.73, then $2.33.

All three describe channels in the Gadgets and Technology category. The first is a third-party advertiser-side estimate the study reproduces beside its own measurements. The second is the median CPM those 300 channels recorded in YouTube Studio across 3,595 channel-months from May 2025 to May

1. The third is the median RPM the same 300 channels were paid across those same 3,595

channel-months to May 2026. Only the third is money that reached a creator.

The three prices are not a disagreement and none of them is false. Each is correct about the thing it measures, and each measures a different thing. The word all three share is thousand, and it counts a different event in every one of them.

Four metrics carry that word in the platform's own documentation, and confusing them is the most expensive mistake available to a person choosing what to make. YouTube's ad-revenue analytics help page, retrieved 5 September 2026, defines all four. They are set out here in the platform's own wording rather than in a paraphrase, because the paraphrase is where the conflation starts.

CPM, on that page as it read on 5 September 2026, is "how much money advertisers are spending to show ads on YouTube" — measured before YouTube's revenue share, denominated in 1,000 ad impressions, and counting revenue from ads and from YouTube Premium only.

RPM, on the same page of 5 September 2026, is "how much money you've earned per 1,000 video views" — measured after YouTube's revenue share, denominated in 1,000 views rather than 1,000 impressions, and counting channel memberships, YouTube Premium, Super Chat and Super Stickers alongside ad revenue.

Playback-based CPM, on that same page of 5 September 2026, is the cost an advertiser pays for 1,000 video playbacks where an ad is displayed.

Estimated monetized playbacks is the number of times a video was watched with ads, and the page carries YouTube's own worked example of it as retrieved on 5 September 2026: ten views, eight of them with ads, gives ten views and eight monetized playbacks.

The four, as that help page defined them on 5 September 2026:

Three of the four are rates, and each divides by a different thousand. CPM divides by a thousand ad impressions. Playback-based CPM divides by a thousand playbacks that served an ad. RPM divides by a thousand views, and a view that served no ad at all sits in that denominator exactly as a view that served three.

The fourth is not a rate. It is a count, and it is the count that decides how far apart the other three land.

Of the three rates, the middle one is the one almost nobody quotes. Playback-based CPM, as YouTube's ad-revenue analytics help page defined it on 5 September 2026, divides by playbacks where an ad displayed, which makes it the only one of the three whose denominator moves with how often ads actually served. Not one of the vendor tables read for this volume's research through 5 September 2026 publishes it. The metric that sits between the number in the sales table and the number in the bank account is the one with no public table behind it at all.

Google publishes no conversion between them. The ad-revenue analytics help page, retrieved 5 September 2026, defines all four metrics and says that RPM is lower than CPM, and it prints no ratio and no worked example connecting the one to the other. There is no multiplier on that page. There is no table of typical conversions. A reader who wants to move from an advertiser's price to a creator's line has to do it with numbers from their own channel, because the platform that holds both sides of the transaction has published nothing that does it for them.

What circulates in place of the missing conversion is a band: RPM is 10% to 40% of CPM. That band is an estimate assembled from creator-reported figures rather than a published Google ratio, and it is labelled that way here because it is usually reproduced with a link to the help page that does not contain it. It is a reasonable summary of what creators have said. It is not a measurement, and it is not the platform's.

Three separate things move between the two metrics, and each of them is visible in the definitions on that page of 5 September 2026.

1. The platform's share comes off. YouTube's partner earnings overview, retrieved 5 September 2026, sets the creator's cut at 55% of net revenues from ads displayed or streamed on their public videos. CPM is measured before that subtraction and RPM after it.
2. The denominator changes. A CPM is divided by ad impressions and an RPM is divided by views, and YouTube's own worked example of 5 September 2026 — ten views, eight monetized playbacks — is the platform stating in its own arithmetic that the two counts are not the same count.
3. The numerator changes in the other direction. RPM includes channel memberships, YouTube Premium, Super Chat and Super Stickers, per that same help page retrieved 5 September 2026, and CPM excludes all of them.

So RPM equals CPM times 0.55 is arithmetic that holds only for a channel where every single view serves exactly one ad impression and nothing is earned from anything else, which describes no channel in the 2026 dataset this chapter is about.

The same partner earnings overview, retrieved 5 September 2026, sets two further splits, and they are printed here so that the 55% is not mistaken for the only one: 45% of the revenue allocated to a creator from the Creator Pool, and 70% of net revenues from channel memberships, Super Chat, Super Stickers and Super Thanks.

