Seven in ten views on a company YouTube channel were paid for

Why nobody has this number

Search for how a company channel performs and you will find, in order: how much YouTube pays per view, how many subscribers you need to monetise, and a survey in which some percentage of marketers say video gives them good ROI. I pulled the live results page for that question on the day I wrote this. The first ten results are all about creator earnings. Not one of them counts a company's videos.

That is not an accident. Surveys are cheap and flattering: the people who answer them are the people who bought the video, and nobody who has just spent forty thousand on a brand film tells a questionnaire that it got six hundred views. Counting is expensive and unflattering. It is also the only thing that answers the question a founder actually asks before signing: if we do this, what happens?

So I counted. Not opinions about company video — company videos.

How the sample was built

I did not search YouTube for company channels. That would have found the channels YouTube already promotes, which is the same as asking the winners how the game is going. Instead I went the other way round: start from companies, and follow the link the company itself publishes.

Two samples, both public and both reproducible. One — 13,045 companies in the United States, the United Kingdom, Canada and Australia that are listed as businesses in Wikidata with an official website on file. Two — 5,239 small businesses in New York City that carry a website in OpenStreetMap: shops, salons, law offices, contractors, bakeries. Big enough to be recorded, small enough to be normal.

Every one of those 18,284 homepages was requested once on 31 August 2026. 13,752 answered (75.2 per cent — the rest were dead, parked, blocked or timed out). Each answering page was read for links to the seven platforms a company might publish on, and for a YouTube channel address in any of its forms: the modern handle, the old channel id, the legacy vanity name. Every handle was then resolved against the live channel page, and every surviving channel queried through the YouTube Data API for its statistics and its fifty most recent uploads.

What came back: 3,469 distinct company channels, of which 3,455 still exist, of which 3,375 have at least one video. Those channels' most recent uploads add up to 134,164 videos and 10,730,546,238 views. That is the corpus. Where a comparison could be confounded by age — an eight-year-old video has had eight years to collect views — I re-ran it on the 39,433 videos published between 90 and 455 days before the measurement, so every video in the comparison had roughly the same time to work.

First: most companies are not on YouTube at all

Of the 10,111 reachable company sites in the first sample, 38.5 per cent link YouTube somewhere on the homepage — but only 30.2 per cent link an actual channel; the rest are embedded videos and stray watch links. Facebook, the platform every strategy deck has declared dead for eight years, sits at 68.9 per cent. LinkedIn — the one every B2B article insists is the serious channel — sits at 34.5 per cent, below YouTube.

Two things fall out of that. The first is that a company's homepage is a fair witness: it shows what the company decided to be findable on, not what an agency recommended. The second is the New York column. Among small businesses the YouTube share halves to 21.3 per cent, and LinkedIn collapses to 13.0 per cent — while Instagram barely moves. Small businesses have not been convinced by the B2B consensus. They went where the customers already were.

One more number from the crawl, and it is the cheapest lesson in the whole study: of the 2,424 handle and vanity YouTube links published on those homepages, 75 point at nothing. Three per cent of companies are linking a channel that no longer exists — a dead badge in the footer that has been sitting there long enough for nobody to notice.

The graveyard: a quarter of company channels have not posted in a year

Of the 3,375 channels with videos, 46.0 per cent have published nothing in the last 90 days. 27.0 per cent have published nothing in a year. 18.6 per cent have published nothing in two years, and 14.3 per cent nothing in three. The median channel last posted 68 days ago.

The interesting part is not that channels stop. It is when. The channels that have gone quiet were not abandoned in a first flush of enthusiasm — the median dormant channel ran for seven years before its last upload. These are not failed experiments. They are things that worked well enough to keep doing for the better part of a decade, and then somebody left the company, or a budget line moved, and the lights stayed on with nobody inside.

Some of them are recognisable. M&C Saatchi London — an advertising agency — has 15,300 subscribers and has not uploaded since 13 November 2016. Brookstone published 2,608 videos and stopped on 30 October 2020. BioWare, 62,700 subscribers, last upload 25 November 2020. None of these companies vanished. Their channels did.

One more thing the dates say. Sort every channel by the year of its last upload and 67.4 per cent land in 2026, 12.5 per cent in 2025, 4.5 per cent in 2024 — and then a tail that runs back to 2007. The median channel in this study was created 12.8 years ago, and the single most common year of creation is 2011. This is not a young medium full of new entrants. It is an old medium full of buildings from the first wave.

What a company video actually gets

Here is the distribution nobody publishes, across all 134,164 videos.

Restrict it to videos published between 90 and 455 days ago, so that every film in the comparison has had roughly a year to find its audience and none of them is riding on a decade of accumulation: 39,433 videos, median 357 views, 65.0 per cent under a thousand. The picture does not improve when you control for age. It gets slightly worse.

