Test Pyramid Variations

Classic pyramid (Mike Cohn)

       ╱╲   UI tests (few)
      ╱──╲
     ╱ Service ╲  (some)
    ╱──────────╲
   ╱  Unit tests ╲  (many)
  ─────────────────

Trophy (Kent C. Dodds)

For frontend-heavy:

       ╱╲   E2E
      ╱──╲
     ╱ Integration ╲  (most)
    ╱──────────────╲
   ╱  Unit + Static ╲
  ────────────────────

Integration tests give best ROI for FE — closer to user behavior.

Honeycomb (Spotify)

For microservice:

   /────\
  ╱ Integrated tests ╲  (most)
   \────/
   /────\
  / Implementation │
   \ Detail tests /  (few)

Focus on integration between services > unit within service.

Anti-patterns

Ice-cream cone

   ╱──────────╲
  ╱  UI tests   ╲  (70%)
   ╲──────────╱
   ╱ Integration ╲  (20%)
    \──────────/
     \ Unit /  (10%)
       \/

Result: slow, flaky, expensive. Often from outsourcing QA team isolated from dev.

Hourglass

• Lots of unit + lots of E2E.
• Few integration → integration bugs slip through.

Cup cake

• Lots of unit at bottom.
• Spike of manual at top.
• No middle layer.

Choose your shape

Depends on:

• App architecture (monolith, microservice, FE-heavy).
• Team maturity.
• Tool ecosystem.

VN context

Most mid-startup default to classic pyramid. Frontend-heavy app (e-com, edtech) often shift to trophy.