Notice who publishes which number. Advertiser-side CPM estimates are published by parties selling to advertisers or selling tools to creators, and they are published freely, in tables, with categories down the side. Studio CPM and RPM are visible to one person per channel, and that person has no reason to publish them and no obligation to. The number that is easiest to find is the one furthest from the bank account, and that is a fact about publishing rather than about YouTube.

One standing rule falls out of that and is worth setting down before any table in this book is opened. A CPM table is not a pay table. Every figure in one is an estimate of what a buyer spends, measured before YouTube's revenue share in 2026, and nobody is paid a CPM — not the third-party estimate, and not the Studio figure either, since the Studio figure is also measured before the split. This volume prints no CPM table as creator pay, and where it prints a CPM at all it says which side of the transaction the number describes.

Which brings the chain back to Gadgets and Technology, where all three numbers exist in one document.

AIR Media-Tech is a multi-channel network that sells paid channel audits, and the study it last updated on 1 July 2026 is, as of September 2026, the only public dataset of its kind built from Studio revenue lines rather than from advertiser-side estimates. Its stated method is the YouTube Analytics of 300 channels on the network's own roster, every monetized month from May 2025 to May 2026, which is 3,595 channel-months across thirteen niches and channels running from 10,000 to 50 million subscribers. Every figure it publishes is a median.

For Gadgets and Technology, that study of 300 channels and 3,595 channel-months, last updated 1 July 2026, gives a median Studio CPM of $5.73, a median RPM of $2.33, and a monetized playback rate of 34%. The advertiser-side estimate the same study prints beside those figures on 1 July 2026 is $9 to $11.50 for the same category.

Read in order, the chain is complete. Advertisers were estimated at somewhere around ten dollars a thousand. YouTube Studio recorded $5.73 a thousand ad impressions in that 300-channel sample of 1 July 2026, before the split. The creator's own line came out at $2.33 a thousand views across those 3,595 channel-months to May 2026.

Take 55% of $5.73 and the answer is $3.15, which is this book's own arithmetic on the split published in YouTube's partner earnings overview of 5 September 2026 and the median Studio CPM published in that 300-channel study of 1 July 2026. The measured median RPM in the same 300-channel sample is $2.33. The revenue share explains part of the distance between the advertiser's price and the creator's line in 2026, and it does not explain the rest.

The rest is in the third figure. A monetized playback rate of 34% in that 300-channel, 3,595-channel-month study last updated 1 July 2026 means that in the median Gadgets and Technology month, about two views in every three never served an ad at all. A rate calculated per thousand ad impressions is then set beside a rate calculated per thousand views, and roughly a third of those views are the ones that produced any impression to divide by. That mechanism has a chapter of its own and is named here rather than developed, because the point of this one is narrower: the numbers were never convertible in the first place, and the reader has been shown three of them and told they were one.

The vendor says as much about its own comparison column. That study, last updated 1 July 2026, states that advertiser-side estimates run roughly 1.5 to 2 times higher than Studio CPM. The $9-to-$11.50 estimate it prints for Gadgets and Technology on 1 July 2026 is therefore not offered as a competing measurement of the same quantity. It is printed as the number in circulation, beside the number in the dashboard, by the party that read the dashboard.

One more comparison inside that dataset is worth carrying, because it is the shape of the whole volume in one line. In that study of 300 channels and 3,595 channel-months last updated 1 July 2026, Gadgets and Technology holds the third-highest median Studio CPM of the thirteen niches measured, and its median RPM of $2.33 sits within a few cents of the all-niche median of about $2.30 in the same study. A category near the top of one table lands in the middle of the other.

The column that follows from that $2.33 median in the study of 1 July 2026 is the one a person planning a channel actually feels. In that same 300-channel study of 1 July 2026, a Gadgets and Technology channel at the median needed roughly 430,000 views to earn $1,000, against roughly 98,000 views for Education and Science and roughly 3,014,000 for Kids and Teens in the same table of 1 July 2026. Those are medians on 300 channels and 3,595 channel-months to May 2026, not a scale anybody was promised, and the distance between the first of them and the last is why this volume treats subject and format as one decision rather than two.

Three things travel with every number from that dataset, in this book, every time one is used.

The sample is one network's own roster of 300 channels, which is a commercial relationship rather than a random draw, and this volume's research reads that roster as heavily non-US in origin, which is an assessment made here and not a disclosure made by the vendor in its study of 1 July 2026. Four of its thirteen niches rest on fewer than ten channels each, and the study labels those four directional rather than firm. And the same vendor's two 2026 studies do not agree with each other.

