When I collected every Short twelve brand channels published in 90 days, two of the twelve turned up empty. Notion and Grammarly both maintain a Shorts tab and neither had posted to it in the window. Chipotle managed one. Zapier managed one.
An empty 90 days is a more interesting thing than a busy one, because somebody built that channel on purpose. So I went back and pulled the complete Shorts history of all four: 276 videos, with their dates and public counts, read on 5 September 2026.
What the four have in common is not failure. It is a shape.
| Channel | Subscribers | Shorts on the tab | First | Last | Days since | Median views | Best Short |
|---|---|---|---|---|---|---|---|
| Zapier | 78,000 | 120 | 2024-10-17 | 2026-07-13 | 54 | 755 | 260,859 |
| Grammarly | 240,000 | 59 | 2022-08-29 | 2026-05-06 | 122 | 1,456 | 517,740 |
| Chipotle | 106,000 | 55 | 2023-05-19 | 2026-07-30 | 37 | 6,311 | 207,550 |
| Notion | 349,000 | 42 | 2023-04-03 | 2026-04-27 | 131 | 5,445 | 25,474 |
The shape of a programme that stops
Zapier is the clearest case. Its Shorts tab starts in October 2024 and immediately becomes a real operation: 17 Shorts in December 2024, 16 in January 2025, 15 in February, 10 in March, 16 in April. That is a team with a schedule.
Then May: 7. June: 4. July: 2. August: 1. September: none. October: 5. November 2025: 15, which looks like a deliberate restart. Then March 2026: one. July 2026: one. Nothing since.
Grammarly’s history has the same signature with a longer gap in the middle. A first run of 23 Shorts from August 2022 to August 2023, then nothing for almost two years, then 14 Shorts in July 2025 followed by 3, 1, 4, and then a final tail through spring 2026 that ends on 6 May.
Notion never really ran a programme at all: 42 Shorts spread over three years, never more than five in a month, last one 131 days ago. Chipotle is similar but with a real burst in early 2025, 11 in February alone, then a slow fade.
Two of the four ran hard for a period, stopped, restarted once, and stopped again. The restart is the tell. Somebody made the case a second time.
The hits did not save any of them
The obvious explanation for a stopped programme is that it did not work. The numbers do not support that cleanly.
Zapier’s three best Shorts took 260,859, 206,370 and 197,300 views. All three landed in its first three months. On a channel of 78,000 subscribers, those are large numbers. The median Short across the whole run is 755 views.
Grammarly’s best Short took 517,740 views, published on 2 July 2025 during its heaviest month. Its second best took 129,399, nine days later. The median across all 59 is 1,456.
Chipotle’s best took 207,550, a video about how to roll a burrito. Its median is 6,311.
So each of these channels produced at least one genuine hit, in Zapier’s case three in a quarter, and each stopped anyway. Whatever ended these programmes, it was not the absence of a big number.
The more likely reading is arithmetic. A channel where the median Short takes 755 views and the exceptional one takes 260,000 is a channel where almost every video is invisible and the occasional one is enormous. That is the same outlier-carried distribution that shows up everywhere in this kind of data, and it is very hard to defend in a quarterly review, because the typical output of the programme is the median, and the median is small.
What the back catalogue is doing now
All 276 of these videos are still public and still accumulating views, and there is no way to tell from a single reading how fast. That question needs the same video measured twice, which is a collection now running and a piece for later.
What can be said from one reading: the median Short on Notion’s dormant tab has 5,445 views and the median on Chipotle’s has 6,311, both of which are higher than the median of several channels that are still publishing weekly. A stopped channel is not a dead one. It is a small library that nobody is adding to.
What this suggests if you are about to start one
- Look at your own median before you commit, not your best. Every one of these four could point at a hit. All four have a median that would embarrass a budget review.
- Plan for the restart. Two of the four restarted after a gap, which means somebody had to rebuild the case, the format and the habit. That second start is more expensive than the first.
- A burst is a decision, and so is stopping. Fifteen Shorts in a month followed by one in the next four is a resourcing story, not an audience story. If the programme depends on one person’s fortnight, its end date is already set.
- The tab persists. Every one of these channels still shows a Shorts tab to anyone who looks, with the last upload date on it.
What this cannot tell you
- Four named channels, chosen because they published nothing or almost nothing in a 90 day window I had already collected. Not a sample of anything.
- Public view counts read once, on 5 September 2026. No impressions, no watch time, no retention, no traffic sources, and no idea what any of these videos cost or returned.
- Nothing about why anyone stopped. There are no interviews here and no internal information. The shapes described above are what the upload dates show, and the reasons are not visible in public data.
- Extraction was capped at 120 videos per channel. Zapier reached that cap, so its Shorts tab may hold more than the 120 collected, and its first upload date is the earliest of those 120 rather than necessarily the channel’s first ever Short.
- Videos deleted or made private are invisible, so a programme that ran and was then cleaned up would not appear here at all.
Source ledger
- Video data: the public YouTube Shorts tab for each channel and each video page, extracted per video on 5 September 2026. 276 videos across four channels: Zapier 120, Grammarly 59, Chipotle 55, Notion 42.
- Every count in the tables and the text is as read on 5 September 2026 and will have moved since.
- The 90 day collection that produced the original empty result is described in Twelve brand channels, 550 Shorts, side by side.
