Fifteen tips on hook rate and hold rate: exact definitions, realistic benchmark ranges, the 2x2 diagnostic, retention-curve shapes and which number to fix first.
Hook rate and hold rate are the two numbers that tell you whether a video ad failed at the door or failed in the room. Most people quote them loosely, compare them across incompatible placements, and end up fixing the wrong half of the ad. Here is how to define, benchmark and act on both.
Not by reach, not by clicks. Hook rate = 3-second video plays ÷ impressions, expressed as a percentage. It measures one thing: did the first frames earn a pause. Write the formula into your reporting template so everyone on the team is quoting the same number — most disagreements about whether a hook is working turn out to be two people using two different denominators.
Hold rate = ThruPlays (or 15-second views) ÷ 3-second views. The denominator matters enormously: dividing by impressions blends the hook problem into the body problem and gives you one muddy number that diagnoses nothing. By dividing by 3-second views you are asking a clean question — of the people who gave this ad a chance, how many stayed.
A Reels placement autoplays full-screen with sound; a right-column or audience-network placement does not. The same creative can show 30% hook rate in one and 6% in another with no creative difference at all. Always break down by placement before comparing two ads, and if you only look at one number, look at the feed placement that carries most of your spend.
Across the ecommerce accounts we see, vertical feed placements typically land at 20–35% hook rate and 8–15% hold rate, with strong creatives pushing above that. Treat those as orientation only — category, price point and audience temperature move them a lot. The number that actually matters is your own account's trailing 90-day median, because that is what a new creative has to beat.
Split at your account median for each metric and you get four quadrants: high/high (scale it), high hook / low hold (the hook oversold), low hook / high hold (a good ad nobody opens), low/low (kill it). This single chart replaces an hour of scrolling through rows, and it tells you the specific repair each ad needs rather than just ranking them.
When the hook rate is strong and hold collapses, the opening wrote a cheque the body could not cash — usually a curiosity gap or a shock frame that has nothing to do with the product. The fix is not a better hook. Rewrite the first two lines after the hook so they immediately deliver on what was promised, and move the payoff earlier. This quadrant is the most common home for ads that look great in the platform's video metrics and sell nothing.
A high hold rate on low volume means the ad is genuinely good and almost nobody is seeing it. This is the cheapest fix in advertising: keep the entire video, replace only the first 2–3 seconds and the first frame. In Klip Kanvas you can regenerate a hook against a locked body, so five new openings on a proven body cost a fraction of five new videos — and one of them usually pulls the whole thing into the top quadrant.
When both numbers are mediocre, start at the hook anyway. Hook iterations are a few seconds of production each and you can run five in the time it takes to rework a script, and a rising hook rate immediately increases the sample size feeding your hold-rate read. Body rewrites are worth doing — just do them second, on the version that has already earned attention.
Two ads can share a 12% hold rate with completely different curves. A cliff in the first two seconds after the hook means the promise broke. A steady slope means the content is fine but slow. A late cliff at 70% usually means the CTA arrived and people who were never going to buy left, which is harmless. Screenshot the curve for every winner you keep — the shapes become a pattern library within a month.
Find the steepest single drop in the curve, note the second, and play the video from one second before it. Nine times out of ten there is a concrete cause: a slow line, a jump cut to unrelated B-roll, a claim that sounds like an ad, a shot where the product finally looks cheap. Fix that one second rather than re-cutting the whole ad, and the curve usually smooths without any other change.
The 25% checkpoint tests whether the hook and the first claim hold together. The 50% checkpoint tests whether the proof section is convincing. The 75% checkpoint tells you how many people are still around when the CTA lands, which sets the ceiling on your CTR. If 75% retention is under about a third of your 25% figure, the CTA is speaking to a nearly empty room and moving it earlier is worth testing.
A 45-second ad and a 15-second ad cannot be compared on hold rate directly — the longer one is being asked to hold three times as long. Either compare only within a length band, or convert to average watch time as a percentage of duration. Teams that skip this step systematically conclude that short ads are better, when what they have measured is that short ads are shorter.
Hook rate falls when the auction gets more expensive and delivery shifts to lower-quality inventory, even if the creative is untouched. If hook rate slid 20% and CPM rose 30% in the same week, you are looking at an auction event, not creative decay. The clean fatigue signal is hook rate declining while CPM stays flat and frequency climbs.
Hook rate moves days before CPA does, because it needs far fewer events to be readable. Chart it weekly for every creative above your spend threshold. A creative that has lost 25–30% of its opening hook rate over two weeks is fatiguing, and you have a week to prepare the replacement before the cost metrics react. This one habit is the difference between rotating on schedule and rotating in a panic.
Shock hooks, mismatched hooks and clickbait all produce excellent hook rates and terrible economics. In the accounts we look at, the correlation between hook rate and cost per purchase is weak once you get past the bottom quartile. Use hook rate to diagnose and to detect fatigue; use cost per purchase and hold rate together to decide what gets more budget.
Hook rate tells you whether people opened the door; hold rate tells you whether the room was worth staying in. Define both with fixed formulas, compare only within the same placement and length band, plot the 2×2 every week, and fix the cheaper half first. Do that consistently and creative decisions stop being arguments about taste.
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
Try Klip Kanvas freeThirteen tips for collapsing the gap between an idea and a live test — pre-built campaign shells, good-enough thresholds, parallel setup and killing perfectionism.
Practical hook-writing tricks for paid video: first-three-words rules, spoken vs written hook splits, hook stacking, negative framing and how to test them cleanly.
Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.
