← Back to blog

Scaling Estimates in Multi-Team Environments

estimates multi-teams agile-at-scale

When you have 5 squads with 5 different velocity levels, 5 point scales, and 5 Product Owners, planning a coordinated release is a mathematical and organizational challenge.

The Fundamental Problem

Each squad calibrates its points independently:

  • Squad A: 1 point = label change
  • Squad B: 1 point = simple CRUD
  • Squad C: 1 point = small refactoring

The points are not comparable. And that’s okay.

Strategy 1: Don’t Normalize Points

Instead of trying to normalize points across squads:

  • Plan by squad: each squad delivers X points/sprint
  • Sum deliverables, not points: “Squad A delivers checkout, Squad B delivers catalog”
  • Use T-Shirt sizing for cross-squad epics for macro planning

Strategy 2: Normalization by Reference (When Necessary)

If you need to compare:

  1. Choose a reference item common to all squads
  2. Each squad estimates this item
  3. Calculate the normalization factor:
Squad A estimates reference = 3 points → factor 1.0
Squad B estimates reference = 5 points → factor 0.6
Squad C estimates reference = 8 points → factor 0.375

Apply the factor to each squad’s velocity to compare.

Strategy 3: Release Planning with Dependencies

When squads depend on each other:

  1. Dependency mapping:

    • Who needs whom?
    • What’s the logical order?

  2. Estimate each squad independently.

  3. Calculate the critical path:

    • Squad A (checkout): 2 sprints
    • Squad B (checkout frontend): 1 sprint (depends on A)
    • Squad C (integration): 1 sprint (depends on B)
    • Total: 4 sequential sprints, not 2 parallel

Strategy 4: Joint Planning for Cross-Squad Features

For features involving multiple squads:

Big Room Planning (or PI Planning)

  1. All squads in the same room (physical or virtual)
  2. Product vision presented by Product Management
  3. Each squad does its Planning for assigned features
  4. Dependencies mapped on a visual board
  5. Risks identified and mitigated

Duration

  • 1 day for 3-5 squads
  • 2 days for 6-10 squads

Output

  • Delivery plan per squad
  • Dependencies visualized
  • Risks and mitigations
  • Consolidated timeline

Consolidated Metrics

MetricHow to consolidate
Total velocitySum deliverables (not points). “5 squads delivered 18 features.”
Lead timeMeasure per feature, not per squad.
Delivery rate% of features delivered by all squads vs. planned
QualityProduction bugs combined from all squads

Scaling Anti-Patterns

Comparing velocity across squads

“Squad A delivered 30 points, Squad B 20. Squad A is better.” Wrong — different scales.

Forcing the same point scale

“All squads must calibrate their points the same way.” Impossible and counterproductive. You lose local calibration.

Centralizing estimates

“The architect estimates for all squads.” Eliminates autonomy and precision from those who know the code.

Conclusion

Scaling estimates is not about normalizing — it’s about coordinating. Respect each squad’s autonomy, map dependencies explicitly, and use joint planning for cross-squad features. The math is simple: dependencies determine the timeline, not points.