The 3-second video retention rate measures the share of viewers still watching at the third second. Organically, it is calculated on views: the healthy zone runs from 60 to 80%. In advertising, the same term refers to the hook rate, 3-second views divided by impressions, where the Meta median sits around 20 to 30%. The two figures are not comparable, and TikTok complicates matters further by counting a view at 2 seconds, not 3. The same creative can show 72% and 24% on the same day without either figure being wrong. According to Médiamétrie’s L'Année Internet 2025, published in February 2026, 15-24 year-olds spend 61% of their browsing time on social networks and messaging apps, and 11-14 year-olds 63%.
3-second retention rate: which metric are we actually talking about?
Four different metrics go by this name in French, and only two are genuinely calculated at the third second. The word “retention” travels from organic to paid without anyone specifying the denominator, and that is where almost every disagreement between an advertiser and its agency about the quality of an opening comes from.
The scene replays every quarter in the reporting meetings we run. The social media manager opens TikTok Studio, reads 71% 3-second retention and concludes the creative is holding up. The media buyer opens Ads Manager the same day, on the same video, reads 24% and asks for the hook to be redone. Both are right. They simply aren’t reading the same fraction.
The four things hiding behind a single word
- The organic retention curve: viewers still watching at the 3rd second ÷ views. This is the descending curve in TikTok Studio or Reels insights. It starts at 100% by construction.
- The hook rate, also called the thumbstop rate or thumb-stop ratio: 3-second video views ÷ impressions, in Meta Ads Manager. The denominator includes deliveries that nobody actually watched.
- The TikTok 2-second view rate: TikTok Ads counts a view as soon as playback starts, and publishes separate “2-second views” and “6-second views” tiers, excluding replays. Applying a Meta threshold to a TikTok column means comparing two different clocks. In its glossary of video playback metrics, TikTok counts a 2-second view each time a video starts and plays for at least 2 seconds within a single impression session, excluding replays.
- The hold rate: by convention, ThruPlays (15 s or completion) ÷ 3 s views. Except that half the market calculates it on impressions, which gives figures five to ten times lower. Both circulate under the same name.
Add the completion rate (VCR) and average watch time, and you have six figures that a poorly maintained agency report can present interchangeably under the label “retention”. The table below works as a conversion chart. Print it out; it settles 80% of misunderstandings.
| Metric | Formula | Platform threshold | Typical range | What it lets you decide |
|---|---|---|---|---|
| Organic 3 s retention, TikTok | Viewers at 3 s ÷ views | 3 s (Studio curve) | 60–80 % | Repost, adapt the format, cut the pre-roll |
| Organic 3 s retention, Reels | Viewers at 3 s ÷ views | 3 s | 55–75 % | Keep or drop an editorial angle |
| Hook rate / thumbstop, Meta | 3 s views ÷ impressions | 3 s | 14–30%, median ~21% | Whether or not to redo the opening |
| TikTok Ads 2 s view rate | 2 s views ÷ impressions | 2 s | 20–35%, median ~27% | Compare creatives with each other, never with Meta |
| TikTok Ads 6 s view | 6 s views ÷ impressions | 6 s | 6–15% (observed on our accounts) | Judge the hook-to-promise transition |
| Meta hold rate | ThruPlays ÷ 3 s views | 15 s or completion | 15–30% depending on length | Judge the body of the video, not the hook |
| Hold rate “on impressions” | ThruPlays ÷ impressions | 15 s or completion | 3–7%, median 4.6% | Nothing on its own: read it alongside the hook rate |
| Completion rate (VCR) | 100% views ÷ started views | End of the video | Depends on length, unusable raw | Decide on the length of the format |
The paid ranges come from Biddyco (an analysis published in late 2025 on a proprietary sample of Meta ads) and AdSights (July 2026 data for TikTok). The two rows marked “observed on our accounts” are orders of magnitude drawn from our own French e-commerce campaigns, not published data, so treat them as such.
3-second retention is only the first floor of a creative performance read. If you want the full chain, from hook to ROAS, it is detailed in the full four-stage measurement protocol. Here, we stay on the ground floor, but we go all the way down into it.
Why does your organic retention show 70% and your ad hook rate 25%?
Because paid puts every impression in the denominator, including those nobody really watched, whereas organic starts from the plays that actually began. The gap has nothing to do with the quality of the creative. It is arithmetic. Google says as much itself in its help page on viewability and Active View: depending on screen size and ad placement, some users never see an ad that may nevertheless be counted as an impression.
