The Video Ad Metrics That Actually Predict Winners

You launch six video ads into a $50/day test. Two days in, none has the spend or conversions to call a winner with a straight face. You either wait a week and burn budget on the four that were never going to work, or you read the early signals and reallocate by tomorrow morning. The operators who scale do the second thing.

The catch is that most of the numbers on the dashboard are useless this early. Impressions, total views, and likes all look fine on an ad that will never convert. The metrics that actually forecast a winner move in the first 48 hours, at low spend, in a strict order: scroll-stop first, retention second, conversion third. Read them in that order and you kill losers before they cost you, and double down on winners before the auction catches up and CPMs rise.

The three metrics that predict winners, in order

A video ad fails at one of three gates, and the gates are sequential. If nobody stops scrolling, nothing else matters. If they stop but don't stay, your message never lands. If they stay but don't click, the offer or the audience is wrong. Diagnose them in that sequence — fixing gate two while gate one is broken is wasted effort, because viewers never reach the part you're tweaking.

Gate 1 — Hook rate (does it stop the scroll?)

Hook rate is the share of impressions that turn into a counted early view. On Meta the formula is 3-second video plays divided by impressions, times 100. It's the same arithmetic as the thumb-stop ratio — different name, identical calculation — and it's the first thing the algorithm reads about your creative. Roughly 90% of viewers bounce if the opening doesn't grab them in the first three seconds, so a weak hook rate caps everything downstream.

It also matters thresholds differ by platform. TikTok counts a hook at the 2-second mark, Meta at 3 seconds, so never compare a TikTok hook rate against a Meta one directly. If your opening three seconds are weak, that's where to spend your effort first; a sharper set of scroll-stopping openers moves this number more than anything else you can do.

Gate 2 — Hold rate (do they stay for the message?)

Hold rate is retention given attention: ThruPlays divided by 3-second video plays, times 100. Where hook rate measures the scroll-stop, hold rate measures whether the body of the ad keeps the people the hook already caught. A high hook with a collapsing hold means the opening over-promised or the pacing sagged — the narrative, not the hook, is the problem.

This is the gate where the script earns its keep. Most ads die here because the first three seconds were engineered in isolation and the next ten were an afterthought. Fixing it is a structural script problem, not a thumbnail swap.

Gate 3 — Cost per outbound click (will it pay?)

The first conversion-side signal that gets reliable at low spend is cost per outbound click — what you pay for a click that actually leaves the platform for your site, as opposed to a click that expands the post. It gets statistically meaningful far sooner than CPA or ROAS, which need conversion volume you won't have on day two. Cost per outbound click tells you whether the attention you bought is the kind that moves toward a purchase, before you've spent enough to measure the purchase itself.

The benchmarks: what "good" actually looks like

Benchmarks are guardrails, not goals — but flying blind is worse. Here's where the cited ranges land, by platform.

Meta hook rate and thumb-stop

Hook rate. A common target band is 20–25%, with top ads exceeding 30%. Motion frames strong performance as 30–40% baseline, with below 25% signaling a creative problem; Billo's tiers put best-in-class at 30%+ and elite at 35–45%.
Thumb-stop ratio by placement. Because it's the same formula, the placement-level data is the most useful slicing. Facebook Feed cold prospecting runs 18–28% (median ~23%), Reels 24–36% (median ~30%), Stories 22–32%, and warm retargeting 30–45%. Reels demand 5–10 points higher than Feed because the swipe is faster.
Below ~20–25% is the danger line. AdSights notes that TSR below 25% means most of your budget is funding unseen impressions — the auction never gives a creative real reach if the thumb-stop is invisible.

Vertical matters too. Gaming creative can hit 38–52% on Meta Reels while SaaS/B2B has a structural ceiling around 20–26%. Don't hold a B2B demo to a gaming benchmark.

Meta hold rate and ThruPlay

Hold rate. Average is 40–50%, with top performers above 50%; Motion calls above 60% strong and under 30% in need of work. An ad clearing 35% hook plus 30% hold is demonstrably performing.
ThruPlay rate is length-dependent — compare like with like. A ThruPlay is a complete play for videos under 15 seconds, or reaching 15 seconds on longer ones. So under-15s benchmark at 70–95% (median ~85%), 15–30s DTC prospecting at 15–30%, and 30–60s at 8–18%. Comparing a 12-second ad's ThruPlay to a 40-second ad's is measuring the threshold, not the creative.

