Interest stacking stopped working. Most accounts never noticed.
Meta’s delivery system now identifies buyers from conversion signals more accurately than any interest combination a media buyer can assemble. Yet accounts still run dozens of narrow ad sets fragmenting the same budget, each too small to exit the learning phase, all served the same three creatives that stopped working two months ago.
Too many ad sets, too little data
Fragmenting budget across narrow audiences keeps every ad set in the learning phase, where delivery is least efficient. Consolidation usually improves performance immediately.
Creative fatigue read as rising costs
Performance decays with frequency. Without a production and retirement cadence, accounts pay progressively more for the same declining attention and blame the auction.
Browser-only tracking
Since iOS privacy changes, pixel-only setups lose a large share of conversion signal. The algorithm optimises on partial data and reported results understate what actually happened.
Retargeting mistaken for performance
Retargeting reports the best returns in nearly every account because it reaches people who already decided. It is also where over-investment almost always hides.
Consolidate the structure. Industrialise the creative.
Broad targeting with strong conversion signals now outperforms hand-built interest stacks in most accounts. That frees the work to go where it matters — producing enough creative variety for the algorithm to optimise against.
We run a standing production and testing calendar rather than commissioning assets when performance dips. By then you have already paid for weeks of decay.
- Consolidated account structure — fewer, larger ad sets so budget concentrates and campaigns exit the learning phase quickly.
- Creative production system — modular assets in volume — hooks, formats and angles varied systematically rather than one polished hero video.
- Conversions API — server-side events so the algorithm optimises on complete data rather than what survives browser restrictions.
- Fatigue monitoring — frequency and creative decay tracked with retirement rules, so assets are replaced before costs climb.
- Incrementality holdouts — geo tests establishing what Meta spend genuinely caused, separating real lift from claimed credit.
The uncomfortable finding: in most incrementality tests we run, retargeting produces far less additional revenue than its reported ROAS suggests. Those people were coming back anyway. The spend was buying credit, not customers.
How we rebuild a Meta Ads account
Six workstreams, signal and structure first — because creative testing on broken tracking teaches you nothing reliable.
Conversions API setup
Server-side event tracking with proper deduplication and event match quality, so delivery optimises on complete rather than partial conversion data.
Structure consolidation
Narrow ad sets merged so budget concentrates and campaigns exit the learning phase, where delivery efficiency is at its worst.
Creative production
A standing calendar of modular assets — varied hooks, formats and angles — produced at volume rather than commissioned reactively when results dip.
Testing framework
Structured creative tests with defined win conditions, sample sizes and retirement rules, so results are decisions rather than opinions.
Fatigue management
Frequency and decay curves monitored per asset, with replacement triggered before costs climb rather than after they already have.
Incrementality testing
Geo holdouts on prospecting and retargeting to measure genuine lift, because attribution windows credit Meta for conversions it did not cause.
We produce creative, not just manage settings
On Meta, the account settings are quickly optimal and then stay there. Creative is where performance is actually won or lost.
Creative at volume
Modular production so dozens of variations ship monthly without a full shoot each time. Volume is what the delivery system needs to work.
Conversions API done properly
Server-side events with deduplication and high event match quality, so optimisation runs on complete data instead of fragments.
Incrementality over attribution
We run geo holdouts. Attribution windows credit Meta for conversions it did not cause, and holdouts are the only way to know the difference.
Fatigue caught early
Decay curves monitored per asset so creative is replaced before costs climb, rather than after a month of paying more for less.
We will tell you to cut retargeting
If holdouts show it is not incremental, we say so — even though it reduces the budget we manage and worsens reported ROAS.
Planned with other channels
Meta budgeted against blended efficiency rather than platform ROAS, so it complements search rather than competing for the same credit.
Why Businesses Choose SDM for Meta Ads
Same budget, very different outcomes. Point by point, here’s how a specialist partner compares to a typical agency or going it alone.
| What actually matters | With SDM | Typical Agency | In-House / DIY |
|---|---|---|---|
| Senior specialist on your account | Always | Often a junior | Stretched thin |
| Custom strategy built for your goals | Tailored | Templated | Guesswork |
| Deep audit before any work begins | ✓ | Surface-level | Skipped |
| White-hat, penalty-safe methods | Guaranteed | Varies | High risk |
| Plain-English reporting tied to revenue | Monthly | Jargon PDFs | None |
| Direct access to your specialist | ✓ | Account-manager relay | N/A |
| Creative produced, not just managed | ✓ | Client supplies | Ad-hoc |
| Conversions API configured | ✓ | Pixel only | Default |
| Structure consolidated | ✓ | Fragmented | Fragmented |
| Fatigue monitored per asset | ✓ | Reactive | Unnoticed |
| Incrementality holdouts run | ✓ | Platform ROAS | Never |
| Ongoing competitor gap analysis | ✓ | One-off | Manual |
| Conversion-focused, not just traffic | ✓ | Traffic-first | Unclear |
| Premium tools included (Ahrefs, SEMrush) | ✓ | Sometimes | Costly extra |
| No long lock-in contracts | Flexible | 6–12 mo lock-in | N/A |
| Established agency, operating since 2017 | ✓ | Varies | Learning curve |
| Fast onboarding & early quick wins | ~2 weeks | Slow | Trial & error |
| Human, SEO-led content (no AI spam) | ✓ | Outsourced / AI spam | Time-heavy |
| Focus on compounding, long-term ROI | Core promise | Short-term wins | Slow & ad-hoc |
| Recovery from Google penalties | ✓ | Sometimes | Very hard |
20 reasons growing brands make the switch. See the difference for yourself →
Platform ROAS is a diagnostic, not a result
Meta counts conversions it touched within its attribution window. We report what it actually caused.
Incremental ROAS
Holdout-tested return, which is consistently lower than platform-reported ROAS and consistently more useful.
Blended CPA
Total spend against total new customers from your own systems, so no attribution window can inflate it.
Creative win rate
Share of tested assets that beat control, and how quickly winners decay under frequency.
Frequency & decay
How fast each asset loses efficiency, which sets the replacement cadence rather than guesswork.
New vs returning
How much spend acquires customers versus re-reaching existing ones — where over-investment almost always hides.
Four stages, repeated every month
Fix signal, consolidate structure, then run creative as a production line.
Signal & structure
Conversions API implemented with proper deduplication, then narrow ad sets consolidated so budget concentrates and campaigns exit the learning phase.
Creative baseline
Current assets audited for fatigue, winning angles identified and a modular production system built so variations can ship weekly.
Structured testing
Hooks, formats and angles tested with defined win conditions. Losers retired, winners varied further, learnings documented rather than remembered.
Test lift & scale
Geo holdouts on prospecting and retargeting to establish real incrementality, then budget scaled against marginal CPA.
Meta Ads in action — three sectors
Different models, the same lever: more creative, better signal, honest measurement. The scenarios below are composites drawn from situations we encounter repeatedly — they illustrate method, not the account of any single named client.
The questions we get asked most
Straight answers on creative, targeting and what Meta reporting actually means.
Further reading
Go deeper on this
The thinking behind the work. These guides go deeper on the methods this page describes — written by the team that runs them.
Meta Ads in 2026: Algorithm Changes You Need to Know
Three major Meta algorithm updates hit in Q1 2026. What changed, what it means for ROAS, and the campaign structure fixes we are making.
Read the guide →Social MediaMeta Advantage+ vs Manual Campaigns: The 2026 Playbook
When to use Meta Advantage+ vs manual campaigns in 2026 — how each works, what drives ROAS, and the creative and data inputs that decide performance.
Read the guide →