No, and on Instagram the relationship runs the wrong way. Across 954 clips posted through ClipEmber between July and October 2026, the rank correlation between our pre-publish score and views 72 hours later was −0.06 (95% CI −0.13 to 0.01). On Instagram Reels it was −0.14 (−0.24 to −0.04): clips in the top score quartile got a median of 19 views, the other three quartiles 56.
Every clip ClipEmber renders gets a score from 0 to 100 before it is posted. We took every clip that was published through ClipEmber and had a view count 72 hours after posting, and compared the score with those views. 72 hours is a fixed point for every post, so old and new posts are compared on equal terms.
We used rank correlation (Spearman), because view counts are extremely skewed: a few posts get thousands and most get very few. The 95% intervals come from 2,000 bootstrap resamples. About 88% of the posts in this period came from our own accounts; the rest came from 26 other accounts.
| Measured | Source | |
|---|---|---|
| Clips with views 72 hours after posting | 954 | scheduled_posts + post_metrics (72 h snapshot), posted 19 July–20 October 2026 |
| Score vs views, all platforms | −0.06 (95% CI −0.13 to 0.01) | Spearman rank correlation, bootstrap CI, 2,000 resamples |
| Instagram Reels | −0.14 (−0.24 to −0.04), n = 387 | same method |
| YouTube Shorts | −0.07 (−0.16 to 0.02), n = 488 | same method |
| TikTok | +0.12 (−0.09 to 0.32), n = 79 | same method |
| Same check at 24 hours, all platforms | −0.03 (−0.09 to 0.03), n = 1,022 | post_metrics 24 h snapshot, same period |
Instagram is the one platform where the interval sits entirely below zero. Splitting its 387 Reels into four groups by score shows where that comes from: the three lower groups had the same median views, and the highest-scored group had a third of that.
One likely reason, from our earlier analysis of the same posts: the scorer we used until October read a short stretch of transcript and favored talkative moments, while on our accounts talking-head clips drew far fewer views than clips that show something being made or demonstrated.
| Measured | Source | |
|---|---|---|
| Score below 69 | 56 views median (n = 89) | Instagram Reels, views at 72 hours |
| Score 69–73 | 56 views median (n = 85) | Instagram Reels, views at 72 hours |
| Score 74–78 | 56 views median (n = 99) | Instagram Reels, views at 72 hours |
| Score 79 and above | 19 views median (n = 114) | Instagram Reels, views at 72 hours |
On 8 October 2026 we switched to a scorer that also watches short previews of each candidate instead of reading only text. In an offline check on 160 held-out clips it ranked them closer to their real views than the old one did. On live posts it is too early to say: 80 clips have reached 72 hours so far, and the interval is wide enough to include both a real improvement and none.
| Measured | Source | |
|---|---|---|
| Old text-only scorer, live posts | −0.06 (−0.13 to 0.01), n = 867 | clip_decisions.scorer = gemini, views at 72 hours |
| New scorer that watches the video, live posts | −0.11 (−0.31 to 0.13), n = 80 | clip_decisions.scorer = gemini-rubric+video, views at 72 hours |
How long people kept watching. For 354 Instagram Reels we read Instagram's own average watch time: it ranked the clips almost the same way views did. The median Reel was watched for under 5 seconds whatever its length, which means the first seconds decide most of what happens next.
That isn't a score you can know before posting, but it tells you where to look when you choose a clip: whether it opens on something that holds attention in the first two or three seconds.
| Measured | Source | |
|---|---|---|
| Average watch time vs views | 0.74 (Spearman), n = 354 | Instagram insights ig_reels_avg_watch_time, read 8 October 2026 |
| Median average watch time | 4.8 s | same 354 Reels |
Most of these posts come from our own accounts, which clip other people's videos. Platforms limit how far reused clips travel: our YouTube Shorts had a median of 1 view and our TikTok posts sat in a narrow 260–270 range, so on those two platforms there is little variation for any score to explain. The same clip can also appear on more than one platform.
So this is not evidence that clip scoring can never work, and creators clipping their own videos may see different numbers. It is evidence that our score, as built until October, did not rank clips the way the platforms did, and that you shouldn't treat any tool's virality number as a forecast without seeing data like this.
It ranks how hook-like a moment is inside the window you gave it: whether a clip opens on something that stands alone. That is useful for picking a starting point from twenty candidates.
It is not a prediction of distribution, and we have a documented example of the gap. On a cooking video, the cook's actual opening line — "Check it out, folks. If you like a chopped cheese, you're gonna love this" — scored 68, last of five. A mid-video aside about breaking up the meat scored 82 and came first.
| Measured | Source | |
|---|---|---|
| Real opening hook, cooking run | scored 68 of 5 clips (last) | request 2ced1dde, clip scores 82/80/76/68/68 |
| Mid-video filler, same run | scored 82 (first) | same run |
| First version of this page | 45 posts, Pearson −0.151 | published 20 August 2026, latest snapshot per post |
One row per clip published from ClipEmber's own accounts: platform, date, clip length in seconds, pre-publish score, scorer version, views at 24 and 72 hours, likes at 72 hours. Rows without a view count are posts whose snapshot hadn't been taken. Other users' posts are not included.
Clip scores and views, ClipEmber's own accounts (CSV) · 1,097 rows · CC BY 4.0
No. A correlation of −0.06 means the score tells you almost nothing about views either way. Use it to shortlist, then pick by how the clip opens and what it's about.
Because the alternative is letting you plan around a promise we can't support. We would rather be precise about what the score does and doesn't do, and publish the data so you can check it.
We haven't measured other tools, so we can't say. We also haven't found a vendor that publishes how its score compares with real views; if one does, compare it with data like this.
Yes. We'll update this page when the new scorer has enough posts to judge. The numbers above are from 20 October 2026, and the first version from August is listed above for comparison.
5 free videos every month, no card required.