---
name: al-performance-review
description: 'Use for AL slow paths, large record loops, repeated database calls, or BC CPU-profile hotspots.'
user-invocable: false
---

# Performance Review

Use this skill when a change touches large data volumes, loops over records, posting flows, background processing, or integration-heavy operations.

## Review focus

- broad record loading where narrower field or filter usage is possible
- queries inside loops and other N+1 patterns
- partial-record opportunities with `SetLoadFields`, accounting for later field access and JIT loads
- weak `FindSet` / `FindFirst` / `FindLast` choices
- repeated FlowField calculations or expensive work in loops
- row-by-row updates where bulk operations preserve required trigger and validation behavior
- design choices that should move heavy work to separate tables, summaries, or background processing

## Review questions

- Is the database doing the filtering, or is code filtering too late?
- Are only required fields being loaded?
- Is the iteration pattern appropriate for expected volume?
- Will the design stay acceptable at 10K, 100K, or larger record counts?
- Is there a simpler architectural split that reduces runtime cost?

## Output expectations

For `.alcpuprofile` evidence, use [runtime artifact guidance](../al-debugging-methodology/references/runtime-artifacts.md). Recommend optimizations from measured or clearly bounded cost, not blanket `SetLoadFields` or bulk-update rules.

- findings first
- concrete AL-level improvement suggestions
- note expected impact when it is reasonably clear
- separate must-fix performance risks from optional optimization ideas