Video Ad Benchmarks 2026: Hook Rate 25.44%, CTR 1.93%, ROAS 2.17, and What to Fix

Attention Went Up. The Money Went Down.

You sort Ads Manager by spend on a Monday morning and there it is: a video holding a 27% hook rate, earning clicks all day, returning 1.4 on every dollar behind it. The creative is working. The account is not. Nobody in the thread can say which of those two facts to act on.

That shape defined the first half of 2026. Across 88,329 sales-objective Meta video ads, the cross-industry hook rate rose to 25.44% from 24.42% the half before, CTR averaged 1.93%, and ROAS fell to 2.17 from 2.41, about 10% lower. Attention got easier to win. Revenue got harder to keep.

A benchmark you can only stare at is worthless. So this piece takes the three numbers one at a time: what each actually measures, where the bar sits, what a reading below it usually means, and the specific change that addresses it. One of the three is mostly not a creative problem at all. Knowing which one is the difference between fixing the account and testing your way nowhere.

Where the 2026 numbers come from, and who counted them

What is actually in the dataset

The figures come from Billo, a company that produces UGC video ads for brands, published across three separate benchmark reports in August 2026. The dataset is 88,329 sales-objective Meta video ads that ran between January and June 2026, spanning 14 industries, $122 million in ad spend, and $212 million in purchase value.

The inclusion criteria matter as much as the size. Only sales-objective video ads with reliable metrics and more than 1,000 impressions were counted. That filter quietly removes two populations you might expect to be in there: awareness and traffic campaigns, and the long tail of ads that never got enough delivery to produce a stable number. So this is a benchmark for ads that were asked to sell and got a real chance to.

The 14 industries are categories assigned by advertiser product type, and one of them is an "Other" bucket holding the industries too small to report on their own. It is a wide read of Meta video advertising. It is not every industry.

One vendor's customers, not an industry census

Here is the part most benchmark write-ups skip, so let us be direct about it. This is proprietary first-party data from one vendor's own customer base. No third party audited it, and none could: the raw ad accounts are not public. Billo sells UGC video ad production, which means it has a commercial interest in conclusions of the form "your creative opening matters." We sell AI video ad creation, so we have the same interest, and you should read both of us with that in mind.

Two outside items have picked up Billo data this year, a press release and a newsletter, but each covers a different cut of ads over a different window. They show that this methodology recurs from half to half, which is mildly reassuring. Neither of them verifies a single figure below.

So treat these as the best public read available rather than as a law. The honest use of a vendor benchmark is as a bar to aim at and a way to notice you are far off it, not as the number your board holds you to.

The cross-industry average describes almost nobody

Every one of the three reports says a version of the same thing, and the CTR report says it most bluntly: "benchmarking against the 1.93% cross-industry average is more misleading than ever. The average describes almost no one."

The arithmetic backs the warning. On hook rate, the 14 industries ranged 7.47 pp from highest to lowest while the cross-industry average moved only 2.53 pp across six months, so your industry mattered about three times more than the month did. On ROAS the gap between top and bottom was 3.95 while the average moved 0.29 all half.

Which means the single most useful thing on this page is not a number. It is the instruction to go find your own category's figure in the source reports and judge yourself against that.

Hook rate: the 25.44% bar and the first three seconds

What hook rate measures

Hook rate is 3-second video plays divided by impressions. It is the share of people who were served your ad and did not scroll away in the opening seconds. Nothing more. It says nothing about whether they liked the product, understood the offer, or ever considered buying.

That narrowness is what makes it useful. Of the three metrics here, hook rate is the one most fully inside a creative team's control, because almost nothing outside the first three seconds of the file can change it.

Where the bar sits, and where yours sits

The cross-industry average was 25.44%, against 24.42% last half, and the report's own guidance is to use 25.4% as your baseline. Roughly one viewer in four staying past the opening was an average advertiser this half.

Then look at the spread. Toys & Games averaged 28.79%, the best of the fourteen. Animals & Pet Supplies averaged 21.32%, the worst. And the single strongest month belonged to a category nobody would have picked: Business & Industrial reached 33.03% in March, and no other category reached that figure in any month. A category with nothing to unbox posted the best attention numbers in the dataset.