TikTok: the 2-second hook and the 6-second VTR

TikTok's algorithm uses two thresholds. A 3-second view rate above 40% tells the algorithm the hook landed and it expands distribution; below 40% restricts delivery. Then the 6-second VTR splits at 45% for sustained engagement, 25–35% as industry average, and below 25% gets deprioritized. TikTok's own docs define a 6-second focused view as a play of at least 6 seconds, a full play if shorter, or an engagement within the first 6 seconds.

The most important TikTok insight for anyone over-indexed on CTR: a campaign with 0.7% CTR and 45%+ 6-second VTR outperforms one with 1.1% CTR and 22% VTR in algorithmic efficiency. TikTok's average CTR sits at just 0.84%, so chasing clicks while ignoring retention is optimizing the wrong gate. For the format-specific mechanics, the TikTok specs that actually convert are a better starting point than CTR.

YouTube: view rate, not raw views

YouTube's equivalent is view rate — views divided by impressions, where a view counts at 30 seconds or video end, or on a click. The overall average is 31.9%, with desktop at 35.4% and mobile at 33.2%; skippable in-stream specifically baselines at 15–25%, with strong campaigns at 30–40%. Completion drops with length: 15-second ads complete at 70–85%, 30-second at 40–60%, 60+ at 20–35%. Connected TV is a different animal — 90–97% completion, with non-skippable pre-rolls at 98.6% — so never benchmark a skippable in-feed ad against CTV completion. The Shorts playbook applies different rules again.

The vanity metrics to deprioritize (and why they lie)

Each of these can look healthy on an ad that's quietly failing. That's exactly what makes them dangerous — they offer false comfort and absorb attention you should be spending on the three gates.

Raw impressions. Impressions measure how much budget you spent, not whether anyone engaged. A high impression count with a sub-20% thumb-stop means the platform served your ad and viewers scrolled straight past — you paid for invisible reach.
Total video views with no retention context. "10,000 views" is meaningless without knowing where people dropped. As Motion puts it, impressions, likes, and views alone are misleading; the real signal combines attention, engagement, and conversion. A view that ended at 3.1 seconds tells you nothing about whether your message landed.
CTR in isolation. The TikTok example above is the proof — a higher CTR can hide worse algorithmic efficiency. CTR matters as part of the pattern (alongside VTR and cost per outbound click), never as a standalone scoreboard. For context, traffic-campaign CTR on Meta averages 1.71%, but a high CTR with poor hold rate is just an ad that tricks people into clicking and loses them on the page.
Likes, comments, shares. Engagement vanity. Useful as a tie-breaker between two converting ads, useless as a primary signal.

The rule: any metric the algorithm doesn't use to decide distribution, and that doesn't sit on the path to revenue, is a tie-breaker at best. Hook rate, hold rate, VTR, and cost per outbound click all gate distribution or forecast revenue. Likes and raw impressions do neither.

The diagnostic framework: a 2x2 you can act on in 48 hours

The single most useful tool here is a hook-rate-by-hold-rate matrix. Pull both for each ad at the 48-hour mark and place it in one quadrant. The diagnosis is unambiguous, and so is the fix.

High hook, high hold — The creative works. Move straight to cost per outbound click. If that's also strong, this is your scale candidate; if the click cost is high despite great retention, the leak is the offer, the landing page, or the audience, not the video.
High hook, low hold — The opening stops the scroll but the body loses them. Keep the first three seconds, rewrite the middle. This is the most common fixable quadrant — your hook is an asset, your script is the liability.
Low hook, high hold — The few who stay love it, but almost nobody starts. The body is good; the opening is wrong. Swap the hook and you may have a winner hiding under a bad first frame.
Low hook, low hold — Kill it. Don't iterate. Vaizle's guidance is explicit: fix hook rate before hold rate, because a low hook prevents viewers from ever seeing the message you'd be trying to improve.

The sequencing is the whole point. Spending an afternoon polishing the script on a low-hook ad is fixing a room nobody walks into. Hook first, always. This is also why structured creative testing isolates one variable at a time — if you change the hook and the script together, the matrix can't tell you which one moved the number.

A read-the-numbers checklist for a fresh test

Run this on any new batch, in this order. It's deliberately mechanical so you don't rationalize keeping a loser or panic-killing a slow starter.

