Sample result
Video Retention Analyzer for Restaurants
The owner asked this:«Here is the data from my last 6 reels. I want to know why some hit and others die, and what I shoot next week.»
① The diagnosis, grouped by duration band (3 of 6 shown)
| Video | Duration | Ret. 3 s | Ret. 50% | End | Where the leak is | Which script part fails |
|---|---|---|---|---|---|---|
| The sancocho spoon | 24 s | 78% | 61% | 44% | No leak: soft and even decline | None: this is the account's healthy curve |
| Tour of the new location | 58 s | 71% | 32% | 12% | Collapse from second 8 to 20 | Development: twelve seconds of hallway with no food and no people |
| «Hi, welcome to our restaurant» | 31 s | 39% | 22% | 15% | Death at second 2 | Hook: the logo greeting scares people off before showing anything |
The comparison is normalized. The 24 s reel at 61% and the 58 s one at 32% are not read with the same yardstick, which is why the table groups by band before issuing a verdict.
SUPUESTO: no industry «good retention» figure is hard-coded here, because it shifts by platform and by year. The threshold used is this account's own median over the last 8 weeks, and it says so. If your format's real median were 10 points higher, the location tour would move from amber to red and be dropped sooner.
② The account's pattern, in three findings with figures
1. A close-up of food in motion holds; the greeting kills. The 3 videos opening with product in the first 2 seconds average 74% at 3 s; the 2 opening with a greeting or the storefront, 41%.
2. Your audience holds for 25 to 30 seconds. The two videos under 30 s keep more than double at the end compared with those from 50 s up; long material is not bad, it is badly cut.
3. On-screen text carries the silent stretch. The only long video that does not collapse carries captions every 4 to 6 seconds; the «clean» ones lose whoever watches without sound, which is most of the lunch crowd.
Hypothesis, still without three videos behind it: closers that say the price out loud hold less than those that only show the dish. It gets tested, not declared.
The full example has 1 more part(s): you see them inside the library, with your account.