Guardrail metrics: Cohorts, fraud tolerance, and analysis

Anthony Scarpaci, former Global VP of Growth at Acorns and senior leader at NerdWallet, Betterment, and Blue Apron, joins Jacob Rushfinn (CEO of Botsi) to break down how to build a referral program that performs. He shares his RIGHTT Framework—Relevance, Incentives, Guardrails, Human Centricity, Timing & Tracking—and real examples from fintech, meal kits, and subscription apps.

🧩 The RIGHT Framework

R = Relevance – Incentives should align with your product’s core value. Cash isn’t always king.

Example: GoHunt gives gear credits usable in-app and in its e-commerce store, keeping rewards tied to the customer experience.

I = Incentives – Make them motivating and credible. Urgency (limited-time offers) beats evergreen “set-and-forget” bonuses.

• Consumers are numb to “Give $10 Get $10.”
• Guaranteed rewards outperform sweepstakes—people act when they know they’ll get something.
• Tie incentives to meaningful product actions that predict retention.

G = Guardrails – Prevent gaming and fraud without killing usability.

The “optimal level of fraud is not zero.”
Every layer of anti-fraud friction hurts good users—accept some inefficiency for total-program scale.

• Analyze cohorts for retention / LTV gaps.
• Require real product usage (e.g., multiple deliveries in meal kits).

H = Human Centricity – Consistent, authentic, transparent experience across the entire journey.

• Map every touchpoint (ads → onboarding → referral share → reward delivery).
• Reinforce trust (“Your friend invited you”) and celebrate wins (“You earned $10—share again”).

T = Timing & Tracking –

• Launch after product-market fit and a healthy customer base.
• Introduce referral prompts at the right emotional moment: trial start or delight milestone.
• Maintain urgency windows for bursts of activity.
• Track cohorts, incremental lift, and blended CAC pre- / post-launch.

💡 Key Insights & Takeaways

• Referrals ≠ free users. Model unit economics and compare to your next-best acquisition channel (Meta, Google etc.).
• Halo & Cannibalization. Account for organic word-of-mouth you’d get anyway and the extra reach you gain when offers go viral.
• Accept some fraud. Zero-fraud programs over-optimize and add friction; “tolerable inefficiency” is a healthy cost of growth.
• Design for compounding. Great referrals create chains (friend → friend → friend), not single invites.
• Avoid conditioning. Don’t train users to expect giant promos forever—treat large bonuses as events, not defaults.
• Influencers as fuel. One creator’s post can 10× signups—plan for the viral halo but don’t depend on it.
• Higher-quality leads. Referred users retain better and cost less long-term—social proof raises both acquisition and retention.

🧠 AI Toolbox Anthony Uses

•  Lovable / v0.dev / Replit V0 → No-code prototyping & mockups.
•  Gemini transcription + Claude / ChatGPT → Strategy alignment & theme extraction from founder calls.
•  OpusClip → Video editing & social creative velocity.
•  Perplexity → Everyday research & voice-based learning.

🔗 Links & Resources

Anthony Scarpaci → https://www.linkedin.com/in/anthonyscarpaci/
Tunomatic → https://www.tunomatic.com/
Growth Notes Newsletter → https://tunomatic.substack.com/