Most Meta accounts that "don't work" are running one campaign, to one cold audience, with one ad, pointing at a homepage. The platform is doing what it was told. A funnel simply means saying different things to people at different stages, and measuring each stage separately.
Get the tracking right before you spend more
Meta optimises towards the event you feed it. If the Pixel fires on page load and nothing else, the algorithm learns to find people who open pages. Install the Pixel and the Conversions API together so server-side events fill the gaps left by iOS and ad blockers, fire a proper event for the action that matters (purchase with value, lead, booking), verify your domain, and set the event priority order. Then check in Events Manager that the numbers roughly match your own back office. Roughly is fine; wildly different is not.
Stage one: cold traffic that earns attention
At the top, broad audiences beat clever ones. Meta's targeting has improved to the point where a wide audience with a strong creative usually outperforms a stack of narrow interests. Keep the structure simple: one campaign, few ad sets, several creatives.
The job of a cold ad is not to close the sale, it is to stop the scroll and make a clear promise. What works consistently: short videos where the first two seconds show the problem, user-generated style clips, a side-by-side before and after, and a direct statement of the offer with the price visible. Give the algorithm at least three to five different angles, not three colours of the same image.
Stage two: warm traffic that answers the objection
People who watched 50% of your video, engaged with the page, or spent time on the site already know who you are. This audience should see the proof: reviews, results, a founder explaining how it works, a comparison against the obvious alternative, an FAQ ad that handles the three things that stop people buying.
Keep this stage separate from cold prospecting, otherwise you cannot tell which stage is broken when sales dip.
Stage three: retargeting that closes
Site visitors, add-to-carts, people who started a form. Here you can be specific and direct: the exact product they looked at, a reminder of what is included, a deadline if you genuinely have one, free shipping, a discount only if your margin allows it.
Two rules save money here. Exclude recent buyers from everything except win-back campaigns, and cap frequency so you are not showing the same ad eleven times to a list of 900 people.
Stop ad sets bidding against each other
If three ad sets target overlapping audiences, you are in an auction with yourself and your costs rise for no reason. Use exclusions between stages, consolidate small ad sets so each one can leave the learning phase, and resist the urge to create a new ad set every time you have an idea. Fewer, better-funded ad sets learn faster.
Budget: where the money should sit
As a starting point for most businesses, about 60-70% of budget goes to cold prospecting, 20-25% to warm, and 10-15% to retargeting. Retargeting audiences are small; pouring budget into them just raises frequency. If your retargeting ROAS looks amazing, that is because the top of the funnel did the work.
Test creative, not settings
After the structure is right, creative is the lever that moves results. Launch new angles every one to two weeks, let winners run, and kill losers early. Judge an ad on cost per purchase or cost per lead, not on likes. Keep a simple document of every hook you have tested and what it cost, and within a few months you will know your audience better than any targeting option does.
Send the click somewhere built for it
A funnel ends on a page. Match the headline to the ad, show the offer above the fold, keep the form short, and make it fast on mobile. Most "Meta ads don't work for us" stories end at a slow homepage.
Want to know what your Meta ads are really costing you? I'll audit the account and the tracking and send you the gaps, free. Ask for the audit, or see how I run Meta ads management and conversion tracking.
