Working with Engineering Best Practices
Partner early on schema design, not post-launch firefight. Use this list to align product, application engineering, and data platform teams.
How to Use This List
- Review at epic kickoff for any feature touching persisted data.
- Product manager and tech lead tick items before sprint commitment.
- Escalate unchecked high-risk items to database council.
- Revisit after incidents where comms or scope failed.
A - Early Partnership
- DBA or data lead in discovery for new entities. Schema sketched before UI mock final.
- Retention and compliance requirements in PRD. Not a launch-week legal surprise.
- Load estimate (rows/day, QPS) documented. Drives indexing and partitioning.
- Multi-tenant isolation model chosen explicitly. RLS, schema-per-tenant, or DB-per-tenant.
B - Delivery Alignment
- Migrations scheduled with product blackout calendar. Marketing launches visible to platform.
- Feature flags decouple app deploy from DDL contract step. Expand/contract respected.
- Analytics freshness expectations written down. Replica lag acceptable minutes documented.
- Rollback story includes app revert and schema state. Product knows forward-only paths.
C - Communication
- User-visible maintenance impact in plain language. Not "DB work."
- Single incident customer summary within 24h of Sev-1/2. Product comms aligned.
- Monthly one-page data health brief for leadership. Green/yellow/red with ask.
- Cost drivers explained before finance escalation. Storage and replica story ready.
D - Continuous Improvement
- Post-mortem action items assigned across teams. Not only platform tickets.
- MVCC and pool training for new squads. Within first month of service launch.
- Query review standards referenced in team working agreement. SQL treated like API.
- Case studies replayed after major upgrades. Lessons shared outside platform silo.
FAQs
When to involve product in schema?
When migration risk tier is medium or higher, or feature adds regulated data.
How to say no to launch date?
Present risk table: launch on time with Sev risk vs delay with correctness fix.
Should product own retention?
Product + legal own policy; engineering implements and audits execution.
How to partner with data science?
Separate replica or warehouse contract; no ad hoc heavy SQL on primary without review.
What if eng skips early review?
Council blocks production migration until checklist complete; no shame, enforce process.
How to handle agency vendors?
Same migration and review bar; temporary credentials and schema access logged.
BI tool direct to production?
Read replica with role limits; never default BI to primary.
Joint OKRs example?
Reduce checkout DB p95 30% while holding infra spend flat QoQ.
Remote collaboration?
Async schema doc comments 48h before council; record decisions in ADR.
What should I read next?
See Stakeholder Basics for executive framing.
Related
- Stakeholder Basics - risk translation
- Data Retention & Legal Holds - compliance
- Cost Conversations - finance
- Technical Leadership Best Practices - council model
- Change Management Basics - DDL tiers
Stack versions: This page was written for PostgreSQL 18.4 (stable 18, maintenance 17), pgvector 0.8+, PgBouncer 1.x, Patroni 3.x, and PostGIS 3.5+.