Take a real video, with rounded figures from a Reels campaign we ran for a facial skincare client in spring 2026:
- 100,000 impressions served
- around 37,000 plays that actually started and were visible long enough to count
- 26,000 3-second views recorded by Ads Manager
Hook rate: 26,000 ÷ 100,000 = 26%. 3-second retention in the organic sense: 26,000 ÷ 37,000 = 70%. Same video, same day, same audience. Two legitimate figures, 44 points apart, which lead to two opposite decisions if you mix them up.
The scroll-past, meaning the thumb that moves on before the image has even settled, is counted as an impression on the advertising side. Organically, the platform generally counts a view as soon as playback starts: the curve therefore starts mechanically at 100% and can only go down. For comparison, Google’s Display & Video 360 help centre considers a video ad viewable when at least 50% of the ad remains on screen for at least 2 consecutive seconds of playback.
What we don’t know, and nobody really does: the exact share of Meta impressions that never led to playback starting. Meta does not publish this breakdown. It is reconstructed by inference, not from data. Anyone who gives you this ratio to the decimal point is making it up.
The mental conversion rule
On our accounts, the ratio between the two readings is around 2.5 to 3. A paid hook rate of 25% roughly corresponds to organic retention of 65 to 75% on the same asset. A hook rate of 12% corresponds to organic retention that is already clearly low, around 35 to 40%. Keep this factor in mind when someone shows you a figure without stating its source: if you’re told a paid campaign has 70% retention, either the wrong column was read, or it was doctored.
What is a good 3-second retention rate in 2026, by platform?
Organically, aim for at least 60% of viewers still watching at 3 seconds; on paid Meta, anything above 25% puts you above the market median. One honest caveat first: Meta publishes no official threshold for a “good hook rate”, as Sepia Lab pointed out in its 2026 benchmarks analysis. Everything in circulation comes from agencies, tool vendors or proprietary samples.
| Context | Measured threshold | Weak zone | Healthy zone | Source and date |
|---|---|---|---|---|
| Organic short-form, all platforms | 3 s ÷ views | < 50% | 60% and above | Grow-it, 2026, on 50,000+ assets |
| Top organic performers | 3 s / 5 s / 10 s | , | 65–80 % / 45–60 % / 25–35 % | Get Ryze, 2026 (vendor) |
| Organic TikTok | 3 s view rate | < 55% | ≥ 65 % | Focusee, Q1 2025 benchmarks (dated) |
| Paid Meta, all audiences | 3 s hook rate | < 14% | 21% (median) to 30% | Biddyco, 2026 |
| Paid Meta, Reels placement | 3 s hook rate | < 20% | 24–36 % | AdSights, 2026 (vendor) |
| Paid TikTok | 2 s view rate | < 15% | 20–35%, median ~27% | AdSights, July 2026 |
Three caveats before you use these figures. The Focusee data dates from the first quarter of 2025: it is ageing, as formats and average lengths have moved since. The 30.7% TikTok average relayed by several agencies in 2025 rests on only eleven accounts, which makes it an indication, not a norm. And the claim doing the rounds at some vendors, “a creative with a 40% 3 s view rate costs 60 to 80% more per conversion than one at 75%”, is backed by no published methodology. A plausible order of magnitude, but not verifiable data.
Another point you almost never read in French: on TikTok, the window in which the algorithm decides whether to push or bury a video plays out in the first thirty to ninety minutes after publication, across a bundle of signals in which early retention carries weight, but so do rewatches, shares by private message and comments. A video with 55% 3-second retention but heavily reshared will sometimes come out ahead of a 75% video that nobody comments on.
Three sentences we should stop saying in meetings. “My hook rate is 25%, that’s bad”: it is the upper market median on Meta. “We have 70% 3-second retention on paid, the creative is good”: impossible, you are reading the organic curve. “You have to grab attention in 3 seconds”: on TikTok, the ad view triggers at playback start, and the platform also publishes separate tiers at 2 and 6 seconds: the Meta 3 s benchmark cannot be transposed.
Why is a low hook rate not always a hook problem?
Three variables external to the creative shift the hook rate by ten to fifteen points: audience temperature, placement and image format. Comparing two hook rates without having neutralised these three is like comparing two temperatures taken under different armpits.
Audience temperature, the first false culprit
According to AdSights data relayed by Sepia Lab in 2026, the hook rate in warm retargeting sits between 30 and 45% (median around 36%), against 18 to 28% in cold prospecting (median around 22%). Fourteen points of difference due to the audience alone. The same creative, served to a 30-day visitor list and to broad targeting, changes category without a single frame moving.
We had this case in 2025 with an accessories client: a creative considered “the best hook in the account” at 38%. It was running at 92% on retargeting. Put back into cold prospecting, it fell to 19%. Nobody had looked at the impression breakdown. That creative wasn’t good; it was well addressed.
