How do you decide between Spring Data repository methods, projections, and custom repository code?

Uses generated repository behavior for simple cases and explicit query or implementation boundaries when the use case becomes complex.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I use derived query methods for small, obvious predicates and standard CRUD behavior. I use @Query when a fixed JPQL or native query states the intent more clearly. Interface, record, or constructor projections are useful for read models that need only selected fields. When a query needs dynamic composition, provider-specific APIs, or substantial implementation logic, I add a custom repository fragment or a focused query service.

Full answer below

Practice focus

Derived queries, @Query, projections, custom fragments, repository ownership, paging, and generated SQL.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
Spring Data removes repository boilerplate; it does not make every query belong in a long method name.
I use derived query methods for small, obvious predicates and standard CRUD behavior. I use @Query when a fixed JPQL or native query states the intent more clearly. Interface, record, or constructor projections are useful for read models that need only selected fields. When a query needs dynamic composition, provider-specific APIs, or substantial implementation logic, I add a custom repository fragment or a focused query service.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
Repository methods should describe domain or application intent, not leak every table operation into services. Returning entities is appropriate when the caller will modify a managed aggregate; list and reporting paths usually benefit from projections. Pageable queries may issue a count query, so I inspect both the content and count SQL and avoid expensive totals when the product can use slices or cursors. is not proof of one bulk SQL statement.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Encoding complex logic in a huge derived method name.
Returning full entities for every list view.
Assuming saveAll means one database statement.

How do you decide between Spring Data repository methods, projections, and custom repository code?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section

Tech stacks

Skill areas

Suggest an improvement

Found this answer useful?