subjects / SUB-017

measuring success — okrs, impact, do we even have the data.

open opened 2026-06-02 · touched 2026-06-02 · m5
#subject#measurement#okr#success#claude-os#studio#data-bank

# SUB-017 — Measuring success / OKRs

> Paste SUB-017 to context-shift here. Spun out of the 2026-06-02 Claude OS conversation. The data-bank tags need a "does this matter, and did it work" dimension; this subject owns that question.

The question

Everything in the system should have meaning. But:

  • Can we even measure success? Do we have data that tells us whether something worked?
  • How do we define success for a given piece of work? (money saved/gained? a metric moved? a decision unblocked? a stakeholder satisfied?)
  • Where does success measurement live inside the data-bank schema — is it a tag, an entity, a view?
  • OKRs: sets of objectives + measurable key results, so work can be scored against intent, not vibes.

Why it matters

The schema (_tech-databank-schema.md) tags work with Impact, business-goal link, and Importance. But those are *predictions* (how important we think it is). Success measurement is the *outcome* (did it actually do what we hoped). Without it, "impact" is a guess that never closes the loop.

This connects to: the goals bank (SUB-016 §7 triad), the business-goal tag, and the "what to do next" ranking — if we can measure outcomes, the ranking gets honest.

Early thoughts (to develop)

  • Define success per work type: a fix, a feature, a research piece, a prototype — each has a different success signal.
  • OKR layer: Objectives (qualitative direction) + Key Results (measurable). Could be a vault entity that tasks/scopes/areas link to.
  • The data problem: do we have access to the numbers (Cosmo usage, money, time saved)? If not, what's the cheapest proxy? Maybe self-rated outcome at first, real metrics later (mirrors the Certified/verification honesty: don't fake a number).
  • Where it lives: likely a success/outcome field on a work entity + an OKR entity in the vault Spine, with a Studio view to review outcomes.

Open threads

  • Can we get real outcome data (Cosmo metrics, business numbers)? What's accessible?
  • Define success per work-type (fix/feature/research/prototype/decision).
  • OKR as a vault entity — shape + where it lives (Spine? Augmentation?).
  • Is success a tag, a field, or its own entity in the schema?
  • Self-rated outcome now vs real metrics later — staging.
  • How outcome feeds back into the "what to do next" ranking.

Raw refs

  • chat: 2026-06-02 Claude OS presentation session.
  • schema: vault/08-Design-specs/studio/_tech-databank-schema.md (the tag dimensions this extends).
  • concept: SUB-016 §7 (goals bank + triad).

Last activity

2026-06-02 23:06 — Subject created. Seed. Gets a placeholder slide in the Claude OS presentation; full measurement model TBD.

context-shift paste this id into chat to resume on this thread