Research · Q3 2026
What 1,900 apps say about
creator-led growth.
We went through a tracked sample of 1,900 apps and 370 published growth breakdowns from July to September 2026, counted what could be counted, and wrote down what the numbers do and do not support. Method and limits are at the bottom, not hidden in a footnote.
Counted across the 751 apps in the sample with any tracked short-form views, as of 27 September 2026. Views are social views, not installs.
Finding 1
Reach is more concentrated than anyone budgets for.
Of 1,900 apps, 1,149 (60.5%) show no tracked short-form views at all. Among the 751 that do, 37 apps (4.9%) account for half of every view, and the top tenth account for two thirds.
If your plan assumes a typical result, it is a plan for an app that does not exist. The median is not where the views are, so a channel should be judged on whether it can find a winner, not on the average creator.
View as table
| Top apps by views | Apps | Share of views |
|---|---|---|
| Top 4.9% | 37 | 50% |
| Top 10% | 75 | 66% |
| Top 19.6% | 147 | 80% |
| All | 751 | 100% |
Finding 2
More accounts means less reach per account, and more revenue.
Median views per account fall steadily as a brand adds accounts. The share of apps reaching the top revenue bucket does the opposite. Depth of distribution beats the size of any single account.
| Accounts | Apps | Median views per account | Reach $100K+ |
|---|---|---|---|
| 1 to 2 | 156 | 7.7M | 11.5% |
| 3 to 10 | 252 | 3.1M | 11.5% |
| 11 to 25 | 151 | 1.3M | 18.5% |
| 26 to 50 | 101 | 1.5M | 27.7% |
| 51+ | 86 | 0.9M | 29.1% |
Correlation is not causation: bigger brands both run more accounts and earn more. We read this as direction, and test it by running two or three accounts first. JobStep’s founders reported going from 16 accounts in May to 70 by August on the way to $2M ARR.
Finding 3
Views barely predict revenue.
Of the 504 apps with both 10M+ tracked views and a revenue estimate, 246 (49%) sit in the lowest revenue bucket. Only 115 (23%) reach $100K or more. Across all 746 apps, the rank correlation between views and revenue is 0.36: real, and weaker than most reports imply.
The sharpest single example in the published breakdowns is Saner.AI: about 1.2M views and under 5K downloads. Reach without a bridge to the product is a number for a slide, not for a P&L. This is why we lead reports with revenue and treat views as a diagnostic.
The case file
Nine published results worth studying.
Figures reported publicly by the companies themselves, or counted from public social data. They are other companies’ results, unaudited, and we have not verified any of them. What they teach is in the right-hand column.
| App | Reported result | What it teaches |
|---|---|---|
| JobStep | $0 to $2M ARR in under 8 months; 70 creator accounts; 100M+ views over nearly 9K videos (founders) | Account count and early localisation, not one viral video |
| PushClock | ~$40K MRR; 347M views; one creator ≈70% of reach (reported) | Find the creator who works, then clone the person |
| Potto | $100K MRR in 50 days; one video 25.8M views, 380K bookmarks (reported) | A single hook, re-shot with a new noun |
| Mise | $100K MRR in 5 weeks (reported) | The same discovery hook, a different niche |
| Stella | ~$300K MRR in about two months; MVP built in two days (reported) | Audience first; the build was the cheap part |
| Pecra | $200K MRR on roughly a fifth of a rival’s views (reported) | A harder paywall can substitute for reach |
| Verbi | 125M views over 43 accounts; ~$20K MRR in 5 months (reported) | Views-rich, revenue-poor niches exist |
| SideShift | One post at 16.5M views; downloads stalled near 20K (tracked) | The winner was run once. Re-run winners |
| Saner.AI | ~1.2M views; under 5K downloads (tracked) | Views without a product bridge do not convert |
So what
How this shapes the way we run campaigns.
Judge the channel on its ability to find a winner.
Reach is concentrated, so we plan for a hit rate across many concepts rather than for an average creator. That is the reason for running volume before judging.
Prove a hook on two or three accounts, then clone it.
The breakouts in the data scaled one proven hook by changing a single variable, not by inventing new creative each week.
Report revenue first, views as a diagnostic.
With a rank correlation of 0.36, a views-led report can look excellent for an app that earns nothing.
Re-run the winner.
The clearest preventable loss in the case file is a hook that worked once and was never repeated.
Localise early.
Spanish-language and German or French accounts over-delivered repeatedly in the published breakdowns. It is one of the cheapest multipliers available.
Method and limits
What this does and does not show.
Data. A tracked sample of 1,900 apps compiled from public social data (views, accounts, videos by platform) with bucketed revenue estimates, as of 27 September 2026, plus 370 published growth breakdowns from July to September 2026. Among the apps with tracked views, 89.5% of tracked accounts are on TikTok, and the median video has about 27K views.
What we computed. Concentration (cumulative share of views by rank), account bands (apps grouped by number of creator accounts, median of views divided by accounts), and a Spearman rank correlation between views and the revenue bucket (n = 746 apps with both). Counts from the written breakdowns used keyword matching and are approximate.
Limits.
- Views are social views, not installs or users. Apps that are not tracked socially show zero even when large.
- Revenue is bucketed and estimated. Founder-reported MRR is unaudited.
- Survivorship: breakdowns are written about winners. We do not see the apps that ran the same hook and flopped.
- Brands running 40 to 200 accounts inflate their own totals.
- Formats last roughly a month. Treat anything here as a snapshot, not a rulebook.
None of the apps named on this page are HiveMind clients, and none of these results are ours.
Free distribution audit
Want this applied to your product?
Forty seconds of questions, then a written teardown that tells you which of these patterns fits your category, and which would waste your budget.