In July 2024 Meta released a major algorithm update called Andromeda. Its practical effect is simple but profound: rather than identifying one single “best” ad and pouring all budget into it, Meta now treats your creative portfolio like a pool of options to be matched to specific audience segments and customer journeys. One generic hook no longer wins. Meta wants true creative diversity — and it routes distinct creatives to the micro-segments most likely to respond.
Why this matters now
- Sudden performance drops? If previously winning ads cratered after July, it’s probably not a bug — it’s this rule change.
- The algorithm now values a range of creative signals across formats, angles and personas, then routes each creative to the audience segment most likely to respond.
- Many advertisers are still running the old playbook: a handful of similar ads, minor variations, and complicated targeting. That approach underperforms in the new environment.
The core shift (high level)
Old model
HOT
- Identify a top-performing creative and allocate scale to that single piece.
New model
- Algorithmic matchmaking of many distinct creatives to granular audience segments and stages of the funnel. The system optimizes across the portfolio rather than concentrating on one ad.
Result
- Creative diversity becomes the principal lever of performance.
- Campaign structure should be simpler; creative complexity should increase.
The biggest mistake advertisers are making
- Producing 10 near-identical variants (same video, same message, small hook changes). Meta treats these as the same creative cluster. They don’t reach different niches and don’t give the algorithm meaningful alternatives to match.
What works instead
Principle: create fundamentally different ad concepts — not tweaks. Each concept should differ in at least one major dimension: format, angle, or persona.
Example for a productivity app (concepts)
- Screen-record product demo showing a 9–5 workflow.
- Customer testimonial about recovering 10 hours per week, told in their voice.
- Static comparison graphic: app vs. spreadsheets.
- Founders’ explainer: why most productivity systems fail.
- Animated 3-step process explainer.
- Quick “day-in-the-life” vignette showing real outcomes.
Each concept looks different, targets different motivations, and will be routed to different audiences.
The three dimensions of creative diversity
- Format — how it looks
- UGC/testimonial, product demo, static image/carousel, talking head, screen recording, animation, interactive story.
- Visual framing, pacing and sound usage matter; mobile-first edits and captions are required.
- Angle — what you say
- Pain, aspiration, social proof, contrast/comparison, education, scarcity/urgency, storytelling.
- Different angles appeal at different funnel stages and psychographics.
- Persona — who you’re talking to
- New users vs. power users; budget buyers vs. premium buyers; industry-specific vs. general.
- Messaging, tone, and proof points must change with persona.
Combine format + angle + persona to produce genuinely distinct concepts.
Campaign structure: simplify, centralize creative complexity
- Prospecting: one CBO prospecting campaign with broad targeting. Place 8–15 truly different ad concepts into the campaign. Let Meta match creatives to micro-segments.
- Budget baseline: start small and reliable — $30–50/day for small accounts; $100–300/day for mid-sized; scale as performance warrants.
- Retargeting: separate retargeting campaign at roughly 15–25% of total ad spend targeted to warm audiences (video viewers, website visitors, app engagers). Use sequential messaging and higher-intent creatives.
- Learning window: give campaigns 2–3 weeks before major changes. The algorithm evaluates longer customer journeys now — early fluctuations are expected.
- Complexity lives in creative variety, not ad set proliferation. Fewer ad sets + more truly different creatives = better results under Andromeda.
How to evaluate creatives (rules of engagement)
- Minimum observation period: 2–3 weeks.
- Minimum data thresholds: aim for several hundred impressions per creative and meaningful conversions where applicable; if conversion volume is low, use leading indicators (CTR, view-through rate, add-to-carts, landing page engagement).
- Don’t kill creative on day 3 for underperformance. Let the algorithm allocate traffic and find matches.
- Replace or iterate only after enough exposure and stable metrics. If a creative never reaches break-even CTR or engagement after the learning window, retire and replace it.
Practical testing workflow (repeatable)
- Audit current creatives. Mark each as: fundamentally different concept vs. minor variant.
- Plan 8–15 new concepts if most are variants. For each: define format, angle, target persona, CTA, and assets required.
- Produce and launch in one prospecting campaign with broad targeting. Include 2–3 retargeting creatives in a separate campaign.