And the ceiling matters more than the median. 14.4 per cent of these channels have never had a recent video pass a thousand views — not the median video, the best one. 44.9 per cent have never had one pass ten thousand. Nearly half of all company channels have a personal best that would be an unremarkable Tuesday for a teenager with a phone.

The median company channel has earned nine subscribers for every video it has ever published.

That figure — 9.1 subscribers per video, median across the corpus — is the one I would put on the wall. It is what "we should be posting more" converts into after a decade.

The finding I did not expect: seven in ten views were bought

The pooled numbers have a strange shape. The median video gets 427 views, but the corpus holds 10.7 billion views, and the top one per cent of videos carry 81.9 per cent of them. My first assumption was the usual power law: a few pieces of content break out, everything else does not. Then I checked what those breakout videos look like from the inside.

A view is a view, but a like per view is a fingerprint. Organic reach and paid reach produce completely different ratios, because a paid impression is served to somebody who did not choose it. So I split all 134,164 videos into view bands and measured the like rate in each.

A company video with a thousand viewers collects about ten likes per thousand views. The same companies' videos above a million views collect 0.03 — one like per thirty-three thousand views. There is no version of human behaviour that does that. People do not become three hundred times less impressed at the exact moment a video crosses six figures. What changes is that the video stopped being watched by people who chose it.

So I took the conservative line and marked every video over 100,000 views with a like rate below 1 per 1,000 — a tenth of what the same corpus produces organically. That flags 3,079 videos, 57.2 per cent of everything above 100,000 views, and 2.3 per cent of the corpus. Those 3,079 videos hold 70.2 per cent of all 10.7 billion views in the study.

Take them out and the whole picture settles into something much more honest: median 400 views, 64.9 per cent of videos under a thousand, and the largest genuinely-watched video in a corpus of 3,375 companies at 125.7 million views. Twenty per cent of channels have at least one bought-looking placement in their last fifty uploads. The median channel's entire earned harvest across those fifty videos is 48,389 views — about what a mid-sized local radio spot buys, spread over years of production.

One detail makes this less damning and more useful than it sounds. A skippable in-stream view on YouTube is billed, in Google's own words, "when a viewer watches 30 seconds of your video, or the entire duration of the video if it's shorter than 30 seconds, or interacts with your video, whichever comes first". So a bought view is not a fake view — somebody genuinely sat through half a minute. They just did not choose to. That is the actual product a company channel with reach is buying, and it is worth knowing which of your videos was bought before you conclude that one of them was good.

Inside a single channel, one video is the channel

Concentration is not only a property of the corpus. It happens inside almost every individual channel. Across channels with at least ten recent videos, the single best video holds a median 25.2 per cent of all the views the channel collected — and in 19.8 per cent of channels one video holds more than half. The best video beats the channel's own median by a factor of 35.8; in the age-controlled window, 13.8; at the ninetieth percentile of channels, 963.

Some of the spreads are absurd. Rent-A-Center's channel has a median of 8,630 views and one video at 552.6 million. Wix's channel medians at 2,696 and carries one at 302.4 million. Goldsmiths: median 687, top video 44.5 million. Those are the paid placements from the previous section — which is exactly the point. The gap between your best video and your typical video is mostly a description of your media spend, not your creative.

Two consequences, and they point in opposite directions. If you are planning a channel, plan for the median, because that is what you will get on any given Tuesday. If you are planning a campaign, the concentration is good news: you do not need two hundred good films, you need one and a budget behind it. Almost every company in this study does the opposite — two hundred films and no budget behind any of them.

Subscribers are not an audience

The number companies report internally is the subscriber count, because it only goes up. Here is what it buys. Channels under 100 subscribers get a median 108 views per video — more views than subscribers, five times over. Channels with 100 to 1,000 subscribers get 268. From 1,000 to 10,000: 521. From 10,000 to 100,000: 1,480. Above 100,000 subscribers: 5,927.

Read those as a ratio and the ladder inverts. A channel with fewer than a hundred subscribers reaches 5.1 people per subscriber. A channel with over a hundred thousand reaches 0.012 — roughly one viewer for every eighty-three subscribers. Growing the subscriber count by three orders of magnitude multiplied the typical video's reach by about fifty-five, and divided the share of the audience that turns up by four hundred.

Only 19.7 per cent of the channels still publishing reach more people with a typical video than they have subscribers. For four out of five companies, the subscriber list is a number in a report, not a distribution mechanism. If a marketing plan says "we will build an audience on YouTube", ask what the plan does on the day the audience does not watch — because that is the median day.

Posting more does not make the videos work

The universal advice is consistency. Post weekly, feed the algorithm, compound. It is repeated so often that I built the check expecting to confirm it. So: within the age-controlled window, group every still-publishing channel by how often it uploads, and split by channel size so that big channels do not carry the result.

Under a thousand subscribers, going from one video every two months to two–four a month takes the median video from 166 views to 97. Between one and ten thousand subscribers, from 542 to 314. Only above ten thousand subscribers does frequency buy anything, and the gain flattens after one or two a month.

