Harloop Research · Study 01YouTube

Which videos get watched on a mid-size company's YouTube channel

November 1, 2026 · Chinmay Pingale

1.41x
views for product demos, against the videos posted around them
57 of 81
channels where demos beat the videos around them
18,822
videos compared, from 400 companies drawn at random

Key findings

  1. Videos that show the product working got the most views. Product demos got 1.41x the views of the videos posted around them on the same channel, and beat them on 57 of 81 channels. How-tos came second, at 1.13x.
  2. Videos about the company itself got fewer. Webinars (0.91x), culture videos (0.87x), event recaps (0.84x) and customer stories (0.81x) all came in under a typical video for their channel.
  3. Five to twenty minutes did best. By length, videos of 5 to 20 minutes scored 1.20x and videos under a minute scored 0.83x.
  4. The result held under every check we ran: a different number of neighbours, removing videos that may have been boosted with ads, and reading the sorting rules by hand.

The question

When a company with 50 to 500 employees posts a video on YouTube, which kind gets watched?

Every marketing team has this argument when it plans the next video. Should it be a customer story, a demo, a recap of the conference? We wanted to answer it with data instead of taste, so we counted.

Step 1: Find the companies

We started from a list of private US companies with 51 to 500 employees and drew 400 at random, so the sample is not made of companies we already knew. We loaded each homepage and looked for a link to the company's YouTube channel. About half had one. Then we listed every long video on each channel (Figure 1).

From 400 companies to 18,822 videos

Each step of the sample, and what was kept

  1. 400US companies drawn at random from a list of private firms with 51 to 500 employees, half software and half industrial
  2. 379homepages we could load and read
  3. 191of those link to a YouTube channel
  4. 166channels have at least 10 long videos
  5. 18,822videos, each with its title, length and view count
Figure 1How the sample was drawn. Companies were drawn at random from a list of private US firms with 51 to 500 employees on LinkedIn, 200 software and 200 industrial. Homepages were read on November 1, 2026.

Step 2: Make the comparison fair

Raw view counts don't work. A company with 10,000 subscribers gets more views on everything than a company with 100. And a video from 2019 has had years longer to collect views than one from last month.

So we never compare one company's videos with another company's. We compare each video with the 10 videos the same company posted around it: five before and five after. Same audience, about the same age. The thing that changes is the video.

score=views of the videomiddle of the views of the 10 videos around it(1)

The middle is the median, taken on a log scale so that one viral neighbour can't drag it.1 A score of 1.0x means a video did as well as a typical video around it. 2.0x means twice as well. Figure 2 works one through.

One video against the 10 around it

Fireflies AI's channel, newest at the top · views

  1. How to Add Fireflies to a Meeting: 5 Ways to Invite the Bot10,000
  2. Introducing Fireflies For Healthcare455
  3. How to Record Google Meet Without a Bot: Fireflies Chrome Extension12,000
  4. Fireflies for VCs - Turn Every Investment Call Into Investment Intelligence335
  5. How Fireflies Helps VCs Capture, Analyze, and Act on Every Meeting325
  6. Introducing Fireflies Live Assist1,800
  7. Introducing Fireflies Desktop App661
  8. Fireflies Sales Assist — real-time coaching & answers during sales and customer calls1,300
  9. Can Admins view all meetings inside your Fireflies workplace?252
  10. How to Connect Zoom With Fireflies (Admin Setup Guide)1,200
  11. What Is Fireflies Live Assist? Real-Time AI for Your Meetings5,300

1,800 ÷ 900 = 2.0x

Figure 2"Introducing Fireflies Live Assist" scored 2.0x. It got 1,800 views. The middle of the ten videos posted around it is roughly 900, between 661 and 1,200.

Step 3: Sort the videos by kind

We sorted every video by the words in its title (Table 1). Titles are a rough guide, and we treated them that way: 56% of the videos matched none of the rules, mostly product names on industrial channels ("Bulk Bag Filling System"). Those still count as neighbours, but we don't report them as a kind of their own.

How a title becomes a kind of video

Kind of videoWords in the title that put it there
Product demo or launchdemo, introducing, new feature, overview, what's new, release, launch
How-to or tipshow to, tutorial, tips, setup, guide, explained, best practices
Webinar or podcastwebinar, podcast, episode, fireside, panel, keynote, or any untitled talk over 20 minutes
Company or cultureabout us, our story, culture, careers, meet the team, anniversary
Event or trade showconference, expo, trade show, booth, summit, recap, highlights
Customer storytestimonial, case study, customer story, success story, How (a customer) uses…
Table 1The rules that sorted the videos. Rules are tried top to bottom and the first match wins. 56% of videos matched none and count only as neighbours.

