The Sunk Cost Trap
You have 18 months of runway. You have one shot.
Building authentication takes 8-12 weeks. Email delivery: 4-6 weeks. Payments: 6-10 weeks. Payment reconciliation: another 8 weeks. By month 8, you've shipped nothing customers care about.
Meanwhile, Stripe handles 99.99% of payment cases. Supabase handles auth. SendGrid handles email. They've already made the mistakes. You inherit their battle-tested code.
Your 12 weeks saved = 12 weeks shipping product.
The YC Decision Framework
Michael Seibel and the YC team use a simple filter:
Question 1: Is this your unfair advantage?
If no, buy it. Auth is not your unfair advantage. Email delivery is not. If your idea loses to Stripe or Firebase, you weren't building a startup.
Question 2: Can a SaaS tool do this at 80% quality?
If yes, buy it. Perfection is expensive. 80% shipped beats 100% delayed.
Question 3: Will your customer pay extra for your custom version?
If no, buy it. Your customer doesn't care if you built payments. They care if you solved their problem.
Question 4: Can you build this in under 2 weeks?
If no, buy it. Two weeks is the threshold. Anything longer competes with core product time.
The Matrix: What To Build, What To Buy
Always Buy (Non-Negotiable):
- Authentication (Clerk, Auth0, Firebase Auth)
- Payment processing (Stripe, Razorpay, Cashfree)
- Email delivery (SendGrid, Mailgun, Braze)
- SMS (Twilio, Sinch, Exotel)
- CDN and image hosting (Cloudflare, AWS CloudFront)
- Monitoring and observability (Datadog, New Relic)
- Analytics (Segment, Mixpanel)
Conditional Buy (Your Assessment):
- Search (Algolia if you can't afford latency, Elasticsearch if you can own infrastructure)
- Video streaming (Mux if you need quality, self-host if volume is massive and margins justify it)
- Real-time messaging (Socket.io for custom, Firebase Realtime if standard works)
- Scheduling/cron (Bull, Temporal for complex, off-the-shelf if simple)
Always Build (Core Moat):
- Your product's main value proposition
- Your algorithm or recommendation engine
- Your data model or domain-specific logic
- Anything that differentiates you from competitors
- Your business rules and validation logic
The India-Specific Arithmetic
Razorpay charges 2-3% + ₹10 per transaction. Building payments costs:
- 2 engineers, 10 weeks: ₹20-30 lakhs in salary.
- Compliance and testing: ₹5-10 lakhs more.
- Bugs and downtime: unquantifiable but real.
At 1,000 transactions/month, Razorpay costs ₹3,000-5,000. Your build costs ₹25+ lakhs. The ROI flips only at 100,000+ transactions/month. By then, you should be raising Series A, not debugging payment reconciliation.
Same logic: use Cashfree for payouts. Use Twilio for SMS. Use Postmark for emails.
The Hidden Cost: Opportunity Cost
Scott Belsky calls this "The Messy Middle"—the 12-month phase where most startups die because they're building instead of learning.
You have 10 weeks. You can build:
1. Custom auth + payments + email + your core product.
2. Or: buy auth + payments + email, build your core product twice.
Option 2 means you ship, measure, iterate, kill bad ideas, double down on signals. Option 1 means you ship late and never learn.
Founders who've raised at YC almost always chose option 2.
The Maintenance Tax
Building means maintaining forever. Auth breaks when Google changes OAuth. Payments fail when banks update APIs. Email deliverability degrades with ISP filters.
You stop building features to debug your payment system on Sunday.
SaaS tools take that load. Stripe hires 50 engineers to handle compliance so you don't.
When To Build It Yourself
Three cases:
1. Massive volume, thin margins. Stripe takes 2.9% + $0.30. At 10M transactions/year, you save $290K. Build only then.
2. Custom competitive advantage. Airbnb built custom payment flows because their marketplace had unique trust requirements. Your use case must be this specific.
3. It's a 2-week sprint. If you can ship it in a sprint and own the tech debt, ship it.
Most founders overestimate case 2 and 3.
The Non-Obvious Insight
The best founders are ruthless about this because they understand the real constraint: attention. A founder's focus is more valuable than a developer's code. Protecting your focus means outsourcing relentlessly.
Paul Graham: "Do things that don't scale." But building auth scales against you.
Actionable Framework
For your next feature, ask:
1. Does a SaaS tool exist that does 80% of what I need? (Yes → buy it.)
2. Would I raise less money if investors heard I built this instead of shipped product? (Yes → buy it.)
3. Can I ship my core feature without this? (Yes → delay, build it later if needed.)
If you answer "buy it" to any, you know the answer.
Your job is to build the thing only you can build. Everything else is renting from someone smarter.