Facebook Ads isn’t dead. It got more expensive. Average CPM on most verticals has grown 2-3x in the last 4 years. What worked in 2020 (broad targeting, 5 creative variations) burns your budget in the first week today. What matters in 2026 is a different game.
The pattern below comes from campaigns we’ve run for Motorola, Muller, Blindspot and Orient Ceramic. Hundreds of thousands of euros in ads, tested across very different audiences.
Creative is 80% of performance
Meta itself confirms it: in the era of advanced machine learning, the algorithm finds its audience. What it can’t do: save a weak creative. Test aggressively on creative, lightly on targeting.
Minimum 5 creative variations per campaign. Different angles, not just different crops. One emotional, one rational, one social proof, one FOMO, one demo. Run for 7-10 days, cut anything below 1.5% CTR.
Vertical video, not horizontal
90%+ of Meta consumption is on mobile, in vertical feed or Stories/Reels. A 16:9 video with black bars loses 30-40% of attention in the first 2 seconds. Produce natively 9:16, repurpose to 1:1 for classic feed.
Hook in the first 1.5 seconds
Same rule as on TikTok. Same audience, same fingers. If you don’t stop the scroll in the first second and a half, CPM goes up because retention is low. At Muller, the on-screen-text hook variant pulled CPM 38% lower than the audio-only hook variant.
Advantage+ for scale, manual for validation
Advantage+ campaigns work well after you have 50+ pixel conversions and a validated creative pool. For launch or new product, use manual targeting with 2-3 interest stacks. You learn which audience responds, then move to Advantage+ with that signal as seed.
Budget at campaign level (CBO), not ad set
Campaign Budget Optimization automatically distributes to ad sets that perform. On campaigns with 3+ ad sets, CBO drops CAC by 15-25% vs ABO. At Motorola, switching from ABO to CBO took ROAS from 2.1 to 3.4 with no creative change.
Retargeting in 3 layers, not one
Layer 1: 75%+ video viewers from organic or ads (warm, but not converted). Layer 2: website visitors last 14 days. Layer 3: add-to-cart but no purchase, last 7 days.
Each layer gets different creative. Layer 1: social proof + testimonials. Layer 2: offer + urgency. Layer 3: specific reminders on the viewed product. At Orient Ceramic, this approach dropped retargeting CAC by 42%.
UGC outperforms studio production
CPM 30-50% lower, CTR 40-60% higher. Constant pattern across all our verticals. It doesn’t mean you film badly. It means you film authentically: phone, natural light, one person speaking directly to camera.
Correct tracking: CAPI, not just pixel
The pixel loses 20-40% of conversions post-iOS 14. Conversions API (CAPI) recovers the signal server-side. If you don’t have CAPI set up, you’re running with half the data the algorithm sees. Documentation: Meta Conversions API.
What NOT to do
Stop an ad at 24 hours because “it didn’t perform” (learning needs 3-7 days). Change budget daily (resets learning phase). Run audiences under 500k (you’re limiting the algorithm). Copy creative from TikTok without adapting (different context, different mindset).
Facebook Ads in 2026 isn’t about hacks. It’s about good creative, complete tracking and 7 days of patience on every test. Whoever respects these rules runs profitably. The rest blame the platform.