That last one is not a quibble, and it is stated here in full because it will be stated again every time this dataset is opened. The vendor's Shorts study, published 23 June 2026 on 274 channels and 3,044 channel-months, puts the long-form monetized playback rate for Kids and Teens at 12.1%. The RPM study last updated 1 July 2026, on 300 channels and 3,595 channel-months, puts the same rate for the same category at 32%. One vendor, one year, one metric, one category, published eight days apart in 2026, and a disagreement of a factor of 2.6.

The same two studies disagree on Education as well. The median long-form RPM for Education and Science is $10.22 in that 300-channel study of 1 July 2026, and $14.97 for medium channels and $18.23 for small ones in the 274-channel study of 23 June 2026. Different samples explain some of that and they do not explain its size.

So the Gadgets and Technology chain is printed in this book as the clearest illustration in the public record of how the three numbers separate, and not as a rate anybody should carry into a decision. The mechanism is legible. The specific figures are not reproducible even inside the one organisation that measured them.

That page, updated 24 July 2026 and retrieved 5 September 2026, carries no sample size, no method and no date range for the table those figures sit in. Its own hedge is the more useful half of it, and it is printed above rather than summarised because it names the evidence base out loud: what creators discuss, and what industry tracking platforms aggregate. That is the citation chain, given by the publisher, one line under the numbers.

Then the unit, which takes one division. The $15,000 per million views published on that page on 24 July 2026 is $15 per thousand views. Stated that way it is a claim about RPM, and the page never says whether it means the money a creator receives or the money an advertiser spends — the distinction this chapter has spent four definitions on.

Set beside it, the only 2026 dataset built from Studio revenue lines — 300 channels, 3,595 channel-months, last updated 1 July 2026 — puts the median RPM for Business and Finance at $2.01, eleventh of thirteen niches, on three channels.

Three channels is not a refutation. A page with no sample size at all is not a measurement. The word for the distance between them is unsupported, and this book prints that word rather than choosing whichever of the two numbers it prefers.

The rest of that page is built the same way. It puts Gaming at $2,000 to $8,000 per million views and Entertainment and Vlogs at $1,500 to $5,000 on 24 July 2026, disclosing exactly as much about those rows as about the finance row, which is nothing. Divide the finance row by the gaming row at both ends and the page's own table implies a ratio of five to seven and a half, which is this book's arithmetic on two ranges published together on 24 July 2026 with no sample size and no date range between them. Even the loudest table in circulation does not say ten.

The machinery behind the wider genre is worth naming once. Public estimators such as Social Blade, checked 5 September 2026, read subscriber and view time series from public interfaces and multiply them by an assumed CPM band, and they are the engine underneath most articles asking what a named channel earns. What an estimator of that kind cannot see is the list this chapter has just walked through: the monetized playback rate, the ad load, the audience geography and the revenue share, which are the four inputs that turn a category's advertiser price into a creator's line in 2026. No figure from an estimator of that kind appears anywhere in this book.

Even the third number is a middle. In that study of 300 channels and 3,595 channel-months last updated 1 July 2026, the Gadgets and Technology median RPM of $2.33 sits inside a band running from $0.70 at the 25th percentile to $3.71 at the 75th, a spread of more than five to one among channels inside one category. Half the Gadgets channels in that sample fell outside that band in one direction or the other. A median is the value in the middle once every channel is lined up smallest to largest, and it describes the line rather than any channel standing in it.

One thing that 300-channel study does not contain is a failure. It reports medians and percentile boundaries across 3,595 channel-months to May 2026, and it names no channel — not a winner, not a loser, nothing a reader can look up and nothing anybody can check against a Studio export. No failure data was published beside it. That is not a criticism of the vendor, which owed nobody a roster; it is a description of what the best public measurement in this area is, which is a set of middles with no names attached.

What would close that gap is a cohort of creators publishing dated Studio revenue exports, with the channels that stopped uploading still in the sample, because a dataset assembled only from channels that lasted has already answered the question it was built to ask. As of September 2026, no such cohort had published.

The test that survives all of this is two questions long, and it runs on any per-thousand figure a reader meets anywhere. Which thousand is on the bottom of the fraction — ad impressions, playbacks that served an ad, or views? And is the figure quoted before the platform's cut or after it? A seller who cannot answer both about their own table is not describing anybody's pay, and the answer takes a sentence, so ask for it in writing.

So take the three numbers off your own dashboard before you accept anybody's. Open YouTube Studio, then Analytics, then the Revenue tab, and set the date range to the last 12 months. Write down three figures and label each one with the thousand it counts: your RPM, per thousand views; your playback-based CPM, per thousand playbacks that served an ad; and your estimated monetized playbacks, which is not a rate but the count the other two disagree about. Put today's date beside all three, because every one of them moves, and read no further table in this book until they are written down.