Day 1–2: check spend, not performance. Has each ad cleared enough impressions for hook rate to stabilize? Below that, the numbers are noise. Don't judge anything yet.
Gate 1 — hook rate / thumb-stop. Sort the batch by it. Anything under the placement floor (roughly ~20% for Feed, ~25% for Reels) goes on the kill list unless hold rate is exceptional.
Gate 2 — hold rate, within the same length bracket. Survivors of gate 1 get ranked on retention. Remember to compare ThruPlay only within the same length bracket, or the threshold mechanic will mislead you.
Gate 3 — cost per outbound click. Of the ads that stop and hold, which buy the cheapest qualified clicks? That's your conversion-side forecast before CPA is reliable.
Decide: scale, fix, or kill. Scale the high-hook/high-hold/cheap-click ads. Fix the ones with a clear single-gate failure. Kill the rest and free the budget.
Only now glance at CPA/ROAS — for confirmation, not decisions. Meta ROAS commonly lands 2.5–4.0 depending on customer value; treat it as the autopsy that confirms what the leading metrics already told you.

The discipline is the same everywhere: act on the leading gate, confirm with the lagging one. A slow starter that hasn't cleared enough impressions isn't a loser yet — it's unjudged.

Why reading metrics fast is worthless if you can't produce fast

Here's the operator math that makes all of this matter. The framework above tells you, on day two, that ad #3 has a great hook and a dead middle — fixable — and ad #5 needs a new opening. Now you have to actually make those variants.

Done by hand, that's the bottleneck. A new hook plus a re-cut middle is a brief to an editor, a day or two of turnaround, and a revision cycle — or a freelancer retainer, or an agency line item. By the time the fix lands, the test window has closed and you're reading stale data. So most solo founders quietly stop iterating: they run the two ads they could afford to make, let the numbers decide which is less bad, and call it a strategy.

That's the trap. Leading metrics let you act early — but acting means shipping a new variant, and if that costs two days you can't act on a 48-hour signal. Diagnosis without throughput is a better-informed wait.

When production drops to minutes instead of days, the math inverts. Instead of 2 hooks you test 20. Instead of one creative per campaign you ship a fresh one per ad set. A founder running three products can keep each account fed; a two-person shop can take on a dozen clients without hiring, because the constraint was never strategy — it was the editor's queue. The leverage isn't "cheaper ads," it's making the diagnostic loop fast enough that reading the metrics changes what you do tomorrow. That's the core idea behind treating iteration speed as a moat, and it's what lets a solo founder run a creative volume that used to need a team. For agencies, the same throughput is the difference between 3x-ing client load and burning out the one editor.

FAQ

What is a good hook rate for video ads?

On Meta, a common target is 20–25%, with top ads above 30%, though it varies by placement — Feed cold prospecting runs ~18–28% and Reels ~24–36%. On TikTok the threshold is 2 seconds and the algorithm expands distribution above a 40% 3-second view rate. Always benchmark against your own placement and vertical, not a single universal number.

What's the difference between hook rate and hold rate?

Hook rate measures the scroll-stop: 3-second plays divided by impressions. Hold rate measures retention among the people who stopped: ThruPlays divided by 3-second plays. Hook tells you if the opening works; hold tells you if the body keeps them. A high hook with a low hold means a strong opening and a weak script — fix the middle, keep the first three seconds.

Which metrics should I ignore when testing video ads?

Deprioritize raw impressions, total views without retention context, likes, and CTR in isolation. As Motion notes, impressions, likes, and views alone are misleading — they can look healthy on a non-converting ad. A TikTok campaign with a lower CTR but higher 6-second VTR can outperform a higher-CTR one, which is why CTR is only useful inside the pattern.

How long before I can trust video ad metrics?

Hook rate and thumb-stop stabilize first because they're impression-based, so they're usable within a day or two once an ad clears enough impressions. Hold rate follows. Cost per outbound click is the earliest reliable conversion-side signal. CPA and ROAS are lagging — they need conversion volume you rarely have early, so use them to confirm, not to decide.

Why does my ad have high CTR but bad sales?

A high CTR with weak hold rate usually means the ad pulls clicks it can't back up — the hook over-promises and the body or the landing page doesn't deliver. Check hold rate and cost per outbound click together: if hold is low, the creative loses people before the message lands; if hold is fine but sales aren't, the leak is downstream on the page, and an ad-to-landing-page mismatch is the usual culprit.

Sources

Reading the metrics fast only pays if you can produce fast — which is the gap Aitachyon is built to close. Describe what you're selling in one prompt, or paste your site URL, and it returns a finished, captioned MP4 in about two minutes, with three script variants out of the gate and exports in 9:16, 16:9, or 1:1 for TikTok, Reels, Shorts, Meta, and LinkedIn — so when the 48-hour data says "great hook, dead middle," you ship the fix the same morning instead of waiting on an editor. Plans run $29 to $299 a month with a 14-day money-back guarantee. Start a test batch and let the leading metrics actually change what you do tomorrow.