01Every reel, every channel, one row
One row per reel with a lane per channel: views, the share of the video people watched, and how many views it collected in its first forty-eight hours — the number that says whether a reel landed or merely accumulated. A chip shows which channel led and by how much. Watch-through and the forty-eight-hour figure come from YouTube's analytics; the Instagram side publishes no equivalent, so those cells stay empty rather than guessing, and connecting an Instagram account is waiting on Meta App Review.
02What's working, what's hurting
Your videos are bucketed by what they have in common and each bucket is compared against your own channel's median, so a pattern reads as a multiple of your median across a named number of videos, with its average watch-through. Nothing here is a model writing a narrative — it is arithmetic you could redo by hand, which is exactly why you can argue with it. This card reads your YouTube channel; the per-platform version for other channels is not built yet.
03It refuses to judge too early
A video needs seven days before its views are allowed to count, and recent uploads are listed separately with their age rather than dragging an average around. Below four settled videos it declines to name patterns at all, instead of reading noise back to you as insight.
04Live channel stats
For YouTube: views over thirty days, engagement rate, average watch-through with total minutes watched, and subscribers gained. For Instagram: followers, views, reach and interactions — available once an Instagram account is connected, which for accounts other than ours is waiting on Meta App Review.
05A history that builds itself
A background job records each connected channel about twice daily, so your trend line survives platform rate limits and the page can answer from data you already hold instead of blocking on someone else's API.
06Who is actually watching
Audience composition per channel and, for your strongest videos, per video — enough to notice that the reel that did well did well somewhere you were not writing for.
07It feeds the next script
The same patterns become context for writing, so the next script is drafted knowing which of your hooks earned retention. Suggestions for what to make next stay suggestions until you apply them.