We price almost everything about a streaming ad break except the part viewers feel most: where it lands in the story.
You know the moment. You're an hour into a thriller, the detective has just pushed the door open, and the screen cuts to an ad for car insurance.
The rest of the break is getting priced. In September, FreeWheel started selling attention, quality and emotional resonance as signals for CTV buying. Its scores also take in the number of breaks, how long the ads run, the time of day, live or on demand, and the device.
A couple of years ago most of that wasn't priced at all, so this is real progress.
Every one of those inputs still describes how much interruption there is, though. To a buyer, a break at the end of a scene and a break in the middle of a reveal look the same: one break, two minutes.
Who pays for the second before the break
The second before a break decides how the viewer meets the ad. If the story has just come to rest, the ad arrives as a pause. If the detective is halfway through the door, the ad is what stands between the viewer and the reveal.
The publisher chooses that second. I think most of the cost of a bad choice lands on the advertiser, because theirs is the brand the viewer meets while annoyed. The platform pays later if people start giving up on episodes.
We already know advertisers lose out when the viewing experience gets worse. In a FreeWheel study with MediaScience, reported by PPC Land, unaided recall was 41% in a lightly loaded programme and 26% next to nine minutes of ads per half hour.
That study measured the amount of advertising. I couldn't find an independent industry study that compares a mid-scene break with one at a natural pause, on recall or on drop-off, so nobody can yet put a price on the difference.
More breaks, and fewer of them planned
Ad tiers carry more ads this year. Ampere puts the average across the big US services up about 18% between January and August.
Each of those extra minutes has to go somewhere in a story. Network dramas make that easy, because the writers build act breaks into the script.
Films and older catalogue titles were made to play straight through, so their breaks get added afterwards to a cut that was never meant to stop.
Usually an editor finds them by scrubbing through the timeline, and a single 40-minute episode can take 20 to 60 minutes to segment by hand.
The faster option is a rule that looks for black frames and silence. In film that misses a lot, because a fade to black often has music or a voice running over it.
What a natural pause looks like
We've worked on this at Vionlabs for years. The main thing we've learned is that the picture and the story don't always pause in the same place.
A story is resting when several things are true at once. The scene has ended, nobody is speaking, the next shot looks clearly different from the last one, and the tension has eased.
No single signal sees all of that, so our ad-break detection combines four models. One spots black frames and avoids cutting into speech, one follows the flow of the story, one finds scene boundaries, and one compares how similar the shots are.
The more of them agree on a spot, the higher it's rated (excellent, good or passable). An editor then reviews a ranked shortlist instead of the whole film.
It still makes mistakes. In a webinar with SDVI this summer I described one: a shot that pans into darkness can look like a black frame to the model, so it gets flagged as a break.
That's why we built it for a person to check the result. SDVI has shown our detection on its Rally platform with Codemill's Accurate.Video, where editors validate each suggested break frame by frame.
The case for the break
The best argument against me comes from research I like. Across six studies, Nelson, Meyvis and Galak found that commercial breaks can make a programme more enjoyable, because the pause stops us getting used to it.
They didn't test where the break sat, so this part is my reading. I think the breaks that help are the ones a story can absorb, like the cliffhanger network TV has used on purpose for decades.
The breaks that worry me are the ones nobody designed. One gets dropped mid-conversation because the clock reached 14:00, another into a long, quiet scene because the picture went dark.
YouTube seems to agree that position matters. Its Peak Points pilot, announced in May 2025, uses Gemini to place an ad right after a peak moment such as a marriage proposal. I haven't seen published results yet.
Placement should be priced
Context is starting to travel with CTV ads. Several ad-tech companies now add contextual data to bid requests or describe the tone of each scene, so a matching ad can be picked.
All of that describes the content around a break. Where the break itself sits is taken as given, and I couldn't find anything in the industry's ad standards that records it.
Display advertising had a similar gap. For years ads were bought on served impressions, with no way to know whether anyone could see them. The Media Rating Council set a viewability standard in 2014, and buyers now sign viewable-only deals that don't pay for ads nobody could see.
I expect break placement to go the same way. A break-quality score, created where the content lives and carried with the cue point, would let buyers pay more for breaks that keep the story intact, and give publishers a way to prove it.
Where I'd start
If you run an ad tier or a FAST channel, these tests are cheap:
- Drop-off in the 30 seconds after a break, split by whether it landed at a scene boundary or mid-scene
- Completion of the first ad in the pod, with the same split
- How often people finish a title when its breaks follow the story, compared with breaks on a timer
And if you sell CTV inventory, I'd like to hear from you. Would you charge more for a break you could prove lands at a natural pause, or will buyers never pay for position?
Want to see where the natural pauses fall in your own catalogue?
.png)
.png)
