Andromeda Nearly Killed This SaaS.
I Adapted And Cut CPL by 78%.
Meta’s Andromeda update destroyed most brands’ ad economics overnight. I leaned into creative velocity, built a system that feeds the algorithm what it wants, and turned a CPL crisis into a 78% cost reduction over 32 weeks.
0
Creatives
0x
Creative Velocity
0%
CPL Reduction
0
Weeks
HyperTask
B2B productivity SaaS — project and task management with Kanban, list, and calendar views. Brought me in July 2025 to run Meta Ads for cold-audience trial acquisition. No pre-existing paid pipeline. Goal: generate qualified signups at a CPL the unit economics could sustain.
Product
B2B SaaS · Productivity
Goal
Cold-acquisition trials
Engagement
32 weeks · Jul 2025 → Mar 2026
Andromeda nearly killed this SaaS.
I had an ad account producing ~$46 CPL like clockwork. Then Meta rolled out Andromeda — their ML-driven distribution model. Overnight, targeting that worked for years stopped working. CPLs jumped 6–20×. Most brands froze. For a small SaaS on a tight CAC window, the math stops working fast.
The Collapse · Aug 2025 → Dec 2025
Stable $46 CPL. Then Andromeda.
$46
Pre-Andromeda CPL
$936
Peak CPL
20×
Cost Spike
I did the opposite of what most people did.
Andromeda punishes brands that run few hand-picked creatives against narrow audiences. It rewards brands that feed it volume and let it find the signal. So I stopped picking winners and started flooding the feed — five creatives a week became sixty. Then I built a machine around it.
The operating system behind the volume
Sixty creatives a week doesn't happen by accident. I built four repeatable processes to produce, test, and scale — without burning out.
Weekly RADAR Meetings
41 meetings over 9.5 months. Every week I reviewed data, set priorities, and adjusted strategy in real-time. Relentless, disciplined iteration.
Systematic A/B Testing
Every angle, landing page, and creative got measured. I found that simple pages outperformed complex ones, and pain-point hooks beat feature lists.
Audience Refinement
Started broad, narrowed by device (mobile beat desktop), refined by behavior signals. Let Meta's algorithm learn from clean conversion data.
Budget Discipline
Scaled spend only after validation. Best weeks ran on $395-$526 budgets, not on brute-force spending. Efficiency over volume.
667 unique creatives in 32 weeks
Every ad I produced, analyzed by AI for hook type, visual style, and emotional tone. The raw material that fed Andromeda.
0
Creatives Produced · Jul 2025 → Apr 2026
Hook Strategy
Visual Style
Tone
Three winning angles, isolated by data
Out of 20+ positioning angles and hundreds of hook variations, only three survived. Each was isolated, validated, and scaled through a repeatable testing process — not gut instinct.
Volume over guesswork
20+ angles across hundreds of hook variations. I didn't try to pick the winner — I let spend, clicks, and signups pick it for me.
20+
angles tested
Data-backed winner selection
An angle only earned 'winner' status once it delivered a measurable outcome: a paying subscriber, a 4x signup lift, or the lowest cost-per-signup in the account.
3
angles confirmed
Concentrate budget on winners
Once validated, losing angles were cut and budget consolidated into the three winners. Best week ran on $395 spend — efficiency, not brute force.
$6.17
best CPL
Landing Page Breakthrough — Dec 2025
After ~20 iterations, a simpler page 4x'd signup volume overnight. Complexity was killing conversions.
When the funnel needed to evolve
26 weeks in, the signup funnel hit a ceiling. I rebuilt the entire front end around a meetings funnel — and the CPL curve flipped overnight.
Phase 1
Self-Serve Signup
Phase 2
Sales-Led Meetings
Andromeda came. I changed everything.
Same budget. Same product. The only variable was creative velocity. CPL dropped 78%, signups scaled 162%, and the scissors closed.
The Scissors Effect · Aug 2025 → Mar 2026
0x creative velocity. 0% CPL drop.
12x
Creative Velocity
$45 → $10
CPL Descent
667
Total Creatives
32
Weeks
0
Creatives Produced
0%
CPL Reduction
0x
Velocity Increase
0
Weeks
5 → 60
Creative Velocity
From 5 creatives per week to 60. Each new asset gave Andromeda more signal to find high-converting audiences — compounding returns on every creative produced.
$45 → $10
Cost Per Lead
CPL dropped 78% as creative velocity scaled. Not from budget cuts — spend stayed flat at ~$800/wk. Pure efficiency gains from feeding the algorithm more creative signal.
667
Total Creatives
A creative engine that produces, tests, and iterates at scale. Winning hooks, angles, and landing pages isolated by data — not gut instinct. The system learns what converts.
The brands still chasing the old playbook — narrow audiences, small creative sets, manual optimization — are still bleeding. Meta's new algorithm favors volume. If you can produce it, you win.
From first call to optimized funnel
41 weekly strategy meetings. Continuous iteration across creatives, funnels, and messaging — toward profitable acquisition.
- Jul 2025First call & kickoff42-week engagement starts
- Aug 2025Meta Ads live$792 spend · 7 signups
- Oct 2025Desktop → mobile audience fixCPM normalized
- Nov 2025Andromeda adaptation complete — creative system stabilizedCPL normalizing
- Dec 2025Landing page simplified — 4x signup lift47 signups / wk peak
- Jan 2026Three winning angles isolatedValidated by data, not gut
- Feb 2026Funnel pivot — signup → meetingsArchitecture rebuiltPivot
- Mar 2026$6.17 CPL — lowest in account history64 signups · $395 spent
41 Weekly Strategy Calls
Real snapshots from my RADAR meetings. Every week for 9.5 months, I reviewed performance, adjusted strategy, and pushed toward better numbers.
Weekly RADAR: reviewing campaign performance and planning iterations
Feb 2026
Analyzing winner ads in Canva — identifying what hooks converted
Mar 2026
Nask AI dashboard: real-time performance tracking
Apr 2026
Want me to build your Andromeda machine?
Same system, your product. I build the creative engine, run the weekly RADARs, and let the algorithm do what it does best — find your buyers at the lowest possible cost.
Book a Discovery CallThis case study covers a 32-week engagement from August 2025 to March 2026. All metrics are from Meta Ads Manager and internal tracking. Client details shared with permission.