I Build Growth Systems That Scale Revenue, Not Just Campaigns
8+ years across US, India & Europe | $18M+ media managed | 127% YoY growth delivered | Ivey MBA '26
Purplle: Engineering a Unicorn's Growth Engine
The Challenge
"Joined during COVID hyper-growth with aggressive quarterly targets from foreign investors. The fundamental challenge wasn't performance—it was building scalable infrastructure where none existed."
No Historical Data
Digital marketing only 4 months old when I joined
High Team Churn
Average tenure <8 months (I became longest-retained)
Aggressive Sales Calendar
Multiple major + frequent minor sales created volatility
Learning Pollution
Mixing easy retention with hard acquisition corrupted ML
Strategic Diagnosis
"The core question wasn't 'how do we optimize?' but 'how do we build a system that can scale aggressively across unpredictable sales cycles without destroying learning?'"
Key Insight 1: Cohort Behavior Pattern
Analysis revealed that ~70% of repeat buyers converted again within 30 days, largely driven by heavy sales cadence. Beyond this window, most RB came through organic/direct channels.
Conclusion: Paid marketing should focus on First Transactions (FT),
not blended LTV optimization.
Key Insight 2: Signal Pollution Discovery
"Meta's ML internalizes account-level history, not just campaign-level signals. Mixing objectives within one account made the algorithm 'lazy'—it optimized for easy wins (retention) instead of hard acquisition."
Key Strategic Decisions
FT-First Strategy Instead of LTV Blended Optimization
Alternatives Considered
Standard e-commerce practice would optimize for blended LTV—mixing first-time buyers and repeat buyers into a single funnel, targeting users most likely to generate high lifetime value.
Why This Won
Cohort analysis revealed that ~70% of repeat buyers converted again within 30 days, largely driven by Purplle's heavy sales cadence and offer density—not incremental paid exposure. Beyond 30 days, most repeat purchases came through organic and direct channels anyway.
This meant paid marketing was not the driver of repeat purchase—the product and sales calendar were. Optimizing for LTV with paid media would inflate CAC without improving outcomes.
FT-First strategy:
- Avoided distorted blended CAC averages
- Allowed higher CAC tolerance for genuine new user acquisition
- Gave CXO leadership clearer visibility into true acquisition efficiency
- Naturally fed the repeat buyer base through aggressive new user inflow
Multi-Account Architecture Instead of Single-Account Campaign Segmentation
Alternatives Considered
Keeping everything inside one Meta account, separated at the campaign level—a faster, simpler setup that most accounts use by default.
Why This Won
Meta's machine learning internalizes account-level behavioral history, not just campaign-level signals. Mixing objectives (app installs, first transactions, retention) inside one account causes the algorithm to learn a blended, conflicted objective—optimizing for the easiest wins (retention) rather than the hardest ones (cold acquisition).
Account-level separation meant:
- Each account's ML learned one pure objective
- Failures in one segment were fully isolated from others
- Recovery from volatility took 2-3 days instead of 2-3 weeks
- Aggressive scaling in one segment could not destabilize others
This was a hard-to-reverse structural decision—which is exactly why it had to be made correctly at the start.
Structure Before Experimentation—Accepting Slower Early Momentum for Long-Term Control
Alternatives Considered
Launch campaigns immediately, optimize week-to-week, and build structure reactively as the account grew. Faster short-term results, but no architectural foundation.
Why This Won
A reactive account—optimized week-to-week without clean signals—would not survive Purplle's scale ambitions. Any structural flaw would be amplified at $500K/month, not corrected.
By investing 2 weeks in information gathering, stakeholder alignment, and system design before launching, the account gained:
- Stable learning environments from day one
- Clear levers for scaling and de-scaling independently
- The ability to experiment safely at account, campaign, asset, and catalog levels
- A system that was dynamic without being fragile
The short-term cost was slower ramp-up. The long-term gain was an architecture that handled 400% budget surges without breaking.
