Skip to content

Technical for PMs · Hard

Engineering is debating a relational database versus a NoSQL store for a new feature. How do you contribute?

Short answer

Your job is not to choose the database but to make sure the decision reflects product needs. Clarify the data's shape and relationships, consistency requirements, expected scale and growth, query patterns, reporting needs and the team's familiarity. Explain how each option affects future features and timelines, then let engineering make the technical call with that context.

Detailed answer

What a PM brings to the conversation

Provide product context that shapes the choice:

  • Data shape: highly structured with many relationships (orders, customers, invoices) or flexible and varied (user-generated content with optional fields)?
  • Consistency: must every read reflect the latest write (balances, inventory), or is a short delay acceptable (like counts, activity feeds)?
  • Query patterns: complex joins and ad-hoc reporting, or simple lookups by key at very high volume?
  • Scale and growth: rough expected volume, peaks and growth.
  • Future roadmap: features likely to need new queries or relationships.
  • Team and operations: what the team already runs well.

Simplified trade-offs

  • Relational (SQL): strong consistency, powerful queries and joins, well-understood. Schema changes need planning.
  • NoSQL (document, key-value and others): flexible schemas and some types scale horizontally more easily. Complex queries and cross-record consistency can be harder.

Worked example

Feature: a loyalty-points system.

Points balances must be accurate because users redeem them for money-like value, and finance needs reliable reports. These point towards strong consistency and good reporting, which a relational database handles well. A PM would surface these needs; engineering would decide, perhaps also using a cache for fast balance reads.

Common pitfalls

  • Advocating a technology because it is fashionable.
  • Ignoring reporting and analytics needs.
  • Overstating your technical authority instead of providing context.