Okrs

OKRs

OKRs work when the Objective says where we're going and why it matters, and the Key Results prove we actually got there with movement on a number — not a checklist of shipped work. This skill helps you draft a tight, focused set and then stress-test it against the anti-patterns that quietly kill most OKR cycles. The single hardest discipline: Key Results are outcomes, not the tasks you'll do to reach them.

Grounded in: Measure What Matters — John Doerr: outcome objectives with measurable key results (outcomes, not tasks). Go deeper (The Product Channel): OKRs: Secrets of Success

When to use this

• A new quarter or planning cycle is starting and you need to set goals.
• You have a strategy or company objective and need to translate it into team/individual OKRs.
• Someone hands you a draft and asks "are these good?" — you need to grade and fix them.
• Your KRs feel like a project plan and you suspect they're really just a to-do list.
• You want to align a goal up the ladder (company → team → individual) so the levels reinforce each other.

What good looks like

• Objective: qualitative, ambitious, time-bound, and memorable. A sentence a teammate could repeat from memory and feel pulled toward. No metrics in it.
• Key Result: a measurable OUTCOME with baseline → target and a named data source. It answers "how will we know we hit the Objective?" — not "what will we do?"
• Shape: ~1 Objective with ~3 Key Results. More than that and focus dies.
• Ladder: each level's Objective ladders up to the level above — a team KR is often the parent of a team-member's Objective.

Before you start (gather these)

• The higher-level goal/strategy this ladders up to (company or team objective). If you don't have it, ask — OKRs that don't ladder up are just a wish list.
• The metric(s) you can actually move this cycle, and who owns them.
• The timeframe (usually a quarter).
• Current baselines for each candidate metric — today's number, with its source. If a baseline is unknown, flag it; "establish the baseline" can itself be a KR — but it burns one of your ~3 scarce slots. Decision rule: only spend a KR slot on baseline-setting when the metric is already core to the Objective and instrumented this quarter; otherwise make it a beta/initiative.

Ask for anything missing before drafting. Don't invent baselines or targets.

Process

Ask before you draft. Ask 2-4 questions and wait: "What's the strategic bet this quarter has to advance?" · "What's the baseline for each metric you'd want to move?" · "Which of these are you committing to vs. aspiring to?" · "What's the guardrail you must not break while chasing this?" Drafting OKRs without baselines produces targets nobody can score. If the numbers are in workspace/metrics/, read them and state what you used.

1. Anchor up. Restate the higher-level objective/strategy this serves. Every KR you propose should be a believable contributor to it.
2. Draft the Objective. Write the qualitative, inspiring "where we're going and why." Make it memorable and time-bound. Strip out any numbers.
3. Derive 3 outcome KRs. For each, name a metric, its baseline, its target, and its data source. Ask of each: if this number moves, did the world actually change for a user or the business? If not, it's an output — rewrite it.
4. Separate KRs from initiatives. List the projects/features/work separately as initiatives. Initiatives are the bets that drive the KRs; they are never KRs themselves. "Ship feature X" belongs here, not in the KRs.
5. Pressure-test against the anti-patterns below. Grade each KR pass/fail and rewrite the failures.
6. Set cadence + scoring. Define a weekly or biweekly check-in and the 0.0–1.0 scoring convention (1.0 = target hit, ~0.7 = solid stretch progress, 0.0 = no movement). Label type per KR, not just for the set as a whole — a single set often mixes aspirational growth KRs (~0.7 is a win) with a committed guardrail KR you must hold (e.g. KR3 = committed). Mark each KR [committed] or [aspirational] so the scoring bar is unambiguous.

Output template

OBJECTIVE: 
  Ladders up to: 
  Timeframe: 

KEY RESULTS (outcomes — tag each [committed] or [aspirational]):
  KR1 —  | baseline  → target  | source:  | [aspirational]
  KR2 —  | baseline  → target  | source:  | [aspirational]
  KR_g —  | hold ≥  (target ) | source:  | [committed]   # guardrail/floor KR — no baseline→target arrow

INITIATIVES (the work that drives the KRs — NOT key results):
  -  → drives 
  - 

HOW WE'LL SCORE & CHECK IN:
  Cadence: , owner: 
  Scoring: 0.0–1.0 per KR (1.0 = target, ~0.7 = strong progress); type: 

RISKS / SANDBAGGING CHECK:
  - Targets we're confident we'll hit (too soft?): 
  - Targets that may be impossible (too hard?): 
  - Baselines we're unsure of: 

Avoid (anti-patterns)

• Output/task KRs. "Ship feature X", "launch the redesign", "hire 3 engineers" — these are work, not results. Move them to Initiatives and ask what outcome they should produce.
• No baseline or no number. "Improve activation" is not a KR. "Activation 38% → 50%" is. If there's no baseline, you can't score it.
• Too many objectives. Five "top priorities" means none. Cut to one Objective per team per cycle.
• Sandbagged targets. A target you're 95% sure you'll hit isn't a goal, it's a forecast. Stretch it until it's uncomfortable but plausible.
• Impossible targets. The opposite failure — a 10x target nobody believes demotivates and gets ignored. Aim for ~0.7-likely.
• A KR that's a to-do list. If your KR has multiple sub-bullets of tasks, it's a project, not a key result.

Tips

• If you can complete a KR without anyone's behavior or the business changing, it's an output — rewrite it.
• One Objective, three KRs is the sweet spot. Resist the urge to add a fourth.
• Pair "grow" KRs with a "guardrail" KR (e.g. grow signups without dropping retention below baseline) so you don't win the metric and lose the war. A guardrail KR uses a floor, not a baseline→target arrow — write it hold ≥ and tag it [committed].
