BenchmarksHooksCheatsheetAnalytics

Cold Traffic Hook Benchmarks Cheatsheet: 30%+ vs Under 20%

Cold-only hook-rate bands by platform and opener type, first-frame diagnostics, sample-size floors, warm-bleed checks and recut rules — good is 30%+, weak is under 20%.

Updated 2026-08-2411 min read

LanguageEnglish

On cold traffic a hook either stops the thumb or it does not. This sheet is the cold-only bands we use: 30%+ is good, under 20% is weak, plus opener-type ranges, first-frame fail lists, warm-bleed checks and the impression floors before you kill a file.

Cold hook rate is 3-second views ÷ impressions, nothing else

Do not diagnose a cold ad with average watch time, ThruPlay, or a hold rate you have not earned yet. Hook rate is 3-second video views divided by impressions on an audience that has not already chosen you. Good on Reels, TikTok and Shorts is 30% and up. Weak is under 20%. The 20–29% band is a near-miss: iterate the first frame and first line, do not torch the body. YouTube in-stream is a different clock — judge the 5-second skip point, not a 3-second social hook. If your report is mixing 4:5 feed with 9:16 Reels, split the placement before you split the creative. A '24% hook' on blended inventory is not a number you can act on.

Opener type moves the band more than polish does

A talking-head problem-callout, a pattern-interrupt object drop, a split-screen proof frame and a logo-then-talk open are not the same test. Cold audiences reward a first frame that names a problem or breaks a pattern; they punish brand cards, slow smiles and questions that dump the category ('need better sleep?'). The locked opener table is observed ranges on cold UGC, not a promise. If your vertical sits a few points lower (SaaS, info), still use 30% as the scale floor on Reels/TikTok rather than moving the goalposts because the file was expensive. Delivery cannot rescue an opener whose first frame is a black slate.

Most 'cold' hook rates are warm bleed

Site visitors, engagers and a lookalike that is secretly your buyers will print hook rates that look elite while prospecting is still dead. Before you scale a 38% hook, check audience overlap and frequency. If hold is also suspiciously fat on a brand-new file, assume mix until proven otherwise. Exclude purchasers and 7-day viewers from the cold cell. Read hook only after the impression floor. When the diagnosis is truly a dead opener, recut 0–3 seconds — generate new first lines from the same product URL in Klip Kanvas rather than reshooting a body that never got a chance to play.

How to use the tables below

Confirm the ad set is actually cold — purchasers and recent video viewers excluded — then pick the platform row, then the opener-type row, then the first-frame fail list. Do not average Reels with Feed. Do not scale under 30% at a confident sample (5k–8k impressions). Under 20% at the early-kill floor: kill the opener and recut 0–3 seconds only. If hook is 30%+ and CTR is dead, stop editing the open; that leak is offer, CTA or page. Locked values are bands and rules, not essays. Warm-looking hook plus fat hold on a brand-new file is bleed until you prove otherwise. One variable per next batch: three hooks, two faces, same body.

5 lookup tables inside: cold hook-rate bands by platform (good 30%+ / weak <20%), opener-type ranges, first-frame fail list, impression floors plus warm-bleed checks, and recut/kill/scale rules for cold cells only.

Cold hook rate by platform (3s views ÷ impressions)

House rule (Reels / TikTok / Shorts): Good 30%+ · Near-miss 20–29% · Weak / kill opener <20%
Meta Reels / IG Reels ads: Kill <20% · Average 20–28% · Good 30–40% · Elite 40%+
TikTok In-Feed / Spark (cold): Kill <22% · Average 22–30% · Good 32–42% · Elite 42%+
Facebook / Instagram Feed 4:5 or 1:1: Kill <18% · Average 18–26% · Good 28–36% · Elite 36%+
Stories (IG / FB): Kill <18% · Average 18–25% · Good 27–35% · Elite 35%+
YouTube Shorts: Kill <20% · Average 20–28% · Good 30–38% · Elite 38%+
YouTube in-stream skippable: Do not use 3s hook · Weak <35% still watching at 5s · Good 45%+ at skip point
Scale permission: Do not scale a cold UGC file under 30% hook at confident sample on Reels/TikTok/Shorts

Cold opener types (observed Reels/TikTok UGC)

Problem-callout (spoken, first frame is face): Typical 28–38% · needs specific customer language · vague 'tired of X?' often <22%
Pattern-interrupt (object drop / freeze / whip-in): Typical 32–42% · first frame cannot be a smile-to-camera hold
Proof-first (split-screen or number on frame 1): Typical 30–40% · number must be substantiated · fake stats do not 'test'
Myth-bust / contradiction: Typical 29–38% · the turn must land by 0:04 or it becomes a lecture
UGC confession / 'I was today years old': Typical 30–39% · dies if the confession is a slogan
Logo / brand-card open: Typical 12–22% · treat as fail on cold · never a control
Category dump question: "Need better [category]?" · typical <20% · rewrite as a specific symptom
Direct offer / discount lead on cold: Typical 22–30% · often loses to problem/interrupt · save hard offer for warm

First-frame fail list (0–3s)

Black frame / fade-up / logo sting: Fail · first decoded frame must be a readable still
Talking without a new fact in sentence 1: Fail · "hey guys" / "so I found this" · rewrite
Text in top 14% or bottom 20% (9:16): Fail · UI cover · move hook text to the safe block
Muted-unreadable: Fail if the hook idea is only in audio · overlay ≤5 words on frame 1
Face too small / room tour: Fail · MCU or product fill by 0:00.5
Slow zoom on packaging: Near-fail · cut to the problem or the interrupt by 0:02
Pass: Frame 1 readable as a thumbnail · line 1 is specific · motion or overlay by 0:01

Sample floors and warm-bleed checks

Hook directional: 2,000–3,000 impressions · no scale · no 'winner' label
Hook confident: 5,000–8,000 impressions · scale/kill allowed
Early kill — dead opener: Hook <15–20% at 3,000–5,000 · kill opener regardless of CPA
Too early: Under floor · status = too early · do not iterate on the percentage
Exclude from cold cell: Purchasers · 7-day ATC · 7-day 50%+ video viewers · engagers if they dominate delivery
Bleed suspicion: Hook 38%+ and hold looks warm on a new file · check overlap · frequency >2 early is a clue
Placement split: Do not average Reels + Feed + Audience Network into one hook % · break out before you recut

Recut / kill / scale (cold cells only)

Hook <20%: Kill opener · recut 0–3s only · keep body and CTA
Hook 20–29%: Near-miss · new first frame + first line · same body · 1–2 more tries then park
Hook 30%+, rest of funnel unknown: Watch problem solved · read outbound CTR then CPA · stop editing the open
Hook 40%+: Scale candidate on hook · still require conversion action before budget jumps
Hook 30%+ but CTR dead: Not a hook problem · offer/CTA/page · do not 'fix' with a new 0–3s
Next batch: 3 hooks × 2 faces on the winning body · one variable · same cold exclusions

Locked

Unlock the full cheatsheet

Enter your email and we'll instantly unlock the complete cheatsheet — plus send you fresh ad playbooks every month.

No spam. Unsubscribe anytime.

Ready to put this into practice?

Create your first AI UGC video ad in minutes — no filming, no actors, no editing.

Try Klip Kanvas free