The same reel got 30x more views - one line changed

Under Meta's new algorithm the creative is the targeting, so the hook, the system behind it, and the account it runs in are the only levers left that actually move money. Here is what's working right now, with the exact plays to steal.

A single hook swap took one reel from ~300 views to ~9,000 with zero other changes. A creator posted an identical reel to a second account and changed only the first three seconds. Version A opened with "Here are 3 editing tips." Version B opened with "You're doing your first 3 seconds wrong and it's killing your reach." A got ~300 views, B got ~9k. Why it worked: "tips" is a promise of work, a pain-point line is a promise of relief, and the hook is not decoration, it is the distribution mechanism that decides whether the platform keeps showing your ad. The move: before you film, write 5 hooks and pick the one that would stop you mid-scroll, per the creator's own experiment.

A hook that converts front-loads the text and disqualifies the wrong viewer. Most brands write the hook backwards, starting from what they want to say about the product, when the viewer only cares about their own problem. Here is the actual mechanic that takes a hook rate from ~25% to ~50%, broken down here:

The one-line version to keep on your desk: make it about them not the product, be specific ("3x my conversion rate" beats "improved my results"), and speak to the pain not the dream, because pain stops scrolls and dreams get skipped.

Steal the prompt that engineers the hook for you. One direct-response prompt tells the model its job is not to sell but to make it impossible to stop reading until the reader arrives at the product. It forces a pre-write pass on the audience's hidden fear, biggest frustration and stuck belief, then brainstorms 15 hooks and scores each on curiosity and scroll-stopping ability before writing a single line of body copy.

Meta's Andromeda update made unstructured testing a way to burn money, because the creative now does 100% of the targeting. The algorithm scrapes your visuals, copy and landing page to decide who to show the ad to before it spends a dollar, so audience targeting is nearly irrelevant and generic ads die within weeks. An operator who audits accounts spending up to $12M/month says the fix is a system, not more tests: mine the exact language buyers use from Reddit and reviews, then map a full-funnel angle set across all 5 awareness levels so the algorithm always has the right ad for the right person.

The reason you can't out-test the supplement incumbents is unit economics, not creative. An operator with 7-figure/month supplement clients lays it out: brands like Gruns, IM8 and Seed win because they hit 50%+ subscription retention by month 3 and 30%+ by month 6, plus investor capital, so they can pay $100 to acquire a customer you're trying to get for $35. The move before you touch creative: fix LTV (subscription, email retention, repeat purchase), because the best ad in the world can't save math that doesn't work.

Build the creative system before you need the volume. IM8 runs 1,500+ active ads, but they started with 28, found the winners and compounded them. The Creative Data System behind $3M/month in spend is the scaffold: pull personas from real review data, map the 3 awareness-level ad types most accounts skip, rank a live angle leaderboard, and iterate winners instead of starting from a blank page each week. When a winner appears, extract its DNA (the angle, the emotional trigger, the framing) and compound it three ways: swap the hook onto a proven format, run the same angle in a totally different format, and write 5 hook variations where one will usually outspend the others 30 to 1.

If you want it on one page, use 4 phases and 12 questions. A strategist who runs this with every DTC brand frames creative strategy as the hardware and ads as the software: Research (when does this person feel the problem, why, who), Ideation (which problem, what format, where), Concept Development (how to tell the story, which content type, what platform), and Production (who builds it, how long, which metric defines the win). Define the win metric before launch, not on day 3 when hook rate looks fine but CPA looks scary.

Most stalled accounts aren't a creative problem, they're a config problem that manufactures fake winners. Two live audits from a buyer this week (the 5-flag checklist and a $100k/mo teardown) surfaced the same silent leaks. Go check yours:

The landing page, not the ads, was killing the ROAS. A DTC team spent months testing hooks, copy and audiences for marginal gains, then ran a CRO audit on the page and found a 4.8s mobile load (benchmark is under 2s), three competing CTAs, no social proof above the fold, and an add-to-cart hidden below the scroll. Fixing those four things lifted ROAS 31% without touching the ad account. Ads earn the click, the page decides the return.

Same creative, rebuilt structure, back-to-back record days. An operator rebuilt an account on a Friday and hit his 3rd highest revenue day Saturday, then his highest ever Sunday, with the exact same ads. He moved from creative-testing campaigns and auto-bidding to 100% manual bids, with ad sets split by economic unit or product category and bids set relative to AOV. "Media buying is dead" is a myth, structure still moves revenue on its own.

Steal the 3-campaign skeleton per offer. A clean structure to run it: (1) an ABO testing campaign, one ad set per concept at $100/day on lowest-cost; (2) a CBO scaling campaign where ABO winners get duplicated in as fresh ad sets; and (3) a "graveyard" campaign under a strict bid cap where every loser goes, so creative you already paid for still recovers spend instead of being wasted.

Treat every ad like a $250 casino chip you only get a few of per week. A strategist reframes "learnings" as starting before the creative ships: each test costs roughly one ad's spend before you can call it, so your team should defend every iteration, format and avatar choice like they're in court, backed by data and pattern recognition, not "I saw this format on TikTok and thought it was cool." A low hook rate on a past ad isn't failure, it's a signal the ad only filtered your ICP and wasn't top-of-funnel enough.

Stop reaching for multivariate testing too early. It isn't the "advanced" A/B test, it's a different tool. Test 5 things at once and you split traffic into 8 buckets that never reach significance, turning a one-week test into a two-month slog. The rule: if you can't fill every combination with traffic, run sequential A/B tests instead.

Before you spend another $10k testing creative, check the offer. One blunt reminder: $10k of testing on a weak offer just pays to expose it at scale, faster. Fix the offer and the "winning creative" tends to show up on the first try. Endless creative iteration is often the wrong diagnosis for what's really an offer problem.

The field is more AI than Twitter admits, and video is the real edge. A scrape of every ad SF startups are running on Meta right now found the benchmarks: only 4% of startups run paid ads (1.6% of YC companies), 73% of ads are UGC, 53% are AI-generated, and 78% are video, which converts 32% better than static. If you take one number to your next planning meeting, make it that last one.

How one client increased their monthly spend from $85K to $400K without changing their target ROAS or campaign structure. The lever wasn't a bigger budget, it was three axes working together: Personas × Awareness Levels × Formats. Steal the exact steps: (1) export your real customer reviews, drop them into Claude, and segment by pain point, emotional trigger and failed solution to find personas your ads have never spoken to; (2) map each persona to a top, mid and bottom-of-funnel ad so the algorithm can target across awareness levels instead of only reaching people who already know you; (3) prove the new angle in a format you know converts, then port the messaging into a fresh format. The ceiling on your spend is the number of angles you're testing. The full framework if you want the receipts.

Your next Reel or Short should start from a hook that already worked, not a blank first frame. Open the HookAds library and pull one.

Open the hook library →

What's HookAds? HookAds is a swipe library of short-form ad hooks and creative angles, so your next Reel or Short starts from something that already worked instead of a blank page.

Thanks for reading 👋
— Piyush, HookAds

HookAds · short-form ad creative that converts · you signed up at hookads.ai