How to turn short-form video data into repeatable results
The dice system: how to read ten short videos that did 200, 500 and 2,000 views, find what worked, and build a format that reaches 10,000 views again and again.
Founders judge their content one video at a time, which is the fastest way to quit. The dice system starts from a batch instead.
From ten videos to one format
Take ten videos. One did 200 views, another 500, another 2,000. I go through the small elements of each one: the first second, the topic, the order of the story, the setting, the tone. I look at what worked and what did not, and from those differences I build a die, a format that can reach 10,000 views and do it again the next week.
A format is a die
Think of every format as a die. It has a few faces (a floor, a median, a top face). One video is one roll and says little on its own, so I judge a format on several rolls.
When we plan a production day, we roll several formats many times. Most videos land between two thousand and twenty thousand views, and the plan expects it. A few land above a hundred thousand, and sometimes one passes a million. The goal is the highest average result with the smallest spread, so the formats that land well and consistently get rolled again.
Two kinds of dice
Mimetic formats borrow a structure that already performs on the platform: the reaction, the teardown, the blind test, the myth-versus-reality. They are cheap to test because the platform already knows how to distribute them. We adapt each one to a trait the founder already has.
Original formats are built from zero around the founder: the way they talk, the argument they always make, the thing only they can show. They take longer to find. When one hits, it becomes the signature nobody can copy.
A healthy library mixes both. Month one tests wide, month two iterates, month three doubles down on the winners and turns them into templates the team can run.
Answering the comments beats copying the trend
When a trend appears, the lazy move is to copy what other accounts are doing. You will get views and add nothing.
With Spaccanapoli, when a tasting video on a viral spread took off, we sorted hundreds of comments into the three questions people kept asking and answered each one with a new video. The trend gave us three new formats, each built from what the audience was already focused on.
Keep, iterate or kill
After 48 hours, every video gets read against the expected distribution of its format.
- Above the median: keep the format, shoot variations.
- Around the median: iterate one variable (hook, length, setting).
- Below the floor twice: kill it.
That judgment, persevere or kill, is where most of the value sits. It is also the part that is hardest to copy.
What this means for your calendar
The founder’s input is one production day a month. Everything else (mining, format design, direction, data, iteration) happens around them. You never write a script.
Want to run the review yourself? The Content Testing Loop handbook gives the weekly routine, and the open-source skill statur-format-review lets your AI assistant read your dice.
Want the dice designed around your ideas? Apply to work with me.
A format is a loaded die.
Each format is a die you throw every week. I build each founder the set of dice that lands the highest value with the smallest spread.
[ MINE ] ──▶ [ PACKAGE ] ──▶ [ THROW ] ──▶ [ READ ]
ideas hook + gap publish retention
on camera + format the batch saves · sales
▲ │
│ ● keep ◐ iterate ○ kill │
└──────────────────────────────────────────────┘Run the review yourself: the Content Testing Loop handbook gives the weekly routine, and the open-source skill statur-format-review lets your AI assistant read your dice for you.
How many videos should a founder shoot per day?
In the Founder Brand Program, one production day produces a batch of videos across several formats. Volume is what lets the data find the winners.
What is a mimetic format?
A structure that already performs on the platform, adapted to the founder's own traits. Original formats are built from scratch around the founder.