Incremental delivery

Incremental delivery ships vertical slices—thin end-to-end increments with usable behavior, migration safety, observability, and rollback paths. Each slice should be reviewable, deployable, and valuable or learning-rich on its own.

| Slice pattern | Example | |---------------|---------| | Read before write | List exports before creating async jobs | | Dark launch | Write new path, compare metrics, no user switch | | Percent rollout | 5% → 25% → 100% with guardrails | | Strangler | Route one tenant or region first |

					Slice 1: CSV export for admins, sync, 1k row cap + metrics
Slice 2: async job + email link + empty-state UX
Slice 3: scheduled exports + audit log
				

On interviews: break a large feature into slices that each reduce risk or deliver user value; mention migrations, feature flags, and how you would verify each slice.

Common pitfalls: horizontal layers only (all DB work, then all UI); slices too large to review; no observability per slice.

The trade-off is more frequent integration overhead versus smaller blast radius and faster feedback.

Checklist:

• Vertical slices with real user or ops value.
• Migration and rollback per slice.
• Metrics or tests proving slice success.
• Avoid months-long dark branches.