Social media is an economy
Creators supply videos. Viewers supply attention. The price is the views each video gets.
Niches are industries. Formats are markets.
Niche
A broad audience, like AI. It holds many formats (tutorials, news, launches) that compete for the same viewers, the way Coke competes with coffee.
Format
A repeatable video shape. Inside a format you compete head-to-head with everyone else making that shape.
The niche sets the ceiling. A good sports video can reach millions. A good B2B software video might reach 10,000. Judge every result against its own market.
Spot the outlier
A big view count alone tells you little. What matters is how far a video out-performs the audience it started with.
The audience did it
1M views on an account with 10M followers.
The video did it
1M views on an account with 1,000 followers.
When small accounts keep breaking out in the same format, that format is under-supplied. One breakout is luck; a cluster is a market.
| Signal | Threshold |
|---|---|
| Watch | Views at least 3× followers |
| Outlier | Views at least 10× followers |
| Breakout | 100k+ views from an account under 10k followers |
| Proven format | 3 or more unrelated small accounts breaking out in it within about 90 days |
Three ways to win a market
Supply the best video
70M
Best means what that audience prefers: simpler, funnier, more emotional or more edited. If a format has frequent outliers but low average quality, there's room.
A basketball quiz format, rebuilt with better questions, polish and pacing, got 70M views in its first 10 videos on fresh accounts.
Import a proven format
4M+
A format that has worked hundreds of times elsewhere has proven fundamentals. Bring it to a niche that doesn't have it yet.
The “$1 vs $X” challenge format, rebuilt as “$1 vs $1M engineer” for AI and tech, got 4M+ views across platforms.
Dominate through scale
280M
One person's output caps out. Script a winning format and send it to many creators, or to a pipeline that varies it.
A creator program of 100+ creators posted 14,000 videos in a few months and reached 280M views.
The math of scale
- Expect a power law. About 10% of videos will carry most of the views. The long tail still adds up.
- Keep the pipeline fresh. Feed in new formats, kill tired ones, and keep scaling what works.
- Code can be your creator program. A format that can be generated, like a simulation with different rules each time, scales without hiring. Vary every output, because TikTok swaps out near-duplicates.
The window closes
When a format visibly works, everyone copies it. Supply rises and views per video fall, so speed is the edge. The exception is a format with a high barrier: one few creators can execute well.
Signs it's time to leave
- Big accounts start making it
- Outlier ratios drop month over month
- Comments say “this trend again”
- Low-effort copies flood the hashtag
Execute every time
Picking the right market gets you in. Craft decides whether the algorithm carries the video.
- Hook in three seconds. Put the payoff, tension or question on screen. Never open with a greeting.
- Protect watch time. Cut every pause, change something on screen every few seconds and tease the best moment early.
- Stay original. Your own performance or commentary must be most of the video. Reposts and watermarked clips are kept off the For You feed.
- Package for Search. Say, show and write the keyword. Add a custom cover, a caption that works as a second hook, and 3–4 relevant hashtags.
- End with one specific ask. “Pick a color before it starts” beats “like and follow.”
The full production rules (filming, sound, captions, compliance and monetization) are in Playbook 01: the TikTok Bible.
The weekly loop
- 01Scan
Pull 50–100 hashtags or keywords. Compute outlier ratios and how recent the hits are.
- 02Score
Rate each format on the six factors in the scorecard.
- 03Pick
Choose one “best video” play and one “import” play for the week.
- 04Produce
Write the hook first, then the beat sheet. Shoot, edit and package.
- 05Measure
At 48–72 hours, read 3-second drop-off, full-watch rate, saves, shares and follows per 1,000 views.
- 06Scale or kill
Turn winners into a series or pipeline. Drop a format after three misses.
The opportunity scorecard
Score each factor from 0 to 5, for 30 points in total. Anything above 22 is worth a 10-video test.
| Factor | What a 5 looks like |
|---|---|
| Demand | Median views in the format far above the niche median |
| Small-account proof | Many accounts under 10k followers with 100k+ views, and a high typical video, not just one hit |
| Supply gap | Few recent videos, or low average quality among the incumbents |
| Freshness | The outliers were posted in the last 60–90 days |
| Fit | You can make the best version with the skills, tools and access you have |
| Barrier | Hard for others to copy once you're in, through craft, code or IP |
What the data says
We tested this playbook on 3,416 TikTok videos from 90 hashtags across AI, game development, 3D and broad niches in November 2026.
Three lessons
Check the median, not the max
Most small accounts with a giant hit had a typical video of 1–3k views. That's luck, not a market. Only formats where the typical video wins are worth entering.
Volume without craft loses
One account posted 348 videos in a hot format and got a median of under 7k. A 20-day-old account in the same format had a median near 1.5M.
Lead with the outcome
Videos that named a tool were crowded and flat. Videos that led with what was built, or a question about the world, broke out.
| Market | Median views | Median followers | Videos ≥10× | Read |
|---|---|---|---|---|
| #simulation | 1.6M | 14k | 83% | Open and current |
| #physics | 2.05M | 52k | 83% | Big demand |
| #indiegame | 367k | 8.5k | 80% | Small accounts break out |
| #devlog | 49k | 1.5k | 83% | Low ceiling, easy entry |
| #aiseries | 80k | 4k | 61% | Supply flooding in |
| #aitools | 127k | 9k | 43% | Oversupplied |
| #cursorai | 12k | 2k | 41% | Oversupplied |
Views are lifetime totals and follower counts are current, so older hits overstate ratios. We weighted by recency. Hashtag pages show top videos, not a random sample.