Comment Tester & Scaler Sur Meta Ads Avec Andromeda En 2026 (GUIDE COMPLET)

Business-focused summary (Meta Ads testing + scaling in 2026, “Andromeda”)

Meta Ads is presented as a data-driven statistical system (not a “magic button”). Success comes from:

• reading performance correctly
• managing account/creative “environment” factors
• running a structured testing-to-scaling process to find winners quickly

---

Core framework: “AOS” via 4 levers (“aerometrics”)

The video argues that your performance ceiling is driven by four metrics that combine into your effective efficiency (referred to as AOS):

• CTR (click-through rate)
  • Driven by: hook, visuals, ad creative, and avatar relevance
• Conversion rate (CVR)
  • Driven by: landing page + offer + ability to convert
• AOV (average order value)
  • Driven by: offer structure, bundles, upsells, delivery options, etc.
• CPM
  • Cost to reach 1,000 impressions; influenced by: audience difficulty, account history, creative saturation, business model, etc.

Key operational claim

Small changes to one lever can drastically change overall results. Example described:

• Holding other metrics constant, a CPM drop can change output from something like 2 → 1 (relative AOS/SWR illustration)
• Conversely, small conversion improvements at constant CPM can materially improve AOS

---

Quick diagnosis playbook (when performance is off)

Use metric “symptoms” to identify where the failure is happening:

• High CPM
  • Likely causes: expensive targeting environment, business model, account status, or need to test different ad account environments
  • Examples mentioned:
    • targeting golfers → higher CPM
    • targeting cosmetics in the US with lots of competitors → higher CPM
• Low CTR
  • Likely causes: weak hooks/visuals, low creative relevance to the avatar, poor creative fit
• Low conversion rate
  • Likely causes: landing page/funnel weaknesses OR ad-to-offer mismatch
• Low AOV
  • Likely causes: offer/bundle/upsell design; adding upsells or paid delivery can change results

---

“Hidden environment” factors (account + assets) that impact learning, CPM, stability

The video stresses that Meta is not neutral across accounts—two identical setups can perform differently due to environment.

Environment levers

• Business Manager (BM) cleanliness + stability
  • “Older, cleaner, more stable BM” + fewer violations/rejections → Meta “gives you better CPMs / more breathing room”
  • borderline/violating business histories can increase CPM, reduce learning speed, and limit spend
  • implies Meta internally ranks/assesses BM health (even if not shown)
• Ad account type / agency account vs personal
  • renting an agency account can help if it’s clean and not associated with banned/aggressive activity
  • recommendation: test multiple accounts with the same setup/creators and pick the account with best AOS/ROS (not just best CPM)
• Facebook page trust (“heat up” new pages)
  • Meta prefers pages that look like real entities (not freshly created)
  • Recommendation: either use older/recovered pages or warm up new pages
  • Warm-up timeline: 1–2 weeks, with ~1–2 posts/day
  • Content strategy during warming:
    • post organic engagement content (even from outside the niche) to generate engagement signals quickly
  • “Hot page” claim: less friction → improved CPM/performance at test start

---

Testing process: “CBO with concepts” to find winners fast

Testing goal: treat every creative idea as a hypothesis and discover winners quickly via data (not “prove you’re right”).

Recommended test structure (CBO)

• Use a CBO (campaign budget optimization) with multiple concepts/angles inside the same campaign
• Each concept can include different:
  • marketing angles
  • sometimes avatars
  • creative execution

Test budget guidance (no fixed number)

Examples given:

• $300–$500 often used in CBO testing (but $100 can also work)
• Europe: €100–150 “OK”
• US: €150 may take longer (higher CPMs → slower data)

Rules of thumb:

• higher budget → faster data; lower budget → more days needed
• avoid allocating too much budget over too few days
  • minimum test duration: 3 days
  • reason: smooth out “good vs bad” days

Creative sourcing rule

• Don’t “invent” during testing:
  • test concepts/structures that are proven by competitors
  • don’t copy; use your own content/footage adapted to your brand/product
  • mentions using AI tools to recreate competitor-style prompts/scripts (high level)

---

How to read Meta performance (in 2026): 4 metrics + patterns

Beyond CTR/CVR/AOV/CPM, the video adds monitoring using:

1. Spend patterns (Meta spending treated as a quality signal)
   • analyze: hook, format, person/avatar, script relevance, emotional triggers
2. Frequency
   • target placement:
     • around ~1.1 frequency → mostly top funnel (new audience reach)
     • above 2 → tends toward bottom funnel / retargeting-like effects
3. Cost per Purchase (CPA)—interpreted by funnel role
   • high CPA doesn’t automatically mean bad ads (might indicate top-of-funnel education vs bottom-of-funnel conversion)
   • emphasizes overall CPA at account level matters because Meta aggregates multiple ads
4. CPMR (early warning metric)
   • defined as: CPMR = CPM × frequency
   • described as a leading indicator of future performance deterioration