We read random samples of every kind by hand and fixed the rules where they were wrong.2

Step 4: Give every channel one vote

For each channel, we took the middle score of its demos, the middle score of its how-tos, and so on. Then we took the middle of those across all the channels. A channel with 1,000 videos counts the same as one with 12, so a few busy channels can't decide the answer.

What we found

Videos that show the product, demos and how-tos, beat the company's other videos (Figure 3). Product demos did best: 1.41x the views of the videos around them, and ahead on 57 of 81 channels.

Videos that show the product beat the rest

Views against the 10 videos posted around it, by kind of video (1.0x = typical)

Product demo or launch1.41x
How-to or tips1.13x
Webinar or podcast0.91x
Company or culture0.87x
Event or trade show0.84x
Customer story0.81x
Show the numbers
Kind of videoViews vs the videos around itChannels where it wonVideos
Product demo or launch1.41x57 of 811,000
How-to or tips1.13x69 of 1072,402
Webinar or podcast0.91x36 of 933,417
Company or culture0.87x31 of 75523
Event or trade show0.84x17 of 44282
Customer story0.81x15 of 49423
Figure 3Product demos got 1.41x the views of the videos around them. Each bar is the median across channels of each channel's median score, so every channel counts once. N = 18,822 videos from 166 channels; a channel counts toward a kind when it posted at least two of them.

Every kind of video made about the company itself came in under 1.0x: webinars, culture videos, event recaps and customer stories.

Length mattered less than kind, but it points the same way (Figure 4). People who go to a company's YouTube channel seem to want the long, useful version.

Five to twenty minutes did best

Views against the 10 videos posted around it, by length (1.0x = typical)

Under 1 minute0.83x
1 to 2 minutes0.96x
2 to 5 minutes1.12x
5 to 20 minutes1.20x
20 minutes or more0.98x
Show the numbers
LengthViews vs the videos around itChannels where it wonVideos
Under 1 minute0.83x38 of 1403,043
1 to 2 minutes0.96x70 of 1574,466
2 to 5 minutes1.12x96 of 1555,187
5 to 20 minutes1.20x80 of 1232,351
20 minutes or more0.98x39 of 883,775
Figure 4Videos under a minute did worst, at 0.83x. Same comparison and counting as Figure 3. Shorts sit on a separate tab on YouTube and are not included.

How we checked it

There were four ways this could have been wrong, so we tested each one.

  1. Are the neighbours really the same age? We looked up the upload dates of 39 videos and the videos around them. The 10 neighbours were posted within about four months of each other (a median of 132 days), while the typical video was about three years old. All 39 were in upload order.
  2. Does the answer depend on using 10 neighbours? With 6 neighbours, demos scored 1.29x. With 20, they scored 1.44x. Either way demos stayed on top, how-tos stayed second, and all four kinds of video about the company stayed under 1.0x.
  3. Were launches just boosted with ads? Companies sometimes pay to push a launch, which inflates its views. We removed every video with 10x or more the views of its neighbours. Demos still scored 1.31x and were ahead on 53 of 79 channels.
  4. Did the title rules sort the videos correctly? We read samples of each kind and fixed the rules five times before these numbers. Every title and the kind we gave it is in the data, so you can check our calls.

What this does not show

  • Views, not sales. The video that gets watched more is not necessarily the one that sells more. We only measured views.
  • The demo result is mostly software. Demos scored 1.50x on 56 software channels and 1.12x on 25 industrial ones.
  • No Shorts. YouTube keeps Shorts on a separate tab, and they are not in these numbers.
  • Companies with a channel. We only looked at companies that link a YouTube channel from their homepage, about half of them. The other half may be different.

What to do with it

If you're planning your company's next YouTube video, make it show the product doing the job, and give it room: 5 to 20 minutes did best. The event recap and the culture video may be worth making for other reasons, but on YouTube they are not what people come for.

Suggested citation

Pingale, C. (2026). Which videos get watched on a mid-size company's YouTube channel. Harloop Research, Study 01. harloop.com/research/youtube-video-types

The data behind every number here is linked under "Data" at the top of the page: every video and every channel. View counts were read from YouTube on November 1, 2026.

Footnotes

  1. The ten neighbours in Figure 2 have a middle of 890 on a log scale; halfway between the middle two by hand is 930. Both round to 900 and give the same 2.0x. ↩

  2. The five fixes: integrations ("ORO + Coupa") were being counted as customer stories, and so were "How To Use…" videos, employee testimonials, and companies describing themselves ("How Fireflies Helps VCs…" on Fireflies' own channel). Press coverage ("Fox Features SpotHero") was being counted as a product feature. ↩