Network verification was blocked by permissions for most URLs, so I’ll cite only sources I’m confident exist (Apple ATT confirmed; the FTC, eCFR, Meta developer docs, W3C and Google Ads Help pages are long-standing institutional URLs) and skip the ones I couldn’t confirm.
Direct response social videos are short paid video ads, built for Instagram, Facebook, TikTok and YouTube feeds, whose only job is to make the viewer do one measurable thing right now: buy, sign up, install, or book. Unlike brand videos, which are judged on reach and recall, a direct response video is judged on cost per result. Every element is engineered for that number. The first one to three seconds stop the scroll, the middle proves the product solves a specific problem, and the ending states exactly what to tap. They run vertical, usually 15 to 45 seconds, with captions, and they are produced in batches of variants rather than as a single polished film, because the platform’s algorithm, not the creative director, decides which version wins.
How is a direct response video different from a brand video?
A brand video asks to be remembered. A direct response video asks to be clicked, and it is shut off the moment its cost per purchase drifts above the target. That changes everything about how it is written and shot.
| Brand video | Direct response social video | |
|---|---|---|
| Goal | Awareness, recall, affinity | One tracked action |
| Success metric | Reach, view-through, lift studies | Cost per result, return on ad spend |
| Length | 30 to 90 seconds, often longer | 15 to 45 seconds; 6 to 15 second cuts for retargeting |
| Opening | Logo, mood, build-up | Problem or claim in the first frame, no logo |
| Output | One hero film plus cutdowns | Dozens of hook, body and ending variants |
| Shelf life | Months to years | Two to eight weeks before fatigue at meaningful spend |
What does a direct response video actually contain?
Four parts, in order, each with a measurable job.
- Hook (0 to 3 seconds). A problem the viewer recognises, a bold claim, or an unexpected visual. We track hook rate, meaning three-second views divided by impressions. On Meta placements we treat anything under about 20 percent as a dead hook and anything above 30 percent as worth scaling, though the baseline shifts by category and audience.
- Proof (3 to 25 seconds). Demonstration, before and after, a customer on camera, or a founder explaining one feature. This is where the product has to be seen working, not described.
- Objection handling. Price, shipping, “does it fit me”, guarantee. One line each, on screen as text as well as spoken.
- Call to action. A specific verb and an offer, spoken, shown in text, and matched by the platform’s CTA button. “Shop the 20 percent launch price” outperforms “Learn more” in our testing almost every time, because it tells the viewer what happens after the tap.
Which platforms run them, and what are the specs?
All four major feeds take the same core asset: 9:16 vertical, 1080 by 1920, MP4 or MOV, under 60 seconds for safe placement everywhere. Keep text and faces out of the top and bottom roughly 15 percent of frame, where platform interface elements sit. Current placement rules and file limits are published in the Meta Ads Guide, the TikTok Ads Help Center, and Google’s reference on video ad formats for YouTube.
Captions are not optional. A large share of feed video is watched with sound off, and burned-in or platform-generated captions are also what the W3C’s accessibility standard requires for prerecorded video. We burn captions in so they survive re-uploads and dark-mode interfaces.
What does it cost to produce, and what do you get?
Costs depend on talent, locations and how many variants you need, not on the length of the finished ad. From our own production schedules, three tiers cover most briefs:
- Creator-style batch. One or two on-camera creators, phone or mirrorless camera, one day. Typically 8 to 12 concepts cut into 30 to 60 variants. Low five figures in most US markets, less if the client supplies the creators.
- Studio or location shoot with modular edit. Crew of four to six, lighting, product styling, one to two days. A hero spot plus 20 to 40 variants. Mid five figures, more if union talent or multiple locations are involved.
- Ongoing creative retainer. A fixed number of new variants per month, built from footage libraries and new pickups. This is what most brands spending more than a few thousand dollars a day end up needing, because fatigue sets in faster than one-off shoots can replace winners.
The hit rate is the number clients find hardest to accept. In our experience roughly one in eight to one in twelve new concepts beats the current best ad. The other concepts are not waste. They are the test that found the winner, and their footage gets recut into new hooks.
How do you measure whether it worked?
Three layers, read in order. Platform creative metrics tell you which variant to keep: hook rate, hold rate at 15 seconds, click-through rate, and cost per result. Conversion tracking tells you whether the clicks became customers. Business metrics, meaning blended cost per acquisition and contribution margin, tell you whether to spend more.
Tracking is harder than it was. Apple’s App Tracking Transparency framework, enforced since iOS 14.5 in 2021, means a large share of iPhone users cannot be followed from ad to purchase by the pixel alone. Server-side reporting through Meta’s Conversions API and equivalent integrations on TikTok and Google recovers part of that signal. Set this up before the first dollar is spent. Launching with pixel-only tracking is the single most common mistake we see, and it makes the first two weeks of data unreadable.
What goes wrong in production?
- Shooting horizontal and cropping. A 16:9 frame cropped to 9:16 loses two thirds of the image and usually the product. Frame vertical on set, or shoot 4:3 open gate and protect for both.
- Using trending music. Commercial accounts cannot use most trending sounds in paid ads. Clear a track from the platform’s commercial library or license one, or the ad gets muted or rejected after it has started performing.
- One variant at a time. Testing a single new ad against the control tells you nothing about why it won. Hold the body constant and swap hooks, then hold the winning hook and swap endings.
- Undisclosed paid creators. If a creator is paid or given product, the endorsement rules in 16 CFR Part 255 apply, and the FTC’s endorsement guidance expects a clear, on-screen disclosure, not a hashtag buried in the caption. Results claims need to be typical or qualified.
- Landing page mismatch. The ad promises a 20 percent launch price and the page shows full price with a code field. Every mismatch shows up as high click-through and zero conversions.
What has changed since earlier guides were written?
Older articles on this topic, including an earlier version of this one, leaned on quoted statistics and named experts that cannot be traced to a published study, so they are gone from this version. Three real changes matter more. Tracking moved server-side after Apple’s 2021 privacy changes. Meta, TikTok and Google have each folded manual campaign targeting into automated buying, so creative variety is now the main lever an advertiser controls, with the algorithm handling who sees which version. And platforms have tightened disclosure and music enforcement, so compliance has to be built into the shoot, not patched afterwards.
If you are starting from zero, do this in order: install server-side conversion tracking, write ten hooks for one proven product benefit, shoot them vertical in a single day with captions and a licensed track, launch all of them with the same body and ending, and give the test enough budget to reach about 50 conversions per variant before deciding. Then recut the winner’s footage into the next ten hooks.
