How many reels go viral, counted across 2,586 of them

Ask how many reels go viral and you get a percentage somebody made up. We went back through every Instagram reel we have ever scraped for a client account and counted.

That is 2,586 reels across 18 Canadian business accounts, posted between September 2020 and September 2026, carrying 87,921,359 views between them. Sorted from the top down, half of all those views come from 42 reels. That is 1.6 percent of the library.

How many reels go viral out of every hundred?

About two in a hundred carry the account. The top 1 percent of this library, 26 reels, holds 41.9 percent of all views. The top 2 percent, 52 reels, holds 54 percent. Everything below the top 10 percent accounts for 17.6 percent between 2,327 reels.

Put the other way: 436 reels, one in six, never reached a thousand people. Those were not lazy posts. They were written, shot and edited the same way as the ones that worked.

The ladder is steep the whole way down. The top half of one percent, 13 reels, holds 29.5 percent of views. The top 5 percent holds 70.8 percent. The top 10 percent holds 82.4 percent. The bottom half of the library, 1,293 reels, holds 2.7 percent between them.

That last number is the one worth sitting with. Half of everything produced across six years accounts for roughly one view in forty. Any plan that assumes a typical reel carries its own weight is planning against the data.

Why the average view count misleads

The mean reel in this set did 33,999 views. The median did 4,514. The mean is 7.5 times the median, and that gap is the whole finding, because the mean is the number agencies quote.

If you are shown a case study saying an account averaged 60,000 views per reel, you are almost certainly looking at one reel that did several million and a hundred and fifty that did under three thousand. The average is arithmetically true and tells you nothing about what the next reel will do.

Inside a single account it gets worse. One realtor account here has 217 reels, a mean of 9,104 and a median of 359. The mean is 25 times the median. One reel did 793,833 views. The typical reel did not clear four hundred.

What the distribution looks like by size

Eighteen reels, seven tenths of one percent of everything posted, produced more than a third of all views. The single largest group, 1,482 reels between one and ten thousand views, produced 7.2 percent between them.

Over 1M views: 18 reels, 0.7 percent of the library, 35.5 percent of views.
100K to 1M: 134 reels, 5.2 percent, 38.1 percent of views.
10K to 100K: 516 reels, 20 percent, 19 percent of views.
1K to 10K: 1,482 reels, 57.3 percent, 7.2 percent of views.
Under 1K: 436 reels, 16.9 percent, 0.2 percent of views.

What this means if you are buying

Judge an agency on the median, and on the count. Ask how many reels the account published to produce the number you are being shown. A 4.6 million view reel inside a run of 159 is a different proposition from one inside a run of 12.

Treat anyone promising you a viral reel as selling a lottery ticket. Across these 18 accounts the median account needed 22 reels before its first one crossed 100,000 views. The mean was 53. One account hit it on reel number one. Another did not get there until reel 206.

If you want one threshold to hold an agency to, use 100,000 views. In this library 152 reels of 2,586 cleared it, which is 5.9 percent. An account posting twice a week for a year produces about 104 reels, so at that rate roughly six of them clear the bar, and which six is not knowable in advance.

Volume is the only input you control. The 42 reels that carried half the views were not identifiable in advance, and we would be lying if we said we spotted them coming. Every one of them sits inside a run of consistent posting. You can see the full working on any of our published client case studies, which publish the median beside the mean for exactly this reason.

What we changed because of this

We publish the median beside the mean on every case study, and the count of reels that produced it. Not modesty. The mean measures how lucky an account got and the median measures what the work reliably produces, and a buyer comparing agencies needs the second one.

We also stopped treating a flat reel as a failure worth diagnosing. When 57 percent of everything lands between one and ten thousand views, a reel in that band is the expected outcome rather than a mistake to investigate. The reviewable question is whether the run is producing its share of the top band over months, not whether Tuesday did better than Thursday.

And we quote ranges rather than single figures when somebody asks what to expect. A median of 4,514 with a mean of 33,999 is not one promise, it is two different statements about the same library, and quoting only the larger one is how this industry earned its reputation.

The part that is against our interest

Our clients' median reel is in the low thousands. We publish that on every case study and we are publishing it again here in aggregate.

Four of the eighteen accounts in this dataset never produced a reel over 100,000 views. That is 22 percent of the accounts we hold reel level data for. They are still counted in everything above. We did not drop them to improve the numbers, which is the normal thing to do with a dataset like this and the reason most agency data is worth nothing.

How to check it

Every account in this dataset is public. Every figure comes from a full scrape of the account rather than a screenshot or a dashboard export, deduplicated by post, taking the highest view count where a reel was captured more than once.

The 18 accounts are not all one trade. They run from realtors through immigration lawyers to restaurants and coaches, which matters because the shape of the distribution barely changes between them. The concentration is a property of the format rather than of any one industry.

Two limits worth stating. This is our scrape cache, which holds 18 accounts and the reels captured across 462 scrape runs. It is not everything we have produced. And the window includes each account's own history from before we started working with them, deliberately, because the distribution above is a fact about short-form video rather than a claim about our work.

We also hold 6,257 cached reels from accounts that are not clients. Those are excluded from every number here. They were scraped as benchmarks, meaning they were chosen because they perform well, and averaging them in would be picking the sample on the result.

---