How to know if your video will perform before you post

The standard creator workflow is broken on a structural level. You shoot, edit, post, and then check analytics three days later. By the time you know whether a video performed, the algorithm has already made its call and moved on. The next video is either built on a survivor (the one that worked) or on guesses about why the others didn't. Both are bad inputs.

The fix isn't checking analytics faster. It's checking before posting at all.

This post covers the variables that actually predict short-form video performance, how to score them manually, and what changes when you can score them automatically.

Why post-publish analytics are the wrong feedback loop

Native analytics on Instagram, TikTok, and YouTube Shorts measure outcomes. They tell you what happened. They don't tell you why.

A Reel that got 200 views might have had a weak hook, a strong hook delivered to the wrong audience, a pacing problem at 0:08, a CTA that confused the viewer, or just bad timing in the algorithm cycle. Analytics show drop-off curves, but the curves are the symptom, not the cause. You're left to guess which variable to change for the next video.

Worse, the feedback is heavily survivorship-biased. The video that hits 50K views gets your attention and shapes your next attempt. The 200-view video gets ignored, even though it usually contains the more useful diagnostic. The pattern most creators end up with is: copy what worked once, hope it works twice, eventually plateau.

Pre-publish analysis inverts this. You score the video against patterns that work BEFORE the algorithm sees it. The variables that drive performance are mostly knowable in advance. They're not random.

The variables that actually predict performance

Five variables explain most of the variance in short-form video performance. Other things matter, but these are the load-bearing ones.

1. Hook strength in the first 3 seconds

The opening 3 seconds determine 60-70% of total retention on every short-form platform. The reason is mechanical: the algorithm uses early drop-off as a quality signal. Videos that lose viewers in seconds 1-3 get throttled in distribution. Videos that hold past second 3 get pushed to more viewers, which is the only way views compound.

A strong hook does one of four things: it creates an information gap the viewer wants resolved, it names a specific identity or situation that activates self-relevance, it interrupts a pattern the viewer expected, or it makes a specific value contract about what the next 30 seconds will deliver.

Generic openings ("hey everyone") fail all four tests. Most underperforming creator videos fail because the hook does no work, not because the body of the video is weak.

2. Pacing and cuts per second

Short-form platforms reward pacing that matches the platform's average. TikTok averages roughly 1 cut every 2-3 seconds for top-performing content. Reels run slightly slower. Shorts are slower still, partly because they pull viewers from search and suggested video, where intent is higher and tolerance for slow setups is greater.

Pacing isn't the same as fast cutting. It's the rate of new information per second. A video can have one camera cut and still have high pacing if the speaker delivers a new claim, a new visual, or a new turn every few seconds. A video with 30 cuts and one repeated point has bad pacing.

Drop-off curves cluster around pacing failures. Long pauses, redundant sentences, or visual stasis at second 8-15 are the most common reasons retention craters in the middle of the video.

3. Audio and visual coherence

Most analyzers treat video as transcript. They miss the half of the video the viewer actually responds to. Music drops, tone shifts, eye contact, gesture timing, and on-screen text all carry signal that pure text analysis misses.

A hook that reads strong on paper can fail because the speaker's tone is flat, the music doesn't lift, or the visual cuts don't reinforce the spoken claim. Conversely, a hook that reads generic can hit because the delivery does the lift the words don't.

This is where most pre-publish scoring goes wrong. If you score only the transcript, you score half the video. The audio-visual layer needs to be in the model.

4. CTA clarity and placement

The CTA is where most short-form videos quietly fail. Either the CTA is missing entirely, or it's so soft the viewer doesn't recognize it as one ("hopefully that was helpful"). Both kill the second-order metric the algorithm cares about, which is whether the viewer takes a follow-up action.

Effective CTAs in short-form do three things: they're specific (follow vs follow for X), they're placed where retention is highest (last 3 seconds or as a mid-video pin), and they connect to the value the video just delivered (continuity, not a non-sequitur ask).

