How to Scale Facebook Ads on a Low Budget (Post-Andromeda)

Key business outcomes & thesis

• Meta’s “Andromeda” creative-learning loop works best when the account has enough budget to let multiple creatives gather meaningful data.
• For small budgets (e.g., ~$30/day or <$100/day), common advice—“launch many creatives and let Meta find winners”—often fails because most ads don’t receive enough spend/impressions to prove performance.
• The proposed playbook is to: identify winners early → concentrate spend into them → test landing pages/audiences after creative wins → fix post-click funnel bottlenecks. The goal is scaling efficiency (profit per ad dollar), not just spend growth.

---

Frameworks / playbooks emphasized

Andromeda-aware creative testing (CT = Creative Test)

• Use CBO (campaign budget at campaign level) to avoid turning targeting/audience into another variable.
• Launch many creatives initially (e.g., 16), but manually deactivate losing ads once they hit a minimum data threshold to prevent budget “dilution.”

Two-step optimization loop

1. CT (Creative Test)
   • Score ads using engagement + funnel metrics to pick 2–5 winners.
2. LT (Landing Test)
   • Keep the ad creative constant.
   • Test landing page URL variants using multiple ad sets.

Then simplify and scale using more appropriate campaign types:

• ASC (Advantage Sales)
• Interest-based targeting (to help the algorithm learn where the persona is while still leveraging Meta’s optimization)

Bottleneck analysis in the funnel

• Diagnose where traffic leaks occur across the funnel stages: content views → add-to-cart → checkout → purchase

• Fix site/product/UX/payment/shipping friction so Meta needs fewer dollars to generate more sales.

---

Key metrics & KPIs mentioned (and how used)

Creative test scorecard metrics

• Total CTR (example: 6.56% deemed healthy)
• Link CTR (example range: ~2% vs ~6%; described as ~3x engagement difference)
• CPM
• Cost per Page View
• Cost per Purchase
• ROAS
• Funnel supporting metrics by ad:
  • Add-to-cart rate
  • Checkouts initiated
  • Total purchases
  • Which ad generated the highest ROAS

Data thresholds / testing volume targets

• Minimum learning target: ~1,000 impressions absolute minimum (treated as a “fleeting glimpse”)
• Preferred: ~2,000–3,000 impressions total per ad (or per creative before deactivation) for small accounts
• Practical budgeting logic:
  • For ~$30/day with CPM ~ $28–$30, running 16 ads may take too long (≈ 16 days), so impression thresholds become the lever to adjust.

Scaling / funnel bottleneck examples

• ADVT checkpoint referenced: Content view → add-to-cart should be ≥ 10%

• Explicit funnel checkpoint example:

  • 3,600 content views → 177 additions to cart (below 10% limit) → identified as the #1 bottleneck

• Checkout friction rule of thumb:

  • “I don’t like to see a drop greater than 50% for checkout process started”
  • They observed >50% drop at checkout stage (second bottleneck)

---

Concrete examples / case studies & results

Case 1: Brand ~ $6M/year revenue

• Before/after: with the same ad spend, revenue rose +45% after reorganizing spend toward proven creatives.
• Starting point: 100+ creatives
• Problem: budget was spread across many assets; winners received only a small fraction.
• Intervention:
  • Reduced variety to ~30 best ads (from 100+)
  • Focused spend on assets already proven
• Result:
  • Revenue increased 45% without increasing ad spend
  • Over 15 more months, grew to $35M (~5x to 7–8 digit scale)
• Additional operational impact:
  • Freed time for product/expansion/new regions because performance wasn’t “just survival on ads.”

Case 2: Brand growth with same ad spend (efficiency focus)

• One business grew +33% to $2.3M YoY with the same advertising spend, attributed to directing dollars to more efficient assets.
• Another smaller account:
  • Around $500k/year
  • Growth ~200% with similar ad spend

Case 3: Landing page test (AOV + conversion)

• LT test kept the same winning ad creative, but changed landing page URLs:
  • Homepage
  • Collections page
  • Product page (smaller variant)
  • Product page (larger variant)
  • Featured/credibility page
  • Persona-specific dedicated landing page addressing pain points
• Result:
  • One landing page set produced no sales (possibly due to perceived price deterrence)
  • The dedicated/pain-point landing page improved funnel progression and produced more sales vs the original product page
• Scaling implication:
  • They argue that defaulting to sending everyone to the homepage can lead to lower cart adds, purchases, and ROAS.

Case 4: “Within first month” performance + segmentation

• CT spending: ~$500 total
• Early return: ~6.7x ROAS
• Segmentation via interest audiences:
  • Outdoor audiences (climbing/mountaineering/camping) example
  • Reported ROAS improvements up to ~16x, later described as 8x in segmented comparison
• Important caveat:
  • Performance scaling isn’t assumed to be linear; they use controlled increases.

---

Actionable recommendations (what to do)

1) For small budgets, stop “feeding Andromeda with 20–30 new ads”

• If you run many creatives but most get pennies:
  • Underexposed ads may not generate meaningful signals
  • You can miss “gold nuggets” that only win when given enough volume
• Instead:
  • Run a creative test with many ads, but deactivate manually once ads hit ~2,000–3,000 impressions (or at least ~1,000) so budget concentrates into remaining candidates.

2) Use a “winner threshold,” not just “winner by early ROAS”

• Winners should be based on reaching impression thresholds and producing downstream behavior (adds to cart / purchases), not only a one-off spike.

3) After creative winners, test only landing pages (don’t change creative + URL together)

• LT structure:
  • Same winning ad creative
  • Multiple ad sets differing only by landing page/URL
  • Equal budgets per ad set to compare landing performance fairly

4) Scale winners with simplified campaign architecture

• Reduce complexity:
  • Fewer campaigns
  • Fewer ad sets
  • Fewer active ads (concentrate signal + budget)
• Use:
  • ASC (Advantage Sales)
  • Interest-based targeting campaigns
• Guidance:
  • Feed the algorithm 3–4 winning creatives + the correct tested landing URLs, rather than asking it to discover everything from scratch.

5) Improve “post-click economy” by fixing funnel bottlenecks (often cheaper than new ads)

• Audit drop-offs:
  • Content views → add-to-cart
  • Add-to-cart → checkout start
  • Checkout → purchase
• Concrete site changes suggested:
  • Simplify overly complex product descriptions
  • Remove unnecessary images that disrupt UX
  • Make the “Add to cart” button more visually distinct (e.g., button color)
  • Consider shipping-fee friction; offer shipping discounts in retargeting
• Goal:
  • Make Meta convert more of the traffic you already paid for, improving ROAS without increasing ad spend.

---

Targets / timelines called out

• First 30 days: CT + initial learnings; improving efficiency over sequential tests.
• Creative test duration: about a week (depending on reaching 2,000–3,000 impressions per ad before deactivation)
• Decision cadence:
  • Identify winners/losers quickly enough to not bleed budget while still giving winning assets enough exposure.

---

Presenter(s) / source(s)

• Presented by a single speaker (not explicitly named in the subtitles).