Quick Reference: Solution Architecture (D4)

Fast-track refresher for Domain 4, covering the right mix of declarative and programmatic functionality and evaluating external applications. Read this, then review the decision flowcharts before your mock board.

What the Board Wants to Hear

Board Question	What They Are Testing	How to Answer
”Why Flow instead of Apex?”	That you evaluated trade-offs, not just defaulted	Cite specific criteria: volume, maintainability, team skills, governor limits
”Why did you buy vs build?”	TCO thinking, exit strategy awareness	Walk through the evaluation: native first, then AppExchange, then custom
”Why LWC over Aura?”	Technology currency and pragmatism	Greenfield = always LWC. Existing Aura = keep if stable, new features in LWC
”What are the risks?”	Proactive risk identification	Governor limits, vendor lock-in, team skills gap, upgrade impact

The Declarative-First Spectrum

Not binary. Five positions exist:

Figure 1. The declarative-first principle means always evaluating positions left to right: only move right on the spectrum when you can name the specific gap that the current position cannot fill. Moving from Platform Native to Custom Build without passing through Configure and AppExchange is a common over-engineering trap.

Flow Builder Quick Reference

Flow Types at a Glance

Flow Type	Trigger	Best For
Before-Save Record-Triggered	DML (before commit)	Field updates on triggering record - no DML cost
After-Save Record-Triggered	DML (after commit)	Related record updates, platform events
Screen Flow	User click	Guided wizards, data entry
Schedule-Triggered	Cron	Batch reminders, cleanup jobs
Platform Event-Triggered	Platform Event	Async, decoupled processing
Autolaunched	Invoked by code/process	Reusable logic modules
Orchestration	Multi-step	Long-running, multi-stage processes

Critical Governor Limits (Shared with Apex)

Limit	Value	CTA Gotcha
SOQL queries	100/txn	Each Flow “Get Records” counts
DML statements	150/txn	Each Flow “Create/Update/Delete” counts
CPU time	10,000 ms	Formula-heavy flows eat CPU
Elements per txn	No hard cap (removed in API v57.0+)	Constrained by CPU time limits

Before-Save vs After-Save

Before-save flows modify the triggering record in memory - zero DML cost. After-save flows require DML for related records. Always use before-save when you only need the triggering record. This is a frequent board question.

Apex: When Code Wins

Scenario	Why Apex
Complex data transformations	Maps, sets, conditional loops are cleaner in code
Volume > 50K records	Batch Apex handles millions; Flow struggles
Multi-object transactions	Explicit savepoints and rollback
External callouts	Retry logic, circuit breakers, error handling
Custom REST/SOAP endpoints	Inbound integration
Performance-critical paths	Compiled vs interpreted

The Bridge Pattern: Invocable Actions

@InvocableMethod = best of both worlds. Complex logic in testable Apex, callable from admin-friendly Flow. Always mention this at the board when explaining hybrid approaches.

Modern Platform Features Cheat Sheet

Feature	Maturity	When to Recommend
LWC	Mature	All new UI. Always.
Screen Flows	Mature	Guided processes, wizards
OmniStudio	Mature (Industries)	Industry templates, complex multi-step UI
Dynamic Forms	Maturing	Complex page layouts, multiple personas
External Services	Mature	Simple REST calls without Apex
Agentforce	Early GA	Customer service AI - recommend with phased rollout
Einstein Prediction Builder	Mature	When 6+ months historical data exists
Flow Orchestration	Maturing	Multi-step, long-running processes

OmniStudio vs Screen Flow - The Quick Test

Industry-specific process + Industry Cloud license + templates exist = OmniStudio. Everything else = Screen Flow (add custom LWC if UI needs exceed standard components).

Build vs Buy Decision - Condensed

The 4-Question Filter

1. Does Salesforce do this natively? —> Use platform native
2. Is this core to competitive advantage? —> Build custom
3. Is there a mature AppExchange package meeting 80%+ of requirements? —> Buy
4. Is TCO favorable over 3-5 years? —> Validates buy decision

Vendor Scorecard - Top Criteria

Criterion	Weight	Red Flag
Feature completeness	15%	Meets < 80% of requirements
Performance impact	10%	Excessive governor limit consumption
Support quality	10%	No SLA, limited hours
Upgrade path	10%	Push upgrades with breaking changes history
Exit strategy	10%	Proprietary data format, no export
Security review	5%	Not passed or expired

Vendor Lock-In Red Flags

Proprietary data formats. Replaces standard SF objects. Bypasses sharing model. No API access to stored data. Requires exclusive admin access.

Reverse-Engineered Use Cases

Scenario 1: Insurance Company - Claims Processing

Situation: 200-person insurance company needs guided claims intake with document collection, multi-step approval, and ERP sync.

What you’d do: OmniStudio (OmniScript for guided intake with industry templates, DataRaptor for data transformation) + Apex for ERP callout with retry logic. Screen Flow alone would require too many custom LWC embeds for the complex UI requirements. OmniStudio is justified because they already have Industry Cloud licensing and insurance-specific templates exist.

Trade-off: Higher skill requirement for OmniStudio maintenance. Mitigated by vendor training and documented runbooks.

Scenario 2: Retail Company - Order Management

Situation: E-commerce retailer wants real-time inventory checks during order creation, with 500K orders/month.

What you’d do: Apex trigger for order creation (volume demands it) + Platform Event to decouple inventory callout (async, separate transaction) + Before-save Flow for simple field defaults on Order. Do NOT use Flow for the inventory callout - External Services cannot handle the retry logic and error handling needed at this volume.

Trade-off: Developer dependency for Apex maintenance. Mitigated by Invocable Action bridge so admins can adjust routing rules in Flow while Apex handles heavy lifting.

Scenario 3: Financial Services - Document Generation

Situation: Wealth management firm needs client-facing portfolio reports with compliance disclaimers, generated on demand and quarterly.

What you’d do: Buy Conga Composer (or equivalent). Document generation is not competitive advantage. AppExchange package meets 90%+ of requirements. TCO over 3 years is lower than custom build. Exit strategy: templates are Word-based (portable), data stays in Salesforce.

Trade-off: Per-user license cost, vendor dependency. Mitigated by contractual review, alternative vendors identified (Nintex DocGen), and data portability confirmed.

Key Anti-Patterns (Board Red Flags)

Anti-Pattern	Why It Fails	Fix
Multiple record-triggered flows per object	Unpredictable order, shared limits	One master flow + subflows
DML inside flow loops	Governor limit violation	Collect, then bulk DML after loop
Apex for simple field updates	Over-engineering	Before-save Flow
Hardcoded IDs in Flow/Apex	Breaks across environments	Custom Metadata Types
AppExchange without evaluation	Vendor lock-in, unknown limit impact	Use vendor scorecard
No error handling in Flows	Cryptic user errors, data inconsistency	Fault paths on every Flow

Deep Dive Links

• Declarative vs Programmatic: Full Flow vs Apex analysis
• Build vs Buy: Complete evaluation framework and TCO
• Modern Platform Features: OmniStudio, Agentforce (formerly Einstein Copilot), Dynamic Forms detail
• Decision Guides: All decision flowcharts
• Trade-Offs: Detailed trade-off matrices
• Best Practices: Patterns and anti-patterns
• Flow vs Apex Quick Reference: Decision matrix and build vs buy scenarios

Sources

• Salesforce Architects: Decision Guides
• Salesforce Architects: Well-Architected Framework
• Salesforce Developer Docs: Flow Builder
• Salesforce Architects: Build vs Buy
• Salesforce Help: OmniStudio