The lesson is not that industrial ads are magic. It is that the hook advantage goes to whoever breaks the sameness in their feed, and a category where every competitor opens on a boring product shot is a category where a good opening is cheap to win.

A low hook rate is a first-three-seconds problem

If your hook rate sits under your category's number, the fix is bounded to about 90 frames. Resist the urge to re-edit the whole video. The changes that move this metric:

• Open on motion or a state change, not a logo, a title card, or a slow push-in. The two leading consumer categories both open on something happening: a toy mid-play, a before-and-after.
• Show the product in use in frame one. The lowest category in the table most likely never gets its subject on screen in the opening seconds.
• Cut the first two seconds off your current best ad and run it as a variant. This costs nothing and is the highest-yield test in the list.
• Change the face, not just the words. A different person delivering an identical line is a genuinely different opening, which is why UGC-style casting is a hook-rate lever rather than a style preference.
• Vary the frame, not the script. Handheld versus tripod, indoors versus outdoors, close versus mid. Platform guidance on creative standards converges on the same point: the opening seconds carry disproportionate weight.

If you want a library of openings to borrow from before writing your own, working through real video ad examples is faster than brainstorming.

CTR: the 1.93% bar and the promise you did not make

What CTR measures

CTR here is link clicks divided by impressions: the share of people served the ad who went on to click. It measures whether the ad earned the next step and not just the attention.

Note the denominator. It is impressions again, not 3-second views. This matters more than it sounds, and the next section is built on it.

The spread is the story

The cross-industry average was 1.93%, and eleven of the fourteen categories came in below it. That is not a rounding artifact, it is what an average pulled up by a few strong categories looks like.

The range ran 0.99 pp, from Health & Beauty at 2.36% down to Baby & Toddler at 1.37%. The report's guidance is to use 1.93% as your baseline and check the month before you judge a number against it, because the same 1.93% was slightly short of the market in February and comfortably clear of it in June.

Two categories did something worth naming: Business & Industrial and Sporting Goods both raised their hook rate this half while their CTR went the other way. They are the only two that split that way, and they are the cleanest proof that these two metrics are not the same metric measured twice.

A low CTR is an offer and promise problem

When the hook rate is fine and the CTR is not, people are watching and choosing not to act. That is almost never a production-quality issue. It is that the ad never told them what they would get by clicking.

• Put the offer in the ad, not only on the landing page. A price, a bundle, a first-order discount, a guarantee. When the calendar supplies no urgency, the offer has to.
• Name the outcome out loud in the second half of the video. Watching a product work and being told what it does for you are different experiences.
• Give a reason to click today. The lowest-CTR categories are the ones people buy when they run out. A subscription discount or a bundle manufactures the timing the calendar will not supply.
• Test one promise per variant. If three variants share a promise and differ only in edit, you are testing production, not messaging. This is the discipline that structured creative testing exists to enforce.
• Check the platform, not just the ad. Click behavior differs enough between feeds that a Meta versus TikTok comparison is worth doing before you conclude the creative is at fault.
• Do not chase the number on its own. A CTR above the bar with nothing behind it is its own trap, which is why a high CTR does not mean the ads are working and why the metric is worth reading next to the rest of the creative metric ladder.

ROAS: the 2.17 bar and the part creative cannot fix

What ROAS measures here

The definition used is each ad's purchase value divided by its own ad spend, then averaged across ads. Worth reading twice, because it is not the same as dividing total purchase value by total spend. It describes what a typical ad returned, not what the portfolio returned.

The half that fell

Last half the cross-industry average was 2.41. This half it was 2.17, about 10% lower. Eleven of the fourteen categories came in under their own prior-half figure.

And the spread is brutal. The fourteen categories went from Baby & Toddler at 4.99 down to Software at 1.04. Software averaged 1.04 and Animals & Pet Supplies averaged 1.05, near enough a dollar back for a dollar in, and that is before the cost of making and shipping anything, so the average ad in both categories lost money.

Before anyone builds a theory on the decline, the ROAS report says the quiet part itself: "nothing in this data says why, and six months isn't enough to separate a market-wide move from a change in who was advertising." A vendor that could have sold a story about rising ad costs declined to. Take the same posture with your own account.

