Foundation
What is SAFe?
SAFe (Scaled Agile Framework) is a set of organisation and workflow patterns for implementing agile practices at enterprise scale. It provides guidance for roles, activities, and artefacts to synchronise alignment, collaboration, and delivery across large numbers of agile teams.
SAFe is built on Lean, Agile, and DevOps principles. It applies them at three primary levels — team, program, and portfolio — and adds a fourth Large Solution level for the most complex systems.
When to use SAFe: SAFe is designed for organisations with 50–150+ people working on a single product or product portfolio, where cross-team coordination and alignment are persistent pain points.
Architecture
The Four Levels
L4
Portfolio
Strategy, investment funding, Lean Portfolio Management, Epic-level governance
L3
Large Solution
Coordinating multiple ARTs; Solution Trains, Solution Architects, Value Streams
L2
Program (Essential SAFe)
Agile Release Trains, PI Planning, System Demo, Inspect & Adapt
L1
Team
Scrum or Kanban teams, iterations, team backlogs, team demos
Start with Essential SAFe (Levels 1–2). Only add Large Solution and Portfolio layers when the complexity genuinely demands it.
Foundation
Core Values
🤝
Alignment
Strategy cascades from portfolio to team through PI Objectives and OKRs.
Culture
📦
Built-in Quality
Quality is not inspected in — it is built in at every layer via engineering practices.
Engineering
🔄
Transparency
Radiate information up, down, and across. No hidden status.
Culture
🚀
Program Execution
Reliably deliver value — PI after PI, sprint after sprint.
Delivery
🌊
Lean-Agile Leadership
Leaders model and reinforce lean-agile mindset and practices throughout the org.
Leadership
Program Level
Agile Release Train (ART)
The ART is the primary value delivery mechanism in SAFe — a long-lived, self-organising team of agile teams (typically 50–125 people) that plan, commit, and execute together on a common mission.
ART Roles
🚂
Release Train Engineer
Chief Scrum Master for the ART. Facilitates ART events, removes impediments, drives improvement.
RTE
🏗
System Architect
Defines ART-level architecture. Facilitates Architectural Runway.
Architecture
🎯
Product Management
Owns the Program Backlog. Sets Program PI Objectives. Connects to portfolio strategy.
Strategy
💼
Business Owners
Key stakeholders. Attend PI Planning and score PI Objectives at end of PI.
Governance
ART Events
| Event | Cadence | Purpose |
|---|---|---|
| PI Planning | Every PI (8–12 weeks) | Align teams on goals; create team PI Objectives; identify dependencies |
| ART Sync | Weekly | Cross-team coordination; surface ART-level impediments |
| System Demo | Each iteration | Integrated demo of all teams' work for PI stakeholders |
| Inspect & Adapt | End of PI | PI demo + quantitative measurement + retrospective problem-solving |
Program Level
PI Planning
PI Planning is the heartbeat of SAFe — a two-day event attended by the entire ART (all teams, business owners, product management, architects). It aligns everyone on a common mission and creates team-level plans for the upcoming Program Increment.
Standard PI Planning Agenda (2 days)
| Time | Activity | Who |
|---|---|---|
| Day 1 AM | Business context; Vision; Architecture briefing | Execs, Product Mgmt, Architects |
| Day 1 PM | Team breakouts: Draft team PI Objectives; plan iterations 1–4 | All teams |
| Day 1 PM | Draft plan review with Product Management and RTE | All teams |
| Day 2 AM | Management review; problem solving; adjust plans | Teams + management |
| Day 2 PM | Final plan review; risk identification (ROAM) | All ART |
| Day 2 PM | PI Objective confidence vote; retrospective | All ART |
ROAM — Risk Management
R — Resolved: Risk is no longer a concern (explain why) O — Owned: Risk is owned by a specific person; action plan exists A — Accepted: Risk is understood; no action possible; accepted M — Mitigated: Plan is in place to reduce probability or impact
Confidence vote: At the end of PI Planning, each team votes 1–5 fists. A vote of 3+ fists ART-wide means proceed. Below 3 = significant problems must be addressed before committing.
Program Level
PI Objectives
Every team and the ART as a whole produces PI Objectives — a short list of business outcomes to achieve in the PI. They connect team-level work to business value.