Strategic Solution: Multi-Account Architecture
❌ Before: Single Account Chaos
Single Meta Account
↓
App Install
FT - App
FT - Web
Retention
↓
Result: Learning conflicts, scale volatility, recovery takes weeks
✅ After: Multi-Account Architecture
Account 1
App Install
Cold traffic, downloads
Account 2
FT - App
First transaction, app users
Account 3
FT - Web
First transaction, web users
Account 4
Retention
Repeat purchasesResult: Isolated learning, independent scaling, fast recovery (2-3 days)
Why Multi-Account Architecture Won
| Metric | Before (Single Account) | After (Multi-Account) |
|---|---|---|
| Learning Stability | ❌ Volatile | ✅ Stable |
| Scale Capacity | ❌ Breaks at 200% | ✅ Handles 400% |
| Recovery Time | ❌ 2-3 weeks | ✅ 2-3 days |
| CAC Control | ❌ Unpredictable | ✅ Predictable |
Multi-Account KPI Segmentation → Unified Macro System
1
App Install
Cold traffic acquisition
Download optimization
2
FT - App
First-time buyers
App conversion focus
3
FT - Web
Web first transactions
Channel-specific optimization
4
Retention
Re-engagement campaigns
Value optimization
→
Unified Growth System
✓ Isolated learning per KPI
✓ Cross-account budget allocation
✓ Shared creative intelligence
✓ Macro CAC stability
✓ Sustainable 400% scale capacity
The Stress Test: I ❤️ BEAUTY (IHB) Sale
Challenge: Mega sale requiring $100K spend in 5-6 days (400% budget surge)
$100K
Spent in 6 days
35%
CAC Reduction
3-4×
Transaction Volume
Zero
Learning Resets
Architect's Note
"Scale isn't about spending more. It's about breaking less while spending more. This structure didn't just handle growth—it enabled strategic expansion into new verticals."
Organizational Impact & Recognition
In-House Brands Launch
Enabled launch of 7 in-house brands (higher margin) with isolated systems
Elite Segment
Built compliance-first system for international brands entering India
Leadership Continuity
Longest-retained marketer during high-churn period
Superstar Award
Organization-wide recognition for exceptional impact (April 2021)
Quantified Business Impact
400%
Budget surge capacity
35%
CAC reduction at scale
2-3 days
Recovery time (vs 2-3 weeks)
Zero
Account rebuilds required
7
In-house brands launched
MCube
Awards nomination
Elite Segment: Launching International Brands Into India
When Purplle launched its Elite segment—featuring international beauty and cosmetics brands entering the Indian market—the challenge was fundamentally different from everything else on the platform.
These brands came with:
- Strict global compliance policies and brand safety requirements
- Specific creative guidelines that could not be deviated from
- Entry-into-India positioning that required cultural sensitivity
- No existing India performance benchmarks to reference
This was not a performance marketing problem. It was a brand architecture problem with a performance layer.
My approach:
- Worked directly with internal content, design, and legal teams alongside global brand stakeholders
- Implemented Meta Collaboration Ads for co-branded campaigns—a feature that was new at the time with no established playbooks
- Benchmarked each international brand against the closest internal Purplle equivalent to set realistic early expectations
- Shifted reporting from short-term ROAS metrics to 3-6 month brand health roadmaps, giving brands the long-term visibility they needed to commit to the Indian market
This balanced strict compliance requirements with the business need for performance—and established Purplle as a credible platform for premium international brands to enter India through.
Key Learnings & Principles
01
"Separate intent before optimizing efficiency"
Don't mix easy wins with hard challenges in the same system
02
"Treat learning as a first-class asset"
Protect ML signals with same rigor as budget
03
"Design for volatility, not averages"
Build systems that survive 400% surges, not just steady state
04
"Structure creates optionality"
Good architecture enables experimentation without risk
ThruPlay Optimization