Why a low ROAS is usually not a creative problem

Here is the finding that should change how you triage. "The categories that got the most attention mostly made the least money. Toys & Games had the highest hook rate of the fourteen at 28.79% and finished twelfth on ROAS. Baby & Toddler had the lowest click-through rate at 1.37% and finished first."

A video that almost anyone will watch pulls in a wide audience that was never going to buy. Attention is not intent, and at the extremes they run in opposite directions.

So when ROAS is under your category's bar and the first two metrics are fine, the creative has already done its job. The revenue is being lost after the click, and every usual suspect lives outside the video file:

• Landing page load speed, layout, and whether it repeats the promise the ad made.
• Price against the category, not against your costs.
• Shipping cost revealed late, at the step where carts die.
• Checkout friction: forced accounts, too many fields, one payment method.
• Margin that was never there, which no ROAS figure can rescue.
• Attribution window too short for a considered purchase.

A more charming actor will not fix any of those.

We sell AI video ad generation and we will still say it plainly: creative cannot fix a landing page, and it cannot fix a price. If your ad spend is climbing while ROAS falls, spend the next week outside the ad account.

Reading the three together

The diagnostic table

Read the three as a combination rather than a scoreboard, and each pattern points somewhere different.

What you see	What it usually means	What to change
Hook under bar, CTR fine	Opening seconds, not the offer	Rewrite the first three seconds
Hook fine, CTR under bar	The ad promised nothing specific	Put the offer in the ad
Hook and CTR fine, ROAS under bar	Post-click, price, or margin	Audit the landing page first
Hook and CTR high, ROAS far under bar	Broad appeal, narrow buyer set	Cast and script to the buyer
All three under bar	Wrong angle or wrong audience	Change the angle, not the edit

A worked example

Take a round number. An ad delivers 200,000 impressions. At the 25.44% cross-industry hook rate, about 50,880 people stay past the opening seconds. At the 1.93% CTR, about 3,860 click.

Now notice what did not happen. The 3,860 is not a share of the 50,880. Both figures divide by the same 200,000 impressions, so hook rate and CTR are siblings, not steps in a funnel. Doubling the number of people who watch three seconds does not arithmetically produce more clicks, which is exactly how two categories in this dataset raised hook rate while their CTR fell.

Carry it through to revenue. Say 2% of those 3,860 clicks convert at a $117 average order value: 77 orders, about $9,000 of purchase value. On $4,000 of spend that ad returns 2.25, a little above the 2.17 average. Now leave the video completely untouched and move the landing page from 2% to 3%: the same ad returns about 3.4. The video did not get better. The page did. That is the single most common misattribution in creative reviews.

Fix them in order

Work up the funnel, cheapest and fastest first.

1. Hook rate first. It is the cheapest to change, reads in a day or two of delivery, and is entirely inside the file.
2. CTR second, and only by changing the promise. A new offer is one variant, not ten edits.
3. ROAS last, and start outside the ad account. If the first two are at or above your category's bar, the ad is not the constraint.

Run each as a real test rather than a swap, changing one variable per variant. The same split-testing discipline applies whichever platform you are buying on.

How Novoads solves the volume problem

Diagnosing a hook rate takes an afternoon. Producing twelve genuinely different openings to fix it is where most small teams stop, because each one has historically meant a brief, a creator, and a week of waiting.

Novoads generates UGC-style video ads from a product image or a written script: pick an AI actor, supply the script, get an ad-ready vertical video. That makes "cut the first two seconds and recast the opening line" a same-day change instead of a new production cycle, which is the only way the fix list above becomes a routine rather than a quarterly project. You can start generating ads on the Inicial plan at $49/month, which includes 50 credits every month. Cancel anytime.

Creative owns two of these three numbers

The most valuable thing in the 2026 data is not the bar. It is the separation. Hook rate, CTR and ROAS pulled apart from each other more this half than in any half measured, and the metric that looked strongest in a category was frequently not the one that made money.

Own the first two honestly and relentlessly, because they are yours. Then be disciplined enough to admit that the third one is mostly telling you about your page, your price and your audience. The teams that beat their category's benchmark are not the ones with the best-looking ads. They are the ones who know which number they are actually fixing.