---

CPMR (saturation predictor) + weekly intervention playbook

Definition

• CPMR = CPM × frequency
• Purpose: measure the “true cost” of reaching new unique people, accounting for repeated exposure

What increasing CPMR means

Rising CPMR indicates:

• higher CPM and/or higher frequency
• Meta is reaching fewer new people → expanding costs more or stagnation
• likely creative diversity/quality issue or budget overspend without enough winning creative to sustain learning

The video describes CPMR as a lagging/leading mix, positioned as early detection before:

• AOS drops
• CPA rises
• ROI declines

What to do when CPMR rises

• Check exclusions
  • ensure buyers are excluded for at least 180 days
  • keep buyer list synced (mentions putting customer list on Clavio)
  • reduce algorithm recycling of existing buyers (retargeting handled via email)
• Consolidate campaigns / reduce competition
  • too many adsets/campaigns → increases CPM
  • recommendation: fewer campaigns per product rather than an extreme number
• Increase creative diversity
  • test new angles/concepts/formats rather than only more iterations of the same asset style
• Expand media mix
  • mention awareness/impression-style campaigns or partnerships to find new audiences at lower CPMs, then feed learnings into conversion tests

---

Account scaling logic: “archetypes” for deciding whether to iterate or scale

Based on spend/frequency/CPM/CPA, the video describes two performance archetypes.

Top-funnel archetype (demand creation)

• Spend up
• Frequency down
• CPM down
• CPA up
• Interpretation: the ad likely creates demand
• Action: keep it and iterate it

Bottom-funnel archetype (conversion/retargeting-like)

• Spend down
• Frequency up
• CPM up
• CPA down
• Interpretation: stronger ROS but more retargeting behavior / less new traffic
• Action: scale carefully; create similar/better “ATS” (implied winning assets)

Important constraint

If CPMR rises too much, stop iterating that ad and shift iteration approach (more diversity / different creative directions).

---

“Andromeda” (Meta Ads update) implications for creative iteration

Andromeda is presented as a major shift focusing on:

• more creative diversity
• finer reading of user signals
• requiring more “professional” creative iteration strategy

Creative identity mechanism (key operational warning)

Meta uses creative IDs and entity IDs to group similar ads. If changes are superficial, Meta may treat them as the same entity, reducing incremental learning, e.g.:

• same visual style + similar lighting/setting
• same format but different text/music/creator
• identical “first 3 seconds” hook → may get grouped

Consequence: less incremental audience discovery, slower scaling.

How to create “new entities”

To trigger new creative/entity grouping, diversify along multiple dimensions:

• Change format (e.g., cartoon vs UGC vs explanatory visuals)
• Change vehicle/hook
• Change character
• Change setting/location (outside vs room vs office)
• ensure hooks in the first 3 seconds are genuinely different (avoid the same opening)

---

Three-stage “Andromeda testing process” (concept → analyze → iterate)

1. Define the perfect match persona
   • character identity, what they want, frustrations/fears
   • message that resonates with that persona
2. Analyze what Meta is pushing (by spend/data)
   • determine which format/message/avatar is getting attention
3. Iterate winning patterns more aggressively
   • if the message is winning → test other formats/avatars
   • if the format works → test other messages/people
   • aim for concept-level iteration until winners emerge

Iteration principle contrast vs 2025/earlier

• Prior approach (implied): micro-changes like titles could create new results
• Now: larger creative differentiation changes are needed so Meta doesn’t treat iterations as the same entity

---

Concrete examples and actionable tactics mentioned

• Account testing tactic
  • run the same setup across 2–3 ad accounts
  • choose the one with best long-term AOS/ROS (not just CPM)
• Page warming
  • 1–2 posts/day for 1–2 weeks before ad scaling
• Competitor validation
  • use competitor concepts/scripts as a starting point, but recreate using own footage
• Creative organization
  • separate ads by funnel stage and by asset type:
    • top/mid/bottom funnel buckets
    • separate static vs video to help Meta read data faster
• Adset sizing rule
  • keep ≤ 15–20 ads per adset
• Minimum spend “hack” (suggested)
  • set a minimum setup spend (example $15) so Meta must allocate time to assess adsets
• Buyer exclusion rule
  • exclude purchases for 180 days
  • treat retargeting via email, not Meta audiences

---

Metrics / KPIs referenced or implied

• CTR, Conversion Rate, AOV, CPM
• AOS / ROS
• CPA (and overall account CPA)
• Frequency
• CPMR = CPM × frequency (early saturation indicator)
• Weekly monitoring: CPMR trend, ROS/CPA/AOS direction

Targets / timelines mentioned

• Test duration: minimum 3 days
• Page warm-up: 1–2 weeks
• Exclusion window: 180 days buyers

---

Presenter / source(s)

• Speaker referenced as Matthéo (“il y a dit… c’était Matthéo.”)