For years, advertisers obsessed over targeting. Stack the audiences, layer the interests, hunt for the magic segment. Those days are over. The platforms now sell campaign types that pick the audience for you. Meta says Advantage+ uses AI to match ads to the people most likely to take action, and TikTok's Smart+ automates targeting, bidding and creative in one campaign.

So if your return on ad spend is sliding, a fourth interest audience in Ads Manager is rarely the fix. A better ad is. Your hook decides who stops and your message decides who stays. Then Meta or TikTok goes looking for more people like the ones your ad pulled in.

That's good news, by the way. You will never out-engineer the platform's machine learning at its own game. You can out-create the business down the road.

Creative is the new targeting

When every advertiser has the same auto-optimising delivery, the ad itself is the biggest lever you have left.

Think about how delivery runs now. You hand the platform a broad audience and an ad. It shows the ad to a spread of people, watches who clicks and who buys, then hunts for more people who behave like the buyers. The ad is the filter. An ad that speaks to busy mums who hate meal prep will find busy mums who hate meal prep, whether you ticked a "parents" box or not.

That's why two advertisers with identical settings and identical budgets can get wildly different results. Same settings, different signal. For the full case for letting go of the audience builder, read why broad targeting wins in 2026.

You no longer beat the algorithm with settings. You beat it with a better first three seconds.

How to tell if you have a creative problem

Before you tear anything down, diagnose it. Creative problems leave fingerprints in your ad account, and each one points somewhere different.

  • People aren't stopping. A low three-second view rate means your opening isn't earning attention, so the rest of the ad never gets a chance.
  • People stop but don't stay. Decent early views that fall off a cliff mean the hook promised something the middle didn't deliver.
  • People watch but don't click. Strong watch time with a weak click-through rate. The ad entertains and never gives a reason to act.
  • People click but don't buy. Usually a landing page or offer problem. Check the page matches what the ad promised before you blame the video.
  • Everything was fine, then it wasn't. Strong results that slowly decay point to fatigue. Meta's definition is an audience overexposed to the same creative. Fresh angles fix that. New settings don't.

Work top down. There's no point polishing the call to action on an ad nobody watches past second one.

The anatomy of a high-ROAS ad

  • Hook (first 3 seconds): A pattern interrupt. Something that makes the thumb stop before the viewer can swipe.
  • Tension: Agitate the pain or amplify the desire, so staying feels more interesting than leaving.
  • Proof: Show, don't tell.
  • Ask: One clear action. Confusion kills conversion.

Hooks that work usually do one of four things: call out the viewer ("If you own a café, stop scrolling"), name a painful problem ("Your oven is costing you $40 a week"), make a surprising claim ("We stopped running Google Ads and sales went up"), or show something odd enough that the brain has to look. Meta's own Reels guidance says viewers decide within the first few seconds whether a video is worth watching. Put the hook on screen as text as well as saying it. When Facebook studied its video ads, 41% were basically meaningless without sound.

Tension is the reason to keep watching. Open a loop: "Here's the mistake almost every renovator makes." Now the viewer needs the answer.

Proof is where most ads are weakest. A founder saying "our product is amazing" proves nothing. A customer holding up the before and after photos does the job in four seconds.

The ask should be single and specific. "Tap below to book your free measure and quote" beats "Learn more about our range of services." For a deeper version of this structure, the hook, story, offer formula breaks it down further.

Volume + iteration = control

The brands winning on paid media aren't producing one perfect ad. They ship dozens of variations, let the data crown the winners and put the profit into new angles. We treat creative like a portfolio, constantly testing hooks, formats, and offers so your cost-per-acquisition trends down while spend scales up.

Volume doesn't mean 50 copies of one ad with a different background colour. Meta says its system groups together ads that share key visual and thematic attributes, and shares what it learns across the group. In practice, 50 lookalikes teach it about as much as one. You want variety where it counts:

  • Angles. Different reasons to buy. Saving time, saving money, looking good, avoiding a disaster.
  • Formats. Talking head, customer testimonial, product demo, static image, carousel, founder story.
  • Hooks. Several openings on the same body, because the opening decides whether the rest gets watched.
  • Offers. A free quote against a discount against a guarantee. Swapping the offer can move results more than any edit.

We unpacked the change behind this in Meta's Andromeda update, where a handful of different concepts beat piles of lookalike variants.

Fix my ad creative, lift my ROAS →

A simple creative testing system you can run this month

You don't need a big budget to test properly. You need a structure you'll stick with for more than a fortnight.

