Case study · barhop
BarHop
Real-time group bar-crawl planner
- ● Live pilot · Denver
- ★ Featured in KrispiTech · Jun 2026
A full-stack crawl app with live GPS where groups share one link and move together in real time — now piloting with a Denver pub-crawl company.
- React
- TypeScript
- Firebase
- GCP
- Mapbox
- Vercel
Incident replay · live sessions at 50 users
- Users joined
- 50/50
- Updates delivered
- 749/749
- p95 write
- 0.14s
- Errors
- 0
Restructured the data model. Visualization of measured load-test results.
01 · Problem
Groups planning a bar crawl lose each other and the plan; organizers need each group to run its own crawl.
02 · Diagnosis
Load testing live sessions found a stall at 50 concurrent users — p95 writes hit 50s and only 6 of 749 updates were delivered. Separately, a real billing incident traced back to exposed API keys.
03 · Fix
- Restructured the real-time data model behind live sessions.
- Moved billed keys behind an auth-gated proxy with rate limits and a global guest cap.
- Built organizer mode from a Denver pub-crawl company’s pilot feedback: branded pages where each group starts its own live crawl, no account needed.
04 · Result
- p95 writes 0.14s, 749/749 updates delivered, 0 errors.
- Live pilot with a Denver pub-crawl company.
- Featured in KrispiTech (June 2026).