This is not an argument for posting nothing. Cadence does something real: median subscriber counts rise from 102 for the rarest posters to 4,130 for the weekly ones. Publishing builds the library and the list. What it does not do is make the individual film land. The treadmill produces more videos, each of which is watched by fewer people — and the total is roughly flat while the production cost is not.

There is a second reading, and I think it is the true one. The channels posting four times a month are mostly posting to fill a slot. When the calendar is the client, the work gets made to fit the calendar. The cheapest thing to make weekly is the thing nobody needed.

Shorts: a real effect between companies, almost nothing inside one

Vertical video under a minute is now 44.3 per cent of everything companies publish. Pooled, it looks like a clear win: in the age-controlled window a short gets a median 684 views, one to three minutes gets 288, three to ten gets 257, ten to thirty gets 189, and anything over half an hour — the webinar, the recorded panel, the conference keynote — gets 145.

Then I controlled it properly. Take only the 857 channels that published at least three shorts and at least three longer videos in the same window, and compare each channel against itself. Median short: 459. Median long: 305. Shorts win in 52.4 per cent of those channels — a coin flip — and the median ratio is 1.06×.

So most of the pooled gap is not the format. It is which companies chose it: the channels that went hard into shorts are different companies with different reach. Inside a single company, switching to short vertical video is worth about six per cent, not one hundred and thirty. If a strategy deck promises you a format-driven turnaround, that is the number it is promising against.

The half-hour recordings deserve their own line. At a median of 145 views, an hour-long webinar upload is not content — it is an archive. Archives are worth keeping. They are not worth a launch.

Channels do not get better with practice

Every channel with at least twenty recent videos was split in two — the older half of its run against the newer half — and the medians compared. The median ratio is 0.78×. Only 36.3 per cent of channels improved. The typical company channel is quietly getting less effective as it goes.

Part of that is arithmetic: older videos have had longer to accumulate. Part of it is not. It matches everything else here — the audience does not compound, the subscriber list does not show up, and the twentieth video of a series has less to say than the first. A channel is not a skill that improves by repetition. It is a product that either finds a use or does not.

The geography barely moves anything. British company channels run a median 466 views a video, Australian 350, American 346, Canadian 318, with dormancy between 23 and 28 per cent everywhere. There is no market in the English-speaking world where this is going well.

The corner shop beats the corporation

The two samples split cleanly. Notable companies: 2,856 channels, median 398 views a video, 25.2 per cent dormant for over a year, median 881 subscribers. New York small businesses: 519 channels, median 676 views a video — seventy per cent more — but 36.8 per cent dormant for over a year and a median of 595 subscribers.

The small business posts less, quits more often, and gets more views when it does post. That is not a mystery. A bakery filming its own kitchen is making something only it can make, for people who can walk to it. A mid-sized company filming its own conference is making something that exists in ten thousand identical copies.

Now the robustness check, because a study like this lives or dies on whether a handful of giants is carrying it. Throw out every channel above 100,000 subscribers: median falls from 427 to 344, videos under a thousand views rise to 67.6 per cent. Throw out everything above 10,000 subscribers — every media brand, every accidental publisher: median 257 views, 73.0 per cent of videos under a thousand, 32.0 per cent of channels dormant for over a year. Every finding in this piece is stronger among ordinary companies than in the corpus as a whole.

So how often does this actually work?

Of the 3,375 channels, 1,669 — 49.5 per cent — are still publishing, defined as at least five videos in the last twelve months. Everything below is measured on those, because a channel that has stopped cannot succeed or fail; it has left.

Read the bottom line honestly. Four per cent of company channels reach a median of ten thousand views. If you are pitching a channel internally, that is the base rate you are betting against — and the person you are pitching to has already seen a company channel fail, which is why they are asking.

Then I looked at what the top ten per cent do differently. This is the part I expected to be a list of tactics. It is mostly a list of restraints.

The winners upload less often than the channels they outperform by a factor of 143. They are not younger, hungrier upstarts — they are two years older on average. The one thing they genuinely do more of is short vertical video, and we already know from section 10 that inside a channel that is worth about six per cent. And note the last row: even in the top decile, one video carries thirty per cent of the views. Nobody escapes the concentration. The winners are simply the companies whose one video was worth carrying.

What I would do with this, if it were my money

None of this says video does not work. It says the channel is the wrong unit. What works is a film with a job: the one that answers the question every prospect asks before they buy, sitting on the page where they ask it; the one the sales team sends instead of a deck; the one with money behind it, aimed at people who have never heard of you. Those are three different films with three different budgets, and not one of them needs a weekly upload slot.

For the record on where I am standing: my portfolio is 16 billion views across more than 200 YouTube channels, and roughly 1,500 AI-assisted films with about 500 million views. Almost everything I have learned about this medium came from channels that were built to be watched, not built to be posted on. This study is what the other side of that looks like, measured.