TRAINING PROJECT

Apex backend optimization (FASHA)

Optimization of FASHA's Apex backend: refactoring, removal of DML in loops and batch improvements.

Context

FASHA, a multinational clothing distribution company (present in 60 countries), used Salesforce to manage accounts, orders and revenue calculations. Three critical issues affected CRM reliability: abnormally slow revenue update batches, deadlocks during simultaneous modifications on Account and Order objects, and disorganized code with no naming conventions. Mission via SFQUAL consulting: audit, optimize and refactor the entire Apex backend.

Problems Identified

Weekly revenue update batches: abnormally high execution time after price changes
Deadlocks during simultaneous modifications on Account and Order — intermittent data loss
SOQL and DML inside FOR loops: N+1 queries, Salesforce governor limits at risk
Code with no separation of concerns: triggers with inline business logic, overly long classes
No unit tests — no guarantee of non-regression during future evolutions

Optimizations Made

Extracted all SOQL queries out of loops — collections used for bulk operations
Removed all DML inside loops: upsert/update lists processed in a single operation
Handler pattern applied: triggers emptied of logic, delegated to dedicated service classes
Batch Apex refactored with filtered SOQL queries and bulk-safe processing
Unit test suite covering nominal, boundary and error scenarios for each method — 0 SOQL/DML in loops confirmed

Stack & Technologies

Apex Triggers — Handler pattern, logic/presentation separation
Batch Apex — Bulk processing of revenue updates
SOQL — Query optimization (collections, filters, one-shot)
Apex Test Framework — Unit tests with @isTest and assertions
Salesforce DX — Deployment and versioning

Project Timeline

1
Code audit
Analysis of existing code — SOQL/DML in loops, inline trigger logic and untested classes mapped
2
Trigger refactoring
All trigger logic extracted to dedicated handler classes — clean, testable architecture
3
SOQL/DML optimization
All queries moved out of loops, collection-based processing, dramatic reduction of consumed governor limits
4
Batch Apex
Revenue update batch refactored: filtered queries, bulk-safe processing, corrected net calculation
5
Unit tests