Architect Thinking – How I Design and Run an Architecture Review Board (ARB)

1. Purpose of an ARB

The Architecture Review Board (ARB) is a forum to make sure big technology decisions are not made in isolation. The goal is simple:

  • Consistency – avoid building the same thing twice in different ways.

  • Visibility – make design choices clear to everyone, not hidden in a project.

  • Quality – catch design gaps early before they become expensive to fix.

Look below for “Tools to support the purpose of ARB”.

2. Governance Framework

Structure

  • Who attends: architects from different domains, data leads, security engineers, and sometimes product or business reps if the topic affects them.

  • Who runs it: usually a senior architect.

  • Focus groups: smaller pods when reviews need specialist focus (e.g., data, AI/ML, integration).

Operating Model

  • Strategic Reviews: look at portfolio-level initiatives and make sure they line up with where the enterprise is heading.

  • Tactical Reviews: review individual projects or systems to check they use agreed standards and patterns.

  • What’s reviewed: high-level design docs, data flows, API designs, system diagrams.

3. Process Lifecycle

  1. Intake – Teams submit a short proposal (a page or two).

  2. Screening – Quick check by architects to see if it’s complete and needs a full review.

  3. Review Session – Regular cadence (weekly/bi-weekly). Teams present, board asks questions, discussions happen.

  4. Decision – Approve, approve with conditions, or send back for rework.

  5. Record – Decision and rationale stored somewhere visible (repo, wiki, EA tool).

  6. Feedback Loop – Lessons learned feed back into patterns and standards.

4. Key Principles

  • Keep it lightweight – proposals and sessions should be short and focused.

  • Match effort to risk – deep review for critical projects, quick path for small changes.

  • Be transparent – publish what was decided and why.

  • Improve over time – refine the process based on feedback from teams.

5. Risk-Based Categorization

Dimensions I Use

  • Regulatory/Privacy impact – does it touch sensitive or regulated data?

  • Data & Security – is it creating a new system of record or external connection?

  • Business impact – will downtime or errors directly affect customers or revenue?

  • Technology novelty – is it new/unproven tech or a deviation from standard patterns?

  • Scope – enterprise-wide vs. a local system tweak.

Scoring

  • If any dimension is high, it goes to full ARB.

  • If all are low, it’s fast-tracked.

  • If it’s a mix, it goes to a smaller delegated review.

Example Applications

  • Critical: New enterprise data platform handling sensitive data.

  • Medium: Adding features to existing dashboards.

  • Low: Tuning performance of a pipeline without design changes.

6. Examples From My Work

  • FOREX Risk Management – real-time trading pipelines → full ARB; reporting dashboards → fast-tracked.

  • Consumer Lending – loan origination with credit bureau calls → full ARB; UI dashboard tweaks → delegated review.

  • Merchant Banking – new payment settlement platform → full ARB; reconciliation enhancements → light review.

  • Healthcare Claims Processing – workflows with patient data → full ARB; claim status UI → fast-tracked.

  • Healthcare Diagnostics Analytics – ingestion of diagnostic images → full ARB; visualization built on existing governed datasets → fast-tracked.

  • Vehicle Manufacturing Digital Sales – enterprise sales/financing platforms → full ARB; dealer portal updates → light review.

  • Customer Onboarding – identity verification/KYC → full ARB; UX refinements → fast-tracked.

Tools to Support the Purpose of an ARB

1. Consistency – avoid building the same thing twice

  • Enterprise Architecture Repositories: Ardoq, LeanIX, Sparx EA, Orbus iServer — keep a central inventory of applications, integrations, and reference architectures.

  • Standards & Pattern Libraries: Confluence, SharePoint, GitHub Wiki — publish reusable blueprints, integration patterns, and security standards.

  • API & Integration Catalogs: Postman, SwaggerHub, Azure APIM — maintain a central catalog of existing APIs and services to promote reuse.

2. Visibility – make design choices clear across the organization

  • Collaboration & Documentation: Confluence, Notion, SharePoint, Google Docs — record ARB submissions and decisions in a way everyone can access.

  • Architecture Diagramming & Sharing: Lucidchart, Draw.io, Miro, Microsoft Visio, ArchiMate tools — allow teams to visualize and share designs consistently.

  • Decision Logging: Architecture Decision Records (ADR) in GitHub/GitLab repos, Jira–Confluence integration — capture key design decisions in a structured, searchable format.

3. Quality – catch design gaps early

  • Proposal & Workflow Management: Jira, ServiceNow, Azure DevOps — manage ARB submissions and track progress through review workflows.

  • Compliance & Policy Automation: Open Policy Agent (OPA), Terraform with policy-as-code, Azure Policy, AWS Config — automate checks for security and cloud compliance.

  • Data Governance Tools: Collibra, Azure Purview, Informatica EDC — ensure metadata, lineage, and sensitivity are validated during review.

  • Testing & Validation: SonarQube, Checkov, Prisma Cloud — run code, infrastructure, and security scans before sign-off.

Putting it Together in Practice

  • For consistency, teams check the EA repository and API catalog before designing something new.

  • For visibility, they submit a design doc in Jira/Confluence with linked diagrams, and decisions are published openly.

  • For quality, automated policy scans (cloud, security, data) run before ARB sessions, so the board focuses on design trade-offs instead of basic compliance.

Closing

For me, an ARB is a structured way to review designs without slowing teams down. The trick is to make it proportionate: the bigger the risk and impact, the deeper the review; the smaller the change, the lighter the touch. Over time, this builds up a library of decisions and patterns that makes the next project easier.