  1. Write down three angles. Three different reasons someone would buy from you. Pull them from customer reviews and sales calls. A brainstorm in the office won't give you the customer's words.
  2. Make two formats per angle. One talking-head video and one static image, for example. That's six concepts.
  3. Write three hooks for each video. Same body, different first three seconds.
  4. Launch them in one broad campaign. Let the platform distribute spend, and keep your hands off the switches for the first few days.
  5. Judge on cost per purchase or cost per qualified lead. Likes don't pay invoices. Give each concept enough spend to tell you something before you decide.
  6. Kill the losers, clone the winners. Make three new variations of the winning angle. Try the winning hook on a different format.
  7. Repeat every two to four weeks. Winning ads wear out, and a steady flow of new concepts stops your costs creeping up.

If your winners seem to die faster than they used to, read about ad fatigue and the shrinking ad lifespan.

Producing creative in volume without a film crew

The obvious objection: "I can't afford to make twenty ads a month." You don't need a film crew for this, because the ads that suit a small account are cheap to make. Think of a founder talking to a phone camera about the one mistake customers make, or a customer answering a single question on video. Screen recordings of the product doing its job work too. So does a plain static with a strong headline and a real review.

Batch it. Block out one Tuesday morning and record ten short pieces to camera on your phone, each built around a different angle or hook. Add customer footage and a few statics and you have a month of testing material. The polished production budget can wait for the concepts that have already proven they sell.

A worked example: the kitchen renovator

Say you run a Melbourne kitchen renovation business spending $5,000 a month on Meta ads. You have one ad running: a polished slideshow of finished kitchens with the headline "Quality Kitchens, Affordable Prices." Results were decent a few months ago. Now leads cost more every week, and you've tried three new interest audiences with no change.

That's a creative problem. The ad says what every competitor says, and the audience has seen it.

So you build three angles. First, a guarantee: "We finish on time, or you don't pay the final invoice" (only if you can honour it). Second, education: "The three mistakes that blow out a kitchen budget." The third is proof, a customer walking through their new kitchen and explaining what made them nervous before they started.

Each gets a video and a static, and the videos get three hooks each. Within a few weeks the spend data shows which one the market cares about. Maybe the budget mistakes video pulls the cheapest qualified leads by a clear margin. Now you know your buyers worry about cost blowouts, so every new ad leans into that. Same $5,000 and the same broad audience, with a very different lead sheet. (Every number here is hypothetical, but the pattern is how testing works.)

Common creative mistakes that crush ROAS

  • Leading with the logo. Nobody cares who you are in second one. They care whether this is relevant to them.
  • Making ads that look like ads. Heavy branding and stock footage get scrolled past. Something that looks like a mate's post earns a second look.
  • Talking about features instead of outcomes. "Soft-close drawers" is a feature. "No more slamming at 6am while the kids sleep" is an outcome.
  • Testing tiny changes. A new font teaches you nothing. A new angle will.
  • Turning ads off too early. One bad day is not data.
  • Letting winners run forever. Every ad has a shelf life, so have the next one ready before this one fades.

Where to start this week

Open your ad account and pull up your top three ads by spend. For each one, name the hook and the angle, then point to the proof. If you can't, your audience can't either.

Then read your last 20 Google reviews and write down the exact words people use to describe why they bought. Those phrases are your next hooks. Customers describe the problem the way other buyers think it, which is why their words beat most agency copy.

At LIQID NOISE, our advertising work is built on this: strategy-led creative, produced in volume, tested with discipline and judged on revenue. If your targeting is already broad and your ROAS still won't move, stop fiddling with settings and fix the ads.

Frequently asked questions

Does ad targeting still matter on Meta and TikTok?

Targeting still matters at the edges. Set the right country, exclude existing customers where it makes sense and give the platform accurate conversion data. For most advertisers, detailed interest stacking adds little beyond that. The platform finds buyers based on who responds to your ad, so the creative now does most of the work manual targeting used to do.

How many ad creatives should I test at once?

For most small to medium accounts, start with three different angles and two formats each, around six concepts, with a few hook variations on the videos. That's enough variety to learn something without spreading a modest budget too thin. Add more concepts as spend grows, and make each one differ in angle or format. Small tweaks don't count.

Why did my ROAS suddenly drop?

Sudden ROAS drops usually come from creative fatigue, a tracking problem, a landing page change or a shift in competition. Check tracking first, since broken purchase events make results look worse than they are. Then look at frequency, and at how long your best ads have been live. Fresh creative is usually the fix when people have seen the same ad again and again.

What is a good hook for a video ad?

A good hook makes the right person stop in the first three seconds. It might call out the viewer directly, name a painful problem, make a surprising claim or show something unexpected. Put it on screen as text as well as speech, because a lot of feed video plays with the sound off. Test several hooks on the same ad body to find the one that works.