A CTA that lives at second 25 of a 30-second video, asks the viewer to follow without saying for what, and lands after the value has already been delivered is the most common pattern in underperforming videos. It's also the easiest one to fix, which is why CTA scoring is one of the highest-leverage pre-publish checks.

5. Pattern match to proven formats in your niche

Every niche has a small set of structural patterns that work. Self-help on TikTok runs different patterns than e-commerce on Reels. Patterns aren't templates, but they constrain what works.

A video built on a pattern proven in your niche has a much higher prior probability of performing than a video built on a pattern that works in a different niche or doesn't appear in your niche at all. This isn't about copying. It's about not fighting structural gravity.

Most creators have an intuitive sense of this for a few patterns they've seen work. The full pattern library for any niche is much bigger than what any one creator has noticed. Pre-publish scoring against the full library catches the cases where you've drifted into a structure that isn't paying out in your specific lane.

The DIY framework: how to score these manually

Before you reach for a tool, you can run a manual version of pre-publish scoring on every video. It takes about 5 minutes per video. Worth it.

Hook check. Mute the audio. Watch the first 3 seconds. Did anything specific happen visually that would stop the scroll? Now play the audio. Did the first sentence create a gap, name an identity, interrupt a pattern, or promise a specific value? If neither layer did work, your hook is weak. Re-shoot the opening.

Pacing check. Count the new beats. A beat is a new claim, a new visual, a new turn. For TikTok, you want one every 2-3 seconds. For Reels, every 3-4. For Shorts, every 4-5. Count them. If you're below the rate, the body is too slow.

Coherence check. Watch the video twice: once with audio only, once with video only. Does the audio version make sense? Does the video version carry the message? If either layer is doing all the work and the other is dead, the video is weaker than it could be.

CTA check. Watch the last 5 seconds. What action are you asking the viewer to take? Is it specific? Does it follow from the value just delivered? Does it land before the video ends, not after? If any answer is no, the CTA needs work.

Pattern check. Find 5 high-performing videos in your niche from the last 60 days. What structures do they share? Does your video follow one of those structures, or have you drifted into something none of them use? If you've drifted, ask why.

This manual version catches most of the easy fixes. The variables it checks are the same ones an automated scorer would check, just slower and more subjective.

What changes with automated pre-publish scoring

The manual version has limits. It depends on your honest assessment of your own work, which is the assessment everyone is worst at. It also can't run pattern matches against thousands of videos in your niche fast enough to catch subtle drifts.

Automated pre-publish scoring removes both. The hook gets scored against a bank of proven patterns, not against your own intuition. Pacing gets measured against platform-specific norms, not your gut sense. CTA placement gets compared against the placement that correlates with conversion in your category. Pattern match runs across the whole library, not just the 5 videos you remembered.

The output is a per-segment breakdown with scores, identified weak points, and specific fixes. The fixes are what matter. "Your hook is weak" is useless feedback. "Viewers drop at second 8 because the second beat repeats the first; cut the second beat or replace it with the data point you mentioned at second 22" is actionable.

Lomero is built around exactly this loop. Paste a draft, get the scores, get the specific fixes, then re-shoot or re-cut before posting. The free tier covers 5 full analyses per month, which is enough to score every meaningful draft for a creator posting 3-4 times per week.

The asymmetry that makes pre-publish worth it

Most creators spend hours per video on production and minutes on review. The math is wrong.

A 10-minute pre-publish review on a video that took 4 hours to make is a 4% time investment that catches most of the variance. If 1 in 5 of your videos is killing your channel by training the algorithm on weak signal, fixing that one before it ships is worth more than producing the next two.

The compound effect is bigger than the per-video effect. Algorithm distribution is reputation-weighted. Channels that ship consistently strong videos get pushed harder than channels that ship a mix of strong and weak. Filtering out the weak ones at the pre-publish stage tightens your average and accelerates distribution over months, not just for the individual video.

The right question isn't "did this video work." It's "would I post this video if I knew it was going to flop." Pre-publish scoring lets you answer that before the algorithm does.

---

Related: the anatomy of viral short-form video covers the structural patterns referenced above in more detail. Hook patterns that stop the scroll is the working library of opening structures.