What content works best for brands on YouTube?

Content about the customer's problem rather than the company's product. The highest-performing brand channels teach the thing their product helps with and let the product appear only where it is genuinely the tool for the job.

How many formats should a brand channel have?

Three to five at most, and usually one or two carry the majority of growth. Format churn is the most common reason a channel accumulates videos without accumulating an audience.

Why do big companies have small YouTube channels?

Because they publish without programming. Videos get made for internal reasons — a launch, an event, a campaign — and uploaded to a channel nobody runs. Company size has almost no relationship to channel size.

How long should brand videos be in 2026?

Longer than most brands assume. Average view duration across the platform has risen sharply as viewing moved to television screens, and loosely cut long-form now performs where it would have failed five years ago. Length matters far less than whether people finish.

Should brand videos mention the product?

Sparingly, and only where it is the natural tool for the job being demonstrated. Channels that lead with the product consistently underperform channels that lead with the problem, even when both are made by the same company.

We Analyzed 50 Brand YouTube Channels

We read 50 company YouTube channels properly — not the subscriber counts, but what they publish, how often, which videos carried the channel, and which ones sank without trace.

The striking thing was how little variation there was. The channels that work do a small number of things, and the ones that do not fail in the same few ways.

What content works for brands on YouTube

The short version, before the detail: content about the customer’s problem, built into one or two repeatable formats, opening on the action rather than on context. Everything below is the evidence for that.

What we looked at

Fifty channels across B2B software, consumer brands, retail, finance and outdoor goods. Ten of them we took apart video by video — among them Ahrefs, Stripe, Shopify, Dropbox, Red Bull, Patagonia, Liquid Death, Duolingo, Lululemon and Alo Yoga — pulling the full upload history, view distribution and format structure.

Figures below were accurate in September 2026 and will have moved since. Subscriber counts always do.

Pattern 1: the product is almost never in the video

The clearest finding in the study, and the one most brands would reject in a planning meeting.

Ahrefs sells SEO software. Its channel has around 677,000 subscribers and has published roughly 420 videos over ten years. None of them are about the software. Of the fifteen most-watched — which between them carry close to half of everything the channel has ever earned — not one mentions the product in the title.

The company does around $150 million a year, self-serve, without an outbound sales team.

This is not modesty. It is arithmetic. The audience for “how to do the thing” is many times larger than the audience for “how our tool does the thing,” and the first audience contains the second.

Pattern 2: one or two formats carry everything

Across the channels that grew, growth was concentrated in a small number of repeatable structures rather than spread across a varied catalogue.

This matches what operators say directly. Paddy Galloway’s account of channels he has grown from scratch puts it at 70% or more of growth coming from one or two ideas built into formats.

A format is a blueprint you refill, not a topic. “Teardown of a real system, with the numbers” is a format. “Thought leadership” is not. The brands that plateaued had published just as much — they had simply published twenty different things once each, which gives you twenty data points and no baseline.

The corollary is uncomfortable for marketing teams: when something works, do it again immediately. The fear of saturating an audience is misplaced. People underestimate appetite for more of the thing they already liked.

Pattern 3: most brand video is an invisible library

Stripe processes money for most of the Dow Jones and moved over a trillion dollars last year. A third of the videos on its YouTube channel have fewer than 5,000 views.

That is not a failure of production. Stripe’s video is well made. It is a failure of programming: videos commissioned for internal reasons — a conference, a launch, a campaign — uploaded to a channel nobody is running as a channel.

The scale of the gap is easier to see in aggregate. Add up the YouTube subscribers of every major payments company — Stripe, PayPal, Adyen, Shopify, Square, Klarna and the rest — and the total sits around one million. A single channel posting nursery rhymes for toddlers has over 200 million.

Company size has almost no relationship to channel size. Nothing else in marketing behaves that way, and it is why the ground is still open.

Pattern 4: the first 30 seconds decide most of it

Across the channels we examined, the largest single drop in every retention graph was in the opening half minute. Platform-wide, the figure commonly cited is that around 40% of viewers leave in the first 30 seconds.

Brand videos make this worse than they need to, almost always through the same three things: an animated logo, a welcome, and context before anything happens. The best-performing videos in our sample opened on the action and let the context arrive afterwards.

Pattern 5: the packaging is simpler than brands make it

Two consistent observations from the high performers:

Short titles. The strongest performers sat under roughly 55 characters. Concise beat clever almost every time.
Three elements maximum in a thumbnail. Contrast, open space around the subject, and a moment from the video rather than a posed shot.

Brand thumbnails failed in the opposite direction: crowded, logo-bearing, built to satisfy a brand guideline rather than to be read at 120 pixels wide on a phone.

Pattern 6: satisfaction beats both metrics

Click-through and average view duration are the two things the platform can measure. Neither is what it actually rewards.

The concept operators keep returning to is satisfaction — value delivered per minute — and the reason it gets ignored is that there is no dial for it. People optimise what they can see.

The practical version: a video with strong packaging that does not pay off its promise gets one good day and then dies. We saw that shape repeatedly in the underperformers — a strong opening spike followed by a flat line, which is the signature of a video that earned its click and did not earn its watch.

Pattern 7: the television shift changed what is allowed

More than half of viewing on the channels we could measure now happens on TV screens rather than phones, against roughly a quarter in 2021. Average view duration across the platform has risen substantially over the same period.

This matters because it changed what a brand can get away with. Long, loosely cut, unnarrated video now performs where five years ago it would have been cut to ten minutes and lost its character. Several of the strongest videos in our sample would not have been commissioned under 2021 assumptions.

What the failures had in common

Videos about the company: launches, culture, event recaps.
A different format every time.
Thumbnails designed as brand assets.
Publishing driven by internal calendars rather than by audience demand.
Channels that were published to but never run.

Not one of the failing channels in our sample failed on production quality. Several were the most expensively produced in the study.

The data-backed content patterns a YouTube strategist should act on first

If you take three things from 50 channels:

Make the next video about the problem, not the product. It is free and it is the biggest single lever.
Find your best-performing video and make it again. Not the topic — the structure.
Cut your first 30 seconds in half. Start where it gets interesting.

Methodology: 50 brand channels reviewed September 2026, ten of them analysed video by video across full upload history, view distribution and format structure. Platform-level figures are drawn from a 16-source compilation of published operator data. Figures were accurate at time of writing and move constantly.

If you want this done to your own channel rather than read about, that is what a channel audit is.