Trial Bench: Trial Reels

Instagram’s Trial Reels show a Reel to people who do not follow you first, so a miss never reaches your own feed. Trial Bench tracks each one and tells you what to do with it. Start a trial by switching on Trial Reel when you schedule a Reel on the Plan page. Choose Manual (you graduate it yourself in the Instagram app) or Automatic (Instagram graduates it if it performs well), and optionally note what you are testing in the form “Axis: side”, for example “Opening: question”. You can also start tracking a Reel that is already published from the Record tab.

The Live tab shows each trial with the hours elapsed and when the call is due, a verdict, Instagram’s own numbers (follows per 1,000 views, watch-through, cold reach and shares, each set against your own median) and what happened after people left Instagram (DM replies, page visits and seven-day revenue). The verdict is a rule, not a mood. A trial is judged on four thresholds against your own archive of closed trials. Clearing three or more means Promote to followers. Clearing one or none with more than half the window gone means Don’t promote. Anything in between is Hold.

Two things make a verdict wait. Until you have five closed trials to compare against, every trial reads Hold, because there is nothing of yours to compare it with. And a trial whose opening looks like a recent one is flagged as a near-duplicate: Instagram suppresses reach on trials like that, so the numbers likely understate the content, and the verdict holds until the window closes.

The verdict and Instagram’s own four numbers are on every plan. On Pro and Max each trial also shows the confidence read and its hour-by-hour trajectory, the money panel (DM replies, page visits and revenue), the seven-day revenue on every Record row, the which-trials-made-money ranking in Findings, and the threshold calibration card. Below Pro those show a lock.

Confidence to promote compares how this trial is tracking hour by hour with every past trial’s full 24-hour curve, then shows what share of similarly shaped trials went on to beat your grid median. It relies on the snapshots knit.bio keeps while trials run, so it fills in as you use Trial Bench.

The other tabs build on the same record. Findings turns closed trials into plain lines such as which opening won: it compares the two most-tested sides of each axis on follows per 1,000 views, needs at least eight trials on an axis, and calls a small difference flat instead of a win. Queue holds test ideas you want to run next. Record is your archive of every past trial with its outcome (promoted, held or not promoted).

Two limits are Instagram’s, not knit.bio’s. No tool can graduate a manual Trial Reel; only the Instagram app can, so a Promote verdict tells you to open Instagram and tap Share with everyone. And Instagram does not reveal that a Reel was a trial when it is read back, so Trial Bench only knows about trials that were started or tracked here.