Command Palette

Search for a command to run...

Arbtr

Command Palette

Search for a command to run...

Decision Weight

Core Concept

How durable is this decision? How much would it cost to change it?

What is Decision Weight?

Decision weight is a human signal about how durable a concluded decision should be. It answers the question: "How expensive would it be to revisit this?"

This is not calculated automatically. You set it when concluding a decision based on your understanding of the blast radius.

The Four Levels

Anchor

Company-defining choices

These are decisions that define who you are as a company or team. Changing them would require extraordinary circumstances.

Examples:

  • Company mission and values
  • Core technology platform (e.g., "We are a Python shop")
  • Fundamental architecture (e.g., "Monolith vs. microservices")
  • Data residency decisions (e.g., "EU data stays in EU")
!
Use sparingly. Most decisions are not anchors. If everything is an anchor, nothing is.

Committed

High-impact operational decisions

Significant decisions that should remain stable. Reopen only with strong new evidence or changed circumstances.

Examples:

  • Database choice (PostgreSQL, MySQL, etc.)
  • Frontend framework (React, Vue, etc.)
  • Authentication architecture
  • API versioning strategy
  • Deployment platform

Standard

Default decision weight

The default weight for most decisions. These matter but can evolve as the team learns more.

Examples:

  • Roadmap priorities
  • Internal process decisions
  • Component library choice
  • Coding style guidelines
  • Team structure decisions
i
When in doubt, use Standard. You can always upgrade to Committed later if the decision proves more foundational than expected.

Tentative

Experiments and early explorations

Use this for working theories, MVPs, and decisions you're willing to change quickly based on new information.

Examples:

  • Feature experiments
  • Tool trials (trying a new library)
  • Process experiments
  • MVP architecture shortcuts
  • "Let's try this and see" decisions

How to Choose

Ask yourself these questions:

  1. How expensive is it to revisit?

    A database migration is expensive (Committed/Anchor). Changing a lint rule is cheap (Tentative/Standard).

  2. Who needs to be informed if it changes?

    If the whole company needs to know, it's likely Anchor. If only your team, probably Standard.

  3. What breaks if this changes?

    Check the graph. Decisions with many dependents are more important.

  4. How confident are you?

    Uncertain? Use Tentative. Very confident? Consider Committed.

Rule of thumb:

  • Anchor: 1-3 per company
  • Committed: 10-20 per team
  • Standard: Most decisions
  • Tentative: Active experiments

Data Model

lib/types.ts
type DecisionImportanceLevel =
  | "anchor"     // Company-defining, extremely hard to change
  | "committed"  // High-impact, should remain stable (displayed as "Committed")
  | "standard"   // Default decision weight
  | "tentative"  // Experimental or easy to revisit

Previous

Voting
    Importance | Arbtr Docs