What's a systematic approach to performance tuning a slow Salesforce org?

Performance tuning starts with measurement, not assumptions.

Step 1: Measure.

• Lightning Performance Analysis in Salesforce.
• Browser DevTools for client-side render time.
• Debug logs with FINEST level for transaction analysis.
• Event Monitoring for slow query and slow Apex.
• User-reported pain points — qualitative.

Identify what's actually slow. Don't optimise based on hunches.

Step 2: Diagnose.

For each slow path:

• SOQL queries — non-selective, missing indexes, too many.
• DML — large bulk inserts, validation rules firing on every record.
• Triggers and flows — unbulkified, recursion.
• Lightning page composition — too many components, heavy LWCs.
• Network — slow callouts, unfiled latency.
• Sharing recalc — pending sharing operations.

Step 3: Prioritise.

• High user impact + high frequency — fix first.
• Quick wins — significant improvement with low effort.
• Long tail — many small inefficiencies; address as opportunity.

Step 4: Optimise.

For each hotspot:

SOQL:

• Index hits via WHERE on indexed fields.
• Custom Indexes via Salesforce Support.
• Selective queries (avoid full table scans).
• Skinny Tables for hot reads.

Triggers / Flow:

• Bulkify.
• Move sync work to async (Queueable, Future, Batch).
• Before-save flows for same-record updates.
• Reduce element count in Flow.

Lightning Pages:

• Lazy-load heavy components.
• Reduce component count.
• Cache @wire data.
• Optimise images / static assets.

Apex code:

• Cache repeated queries.
• Use Maps / Sets, not List.contains.
• Aggregate SOQL where appropriate.

Step 5: Verify.

• Re-measure post-fix.
• Confirm improvement.
• Ensure no regressions.

Step 6: Monitor.

• Set up ongoing performance monitoring.
• Alert on regression.
• Trend over time.

Common pitfalls:

• Optimising without measuring — wasted effort.
• Optimising the wrong thing — visible but not the bottleneck.
• Local optimisation, global regression — fix one part, break another.
• No regression test — performance regresses silently.

Architect role: lead performance discipline. Establish standards (e.g., "page render < 2s"); enforce via automation (perf test in CI).

The senior insight: performance is a feature, not an afterthought. Build performance considerations into Discovery, design, code review.