Placement, the second false culprit
Again according to AdSights, Feed in cold prospecting runs between 18 and 28%, Reels between 24 and 36%. An Advantage+ campaign that shifts 60% of delivery to Audience Network will drag the average hook rate down without telling you anything about the opening. Before judging a creative, break it down by placement. It is three clicks in the breakdown, and it saves you from throwing away a healthy asset.
Format, the third false culprit
A native 9:16 cropped to 4:5 for the Feed loses its edges, often its burned-in text, sometimes its subject. The opening becomes unreadable on the first frame. We regularly see a gap of two to five points between a dedicated export and an automatic crop, on the same video.
A fourth case, more subtle: a hook that falls gradually over several weeks with constant audience and placement does not reveal a writing flaw, but wear-out. Different diagnosis, different remedy: you don’t rewrite a worn-out opening, you change the angle. And you don’t deal with that problem on three days of data.
What should you do when 3-second retention is good but sales don’t follow?
A hook rate on its own decides nothing: it is the pairing of 3 s retention × 15 s retention that tells you what to reproduce. Four possible combinations, four different production decisions. This is the grid we use in the weekly creative review, and it fits in one table.
| Low 15 s hold rate | Decent 15 s hold rate | |
|---|---|---|
| Low 3 s retention | Creative to scrap. Neither a variant nor a re-edit: start again on a different angle. | Isolated opening problem. The body works: produce 5 to 8 hooks on it. |
| Strong 3 s retention | Body problem. Move the proof up to the 5th second, cut 20% of the length. | Healthy creative. The bottleneck is downstream: offer, landing page, targeting. |
The top-right cell is the most profitable of all, and the most neglected. A body that retains and an opening that misses is an asset half paid for: you already have the shoot, the product proof, the edit. Only the first three seconds remain to be redone. This is the moment to apply the three-layer method for rewriting the hook, and to produce several openings at once rather than just one; the useful volume of variations is framed in the rule of 3, then 5 to 8 variants per angle.
The bottom-right cell is the one that stings. High retention, decent hold rate, and zero sales. Brands call us asking for “more creatives”. Nine times out of ten, there is nothing to do on the video side: the problem lies in the offer, the price, the shipping costs displayed at step 3 of the funnel, or targeting that brings in the merely curious. Saying so costs us a production order. We say it anyway, because selling five more videos to someone whose funnel leaks comes back to bite you within three months.
How do you measure 3-second retention without being caught out by Meta’s 2026 changes?
Three changes to Meta reporting that came into force in early 2026 break comparison with your historical data: your pre-January benchmarks are no longer aligned with those from after. This point is almost absent from French pages on retention, yet it invalidates half the dashboards we are sent.
- 26 January 2026: deprecation of 10-second video metrics on Facebook, and a switch of vocabulary from Reach / Impressions / Engagement to Viewers / Views / Interactions (source: ExtraDigital, 2026). If your dashboard was pulling the “10-second video views” column, it has been empty since.
- 12 January 2026: the Ads Insights API no longer returns 7-day and 28-day view-through attribution windows, and available history is capped at 13 months on single accounts (source: Aurelius Media, Dancing Chicken, 2026). Your year-over-year comparisons on older assets become partial.
- 3 March 2026: click-through attribution now counts only link clicks, and the engaged video views threshold goes from 10 to 5 seconds (source: Aurelius Media, 2026). A threshold that moves means a benchmark that moves.
The practical consequence is simple. Set a break date in your tracking, 26 January 2026 makes a good marker, and refuse any comparison that straddles it without a note. On our client accounts, we froze a raw monthly export before the switch, to keep a record of the old definition. It took two hours. It has spared us three sterile debates since.
Second precaution: never steer a budget trade-off between TikTok and Meta on raw view metrics. Thresholds differ, attribution windows too, and a TikTok 2-second view is not worth a Meta 3-second view. The method for putting both platforms on a common basis is detailed in our article on how to reconcile your TikTok and Meta figures. In TikTok One reporting, TikTok’s help centre states that the 2-second view rate is calculated by dividing the number of views over 2 seconds by the number of views.
The minimum volume before reading a figure
A hook rate calculated on 2,000 impressions means nothing. On our campaigns, we don’t look at a retention figure before 15,000 to 20,000 impressions on the creative alone, and we wait 72 hours for delivery to stabilise. Below that, you’re commenting on noise. This is the most common flaw in hook tests: six variants launched in parallel, €300 of total budget, and a “winner” declared after 24 hours on a two-point gap.
How much does one point of 3-second retention cost?