- Monitor weekly: engagement rates, CPM, CTR, CVR, CPA/ROAS, and creative-level breakdowns.
- After 2–3 weeks, identify top performers (by ROAS/CPA or proxy metrics), scale those concepts and create 2–3 iterations that remain substantively different.
- Repeat cycle: refresh portfolio every 4–8 weeks to avoid ad fatigue and capture shifting niches.
Creative brief template (use for each concept)
- Objective: acquisition / awareness / retention
- Target persona: demographics, behaviors, primary pain, decision trigger
- Single-minded proposition: one sentence the ad must communicate
- Format & length: vertical 15s/30s, static, carousel, UGC, etc.
- Opening hook (0–3s): attention trigger
- Core message (3–20s): proof, demo, story
- CTA: explicit, simple (install, sign up, learn more)
- Must-have assets: logo, screenshots, captions, testimonial clips
- KPI to evaluate: CTR/engagement for awareness; CVR/CPA/ROAS for conversion
Sample concept frameworks for other industries
- E-commerce apparel:
- Outfit transition/fashion hack video
- Customer UGC before/after styling
- Comparison (fabric, durability) with competitors
- Mini-lookbook (seasonal edit)
- Behind-the-scenes production story (ethical/quality proof)
- B2B SaaS (lead gen):
- Quick demo of a specific workflow improvement
- Customer case study with quantified outcomes
- Industry pain explainer: common cost of current process
- ROI calculator teaser driving to landing page
- Founder explains product philosophy
- DTC skincare:
- Clinical summary by dermatologist (UGC-style)
- Visual before/after with timeline
- Ingredient spotlight (why it works)
- Routine tutorial (how to use with other products)
- Customer testimonial focused on a specific skin type
- Local service lead gen:
- Day-in-the-life service visit (short doc style)
- Client testimonial about time saved/pain eliminated
- Pricing transparency/guarantee creative
- Before/after transformations with captions
- Local credibility: community/awards/associations
Measurement and attribution considerations
- Attribution windows: align reporting with your backend (common: 1-day click, 7-day click, 7-day click + 1-day view). Be consistent when comparing periods.
- Creative-level reporting: use Meta breakdowns and UTM tracking to attribute on-site behavior to creatives. Tag creatives to run cohort analysis in analytics/CRM.
- Incrementality: schedule periodic lift tests for big decisions. True winners are those that produce incremental conversions, not just cheaper last-click metrics.
Common pitfalls and how to avoid them
- Mistake: producing micro-variants and thinking that equals diversity. Fix: aim for radical difference across format/angle/persona.
- Mistake: excessive ad set fragmentation. Fix: consolidate targeting and rely on creative diversity to reach segments.
- Mistake: pausing creative too quickly. Fix: allow 2–3 weeks and minimum data thresholds.
- Mistake: ignoring mobile-first rules (sound-off viewing, subtitles). Fix: include captions and design for vertical viewing.
- Mistake: not tracking creative-to-landing engagement. Fix: use UTMs and landing page experiments to ensure the creative’s promise converts.
A practical budget and timeline example (small SaaS)
- Week 0: Audit and plan 8 concepts. Produce content (simple shoots, screen recordings, UGC requests).
- Week 1: Launch one prospecting campaign with CBO, $50/day; retargeting campaign $10/day.
- Weeks 1–3: Monitor KPIs, don’t make structural changes. Replace only broken assets (technical issues).
- Week 3: Evaluate creatives using CPA/ROAS and leading engagement metrics. Keep top 3–5 and create 2 iterations of each. Increase budget on winners.
- Weeks 4–8: Continue iteration and start scaling budget for top performers.
Immediate action plan (this week)
- Audit your library: mark which ads are truly different vs. minor variants.
- If mostly variants: plan 5–8 fundamentally different concepts (mix format + angle + persona).
- Simplify campaigns: migrate to one prospecting campaign (broad) and one retargeting campaign. Stop micro-targeted ad set proliferation.
- Launch and let run 2–3 weeks. Track engagement and conversion proxies.
- After the learning window, scale winners and iterate with new concepts.
Final note
Andromeda makes creative the central competitive advantage. Campaign setup should be simpler; creative briefing and production must be smarter and more varied. Deliver truly different ads across format, angle, and persona — then give the algorithm time to do its matching.