Team PI Objectives format: → Business objective (outcome, not feature) → Committed or Uncommitted (stretch) → Business Value score from Business Owners (1–10) → Planned Value / Achieved Value (scored at PI end) Example committed objective: "Enable single sign-on for all enterprise customers so they can access the platform without separate credentials." Example uncommitted (stretch) objective: "Integrate with Okta — dependent on API docs availability."
Predictability Measure: ART predictability = Achieved Business Value / Planned Business Value × 100%. Target: 80–100%.
Program Level
IP Iteration
The Innovation and Planning (IP) Iteration is the final iteration of every PI. It is a dedicated timebox for innovation, hardening, integration testing, and planning activities.
IP Iteration activities: → Inspect & Adapt workshop (PI System Demo + retrospective + problem-solving) → Innovation time: hackathons, spikes, experiments, tech debt → PI Planning preparation → Training and enablement → Integration testing and hardening
Never use the IP Iteration as a buffer for unfinished features. Teams that do this consistently are signalling a planning or Definition-of-Done problem that needs to be addressed directly.
Portfolio Level
Portfolio Management
Lean Portfolio Management (LPM) connects the enterprise strategy to ART execution through budgets, epics, and strategic themes.
🗺
Strategic Themes
Business objectives that connect the portfolio to the enterprise strategy. Change annually.
Strategy
💰
Lean Budgeting
Fund value streams, not projects. Guardrails set boundaries; teams allocate within them.
Finance
📋
Portfolio Backlog
Epics waiting to enter a value stream. Reviewed at the Portfolio Kanban.
Planning
📊
Portfolio Kanban
Funnel for epics — from Funnel through Analysis, Portfolio Backlog, Implementing, Done.
Flow
Lean Budget Guardrails
Guardrails control spending without micromanaging teams: → CapEx vs OpEx split guidance → Approved PI Objectives thresholds → Spending authority tiers per value stream → WSJF prioritisation for epics competing for budget
Portfolio Level
Epics & WSJF
Epics are large initiatives that span multiple ARTs or PIs. They are prioritised using Weighted Shortest Job First (WSJF) — an economic decision model.
WSJF = Cost of Delay / Job Duration
Cost of Delay = User + Business Value
+ Time Criticality
+ Risk Reduction / Opportunity Enablement
All values scored 1–21 (Fibonacci).
Divide Cost of Delay by estimated job size (also Fibonacci).
Highest WSJF score → next epic to implement.
Lean Business Case: Before a portfolio epic enters the Program Backlog, it must have a Lean Business Case: hypothesis, expected benefits, MVP definition, and go/no-go decision criteria.
Reference
SAFe Anti-Patterns
| Anti-Pattern | Problem | Fix |
|---|---|---|
| SAFe without culture change | Processes layered on top of command-and-control; nothing actually changes | Leadership must model Lean-Agile mindset first |
| PI Planning as theatre | Plans made before the event; teams just go through the motions | Real dependency discovery and real negotiations required |
| RTE as project manager | RTE assigns tasks, tracks individuals, escalates to management | RTE is a servant leader; teams self-organise |
| Funding projects, not value streams | Budget resets kill long-lived teams; knowledge lost | Lean budgeting funds value streams annually |
| Ignoring Architectural Runway | Teams cannot deliver without infrastructure; speed drops | Allocate 20% capacity per PI to enablers |
| IP Iteration as buffer sprint | Teams dump unfinished features into IP; it ceases to be innovation time | Definition of Done must be enforced sprint-by-sprint |
| WSJF theatre | Scores manipulated to rubber-stamp pre-made decisions | Use WSJF for genuine prioritisation conversations with Business Owners |
Reference
SAFe Cheat Sheet
Levels Portfolio → Strategic Themes · Lean Budgets · Portfolio Epics · LPM Large Sol → Solution Train · Solution Architect · Solution Intent Program → ART · RTE · PI Planning · System Demo · I&A Team → Scrum/Kanban · Iterations · Team Demos ART cadence (per PI) Iterations → 4–5 × 2-week sprints + 1 IP Iteration PI length → 10–12 weeks total PI Planning → 2 days, full ART, every PI PI Planning outputs → Team PI Objectives (committed + uncommitted) → Program Board (dependencies, milestones) → ROAM risks (Resolved/Owned/Accepted/Mitigated) → Confidence vote (≥3 fists to proceed) WSJF formula WSJF = (User+Business Value + Time Criticality + RR/OE) / Job Size Predictability ART Predictability = Achieved BV / Planned BV × 100% Target: 80–100%