A batch of six openings produced on an existing body costs a fraction of a full shoot, and becomes profitable from around €5,000 of monthly media budget on the creative concerned. Here is the reasoning, to redo with your own figures.
Take €10,000 of monthly media on Meta, at an €8 CPM. That makes 1,250,000 impressions. At a 22% hook rate, you buy 275,000 3-second views, or €0.036 per view. Gaining 5 points means moving to 27%, so 337,500 3-second views for the same budget: €0.030 per view. The media-equivalent of the gain is around €1,800 over the month.
Against that, the production cost. Producing six openings on a video body that has already been shot means half a day on set or a partial reshoot with a creator, plus editing. An order of magnitude well below a full production. Even on a prudent assumption, the test pays for itself within the month if the gain materialises.
Two honest points are needed here. First: the gain does not always materialise. Out of six openings tested, we regularly see four that do worse than the original, one that does the same, one that wins. That is the rule of the game, and it is why we test in batches and not one at a time. Second: five points of hook rate gained do not mechanically turn into five points of sales. You are buying more attention at the same price. What you do with it next depends on the body of the video and the offer.
The tipping point, as we observe it: below €3,000 of monthly media, shooting six hooks costs more than the value of the expected gain, so it’s better to concentrate the production budget on a new angle. Between €3,000 and €8,000, one batch of variations per month is defensible. Beyond €10,000, not optimising the opening amounts to leaving several hundred euros on the table every month. The full calibration of the production budget by media tier is set out in our guide on calculating your creative budget by media tier.
Should you optimise campaigns for the 2- or 3-second view?
No, except in specific cases. Choosing a “video views” campaign objective optimised on 2 or 3 seconds means asking the algorithm to go and find profiles that watch a lot and don’t buy. You get flattering view rates and a disastrous cost per acquisition. The hook rate is read as a creative diagnostic, not as a media buying target. The only exceptions we accept: an awareness campaign with a non-negotiable reach objective, or a phase of building warm audiences to feed sequenced retargeting.
The first question to ask when someone shows you a 3-second video retention rate is not “how much”, but “divided by what, on which platform, on which audience and with how many impressions behind it”. The four answers take twenty seconds. And until you have them, the figure you’re being shown isn’t worth discussing.
Frequently asked questions
- What is the difference between 3-second retention rate and hook rate?
- The 3-second retention rate is calculated on started views: it is the organic curve from TikTok Studio or Reels insights, between 60 and 80% when things are working. The hook rate divides 3-second views by ad impressions, which includes deliveries never watched: a median of around 21% on Meta according to Biddyco in 2026.
- Is a 20% hook rate on Meta good or bad?
- It’s average. The Biddyco analysis published in 2026 puts the Meta median at 20.9%, with the middle half of ads between 14 and 30%. Below 14% on cold traffic, the opening genuinely has a problem. Above 30%, first check that delivery isn’t mostly going to retargeting, which mechanically gains around ten points.
- Does TikTok really measure 3-second retention?
- No, not on the advertising side. TikTok Ads counts a view from 2 seconds of continuous playback, with a separate metric at 6 seconds. Applying a Meta 3-second benchmark to a TikTok column distorts the comparison. The TikTok 2-second view rate showed a median of around 27% according to AdSights in July 2026, within a range of 20 to 35%.
- How many impressions before judging a retention rate?
- On our campaigns, we draw no conclusion before 15,000 to 20,000 impressions on the isolated creative, and at least 72 hours of delivery. Below that, the gap between two hook variants is statistical noise, not a difference in quality. Many hook tests are settled in 24 hours on a €300 budget: those decisions are worthless.
- Why don’t my 2025 retention benchmarks match 2026?
- Meta changed its reporting in early 2026: deprecation of 10-second video metrics on Facebook on 26 January, removal of 7-day and 28-day view-through attribution windows from the API on 12 January, and the engaged video views threshold moving from 10 to 5 seconds on 3 March. Any comparison that straddles these dates must be flagged as non-homogeneous.
- Should you redo the video when 3-second retention is low?
- It depends on the hold rate. If 15-second retention remains decent, the body of the video works and only the opening needs redoing: produce 5 to 8 hooks on the same edit. If both figures are low, the creative should be abandoned and you need to start again on a different angle, not a variant.
Do your hooks really hold the third second?
You now know which denominator to read. What remains is producing the openings that move the figure. Esekai is a vertical video studio: creative ads, UGC and organic content for TikTok, Reels and Shorts, with a dedicated Creative Strategist, in-house shooting and editing, and batches of variations delivered fast. Not a 360 agency. Thirty minutes on the phone: formats, budget, action plan. No agency pitch deck.
Book a call →30 minutes to talk formats, budget and action plan.