Judging at 24 hours
Shorts and Reels keep distributing for days. Compare clips at the same age, ideally 7 days.
AI clipping tools tell you a clip's predicted score. Only the platforms tell you what it did. This page covers the handful of metrics that explain clip performance (hook rate, retention, completion, shares per view, follows per view), where to read them on TikTok, Shorts and Reels, how to feed them back into the clipper, and includes a batch analyzer: paste a week of clip numbers and it ranks them, finds the pattern and tells you what to change.
A clipper's score is a prior. The platforms publish the posterior for every clip you post: a retention curve, a completion rate, shares, saves, follows and where the views came from. Most clippers look at view counts and stop. Views are mostly a function of distribution luck and posting time; the ratios underneath them are what you can change.
Each one maps to a step of the clipping process, so a weak number tells you what to fix.
Share of viewers still watching at the 3-second mark (YouTube Shorts shows "viewed vs swiped away"; TikTok and Instagram show the retention curve). Under 60% means the first line or first frame failed; fix the in-point or the title card. Over 75% is strong.
The area under the retention curve. 50%+ on a 40-second clip is healthy; above 100% means loops. A steep mid-clip drop marks a dead stretch to cut; a cliff at the end means the payoff came too late or the clip ran past it.
Viewers who reached the end. This is the number the algorithms weight most for distribution, and the one most affected by length. If completion is low but hook rate is high, the clip is too long or has no ending.
The reach multiplier. Ten or more shares per thousand views means the clip spreads beyond your followers; it's also the best proxy for "quotable". Track as a ratio, not a count, so small clips compare with big ones.
The business metric. For creators, new followers per thousand; for clippers on campaigns, confirmed views and any click-through the brief pays on. A clip with high completion but zero follows entertained without selling the source; add the creator's name and a reason to visit.
Paste one clip per line: title, views, avg % watched, shares, follows (shares and follows optional). Export from TikTok Studio, YouTube Studio or Instagram insights, or type a week's worth. Nothing leaves your browser.
| # | Clip | Views | Avg % watched | Shares / 1K | Follows / 1K | Craft score | Diagnosis |
|---|
Craft score = 0.5 × avg % watched (capped at 100) + 0.3 × min(shares/1K, 30)/30 × 100 + 0.2 × min(follows/1K, 10)/10 × 100. It deliberately ignores views so a 2,000-view clip with great ratios ranks above a lucky 50,000-view one; that's the clip to make more of.
| Source | Retention curve | Completion / loops | Shares, saves | Follows from clip | Traffic source | Notes |
|---|---|---|---|---|---|---|
| TikTok Studio | Yes, per second | Yes, "watched full video" | Yes | New followers per post | For You vs Following vs Search | Export per video; 7/28/60-day windows |
| YouTube Studio (Shorts) | Yes | Viewed vs swiped away; avg % viewed | Shares | Subscribers per video | Shorts feed vs channel | Advanced mode exports CSV |
| Instagram Insights (Reels) | Yes (retention) | Avg watch time, plays | Shares, saves | Follows per Reel | Followers vs non-followers | Professional account required |
| Clipper schedulers (OpusClip, Klap, Katto, Vizard) | — | — | Views, likes on posted clips | — | — | Convenient roll-up across platforms; ratios still need native analytics |
| Campaign boards (StreamClipping, Whop) | — | — | — | — | — | Confirmed views at source after the hold (14 days on StreamClipping); what you're paid on |
Native analytics are the only place retention lives. A clipper's dashboard showing views across six platforms is useful for campaign reporting, not for craft.
A sheet: clip title, source video and timestamp, clipper score, platform, posted at, and the five metrics after 7 days. Thirty rows is enough to see patterns.
Sort by clipper score, then by craft score. If the top-scored clips aren't your top performers, the tool's weights don't match your audience; change the prompt (Katto, OpusClip) or stop trusting the order and review more candidates.
Low hook rate → in-points and title cards. Low completion with good hook → length and endings. Low shares → topic and quotability. Low follows → attribution and a reason to visit. One fix per batch so you can see its effect.
The minutes around a winning moment in the source video usually contain two more. Go back to the source, not the clipper's list.
For campaigns, reconcile your log against the board's confirmed views; disputes are won with screenshots and timestamps. See get paid to clip videos.
Shorts and Reels keep distributing for days. Compare clips at the same age, ideally 7 days.
Avg % watched favours short clips. Compare within a length band, or use completion with the length noted.
If one clip is 70% of the batch's views, the batch average means nothing. The analyzer flags concentration for this reason.
Boards count confirmed views at source after a hold and exclude flagged accounts. Expect the paid number to be lower.
Over 100% avg watched on a 10-second clip is loops, not love. Weight completion and shares on very short clips.
Rage-bait maximises shares and tanks follows. Decide which ratio pays before you tune for it.
Those with schedulers (OpusClip, Klap, Katto, Vizard, Quso) show views and engagement on the posts they published, which is handy for a cross-platform roll-up. None expose the retention curve; for hook rate and completion you need TikTok Studio, YouTube Studio or Instagram Insights.
Above 70% still watching at 3 seconds is good, above 80% is excellent, under 60% means the opening needs work. Numbers vary by niche; compare against your own median rather than a universal benchmark.
About 20–30 at the same age on the same platform. Below that, one outlier dominates. The analyzer flags when the top clip is more than half the batch's views.
Partly. YouTube's Analytics API and TikTok's Research/Display APIs return per-video metrics for your own content; Instagram's Graph API returns Reel insights for professional accounts. Pull weekly into a sheet, compute the ratios, and feed the top topics into your clipper's prompt. See API and MCP for the clipping side.
Confirmed views at the platform after a hold period, sometimes with a per-clip cap and minimum. Shares and follows don't pay directly, but shares drive the views that do. Use the CPM calculator to turn a board's rate into a per-clip figure.