CloudShetra

Building Scalable Salesforce Solutions: Key Architect Insights

—

by

Dilip Singh
in

Being an architect isn’t about having all the answers – it’s about knowing the right questions to ask and building systems

Salesforce gives you endless building blocks. But the real test isn’t how fast you can build – it’s whether your solution can scale without crumbling under its own weight, and most importantly, not over-engineering the solution.

In most orgs, complexity doesn’t arrive overnight. It sneaks in through quick fixes, siloed solutions, and undocumented assumptions. With millions of records, multiple business units, 1,000+ users, and third-party integrations, your solution design either becomes the foundation of growth… or a very expensive bottleneck.

A practical guide for architects and technical leaders who want to build Salesforce orgs that last. Again, this is not the FINAL list as there cannot be any list that is final as time passes, the systems evolve, and lot of other unknown factors will start unfolding. The key attribute for an architect is to be open and have a curiosity to know the unknown.

1: Data Architecture – Model to Grow, Not Just to Survive

Standard

  • Start from business questions; shape objects/fields to answer them.
  • Denormalize when it helps UX/reporting; normalize when reuse is real.
  • For large data volumes, consider Big Objects / External Objects and async patterns.

Red Flags

  • Timeouts, formula chains everywhere, users exporting to Excel “to make it work.”

Start with

  • List 5 core analytics questions → map to objects/fields → pre-calculate (rollups, nightly jobs).

2: Integration Strategy – APIs are your Org’s Heart:

Standard

  • Secure every endpoint – don’t assume internal = safe.
  • APIs are arteries. Middleware is your heart. Don’t DIY what the ecosystem already handles well.
  • Prefer Platform Events or Change Data Capture for real-time.
  • Avoid tight coupling; use canonical data models.

Red flags

  • Batch retries at midnight, mystery resyncs, one-off scripts nobody owns.

Start with

  • Draw a one-page context diagram with flows, SLAs, and auth per connection.

3: Org Strategy – Build like it if it merges tomorrow

Standard

  • Segment by business unit (record types, layouts, roles).
  • Prefer Permission Sets over all-powerful Profiles.
  • Planned sandbox strategy (naming, refresh cadence, data seeding).
  • Change Management Process.

Red lags

  • “God mode” users, copy-paste metadata, layouts that fit nobody.

Start with

  • Inventory Profiles → replace with Permission Set stacks per persona.

4: Security Model – build it like it will be hacked

Standard

  • Least privilege + layered sharing (OWDs, rules, manual).
  • Shield + Event Monitoring for telemetry and encryption at rest.
  • Day-one compliance thinking (HIPAA, GDPR, etc.).

Red flags

  • “Everyone can view everything,” stale public groups, no audit trail.

Start with

  • Turn top 3 threats into controls (encrypt PII, monitor high-risk events, break-glass access).

5: Deployment Strategy – CI/CD or Manual

Standard

  • Git as source of truth.
  • Unlocked packages for modularity and isolation.
  • Automated tests + one-click rollback.
  • Purposeful environments (scratch → dev → UAT → prod).

Red flags

  • Deployment weekends, manual change sets, undead hotfix branches.

Starts with

  • Put everything (metadata, seed data, scripts) in Git; wire a minimal pipeline.

6: Analytics Readiness – Design with Insights in Mind

Standard

  • Reports/dashboards are first-class deliverables.
  • Pre-compute heavy formulas; avoid nesting monsters in the report builder.
  • Use Reporting Snapshots / Data Cloud for trends and aggregates.
  • Keep models Einstein-ready (clean, labeled, timely).

Red flags

  • Executives don’t trust dashboards; shadow spreadsheets take over.

Start here

  • Define 10 “run-the-business” metrics → guarantee each has a fast data path.

7: Communication – It’s a key to all

Standard

  • Consistent visuals (ERDs, sequence, auth flows) plus a sharp story:
    • Problem → Options → Decision → Trade-offs → Risks → Next steps.
  • Treat reviews like a design board; defend trust, not just code.

Red flags

  • Great designs bypassed by louder opinions; architects reduced to ticket routers.

Start here

  • Create a one-pager per initiative with the story arc above; circulate early.

8. AI & Automation – Useful, Safe and Observable

Standard

  • Automation-first experiences (Flow/Orchestration) with guardrails (naming, versioning, DORA-style metrics).
  • AI assists (suggest, summarize, classify); it doesn’t make unreviewed commitments.
  • Unified governance: trigger/flow order, limits, telemetry, and conflict prevention.

Red flags

  • Duplicate flows, recursion, and “who built this?” mysteries.

Start here

  • Publish a Runbook: when to use Flow vs Apex vs Events; add linting and review rules with AI to it.

Final Thoughts

Scalable isn’t “more code.” It’s clear models, loose coupling, disciplined shipping, observable automation, and relentless communication. Nail these eight essentials and you deliver velocity with guardrails.

Drop your thoughts in the comments.

Need help designing a scalable org? Reach out for a discussion.

Let’s build smart, scalable Salesforce solutions – together.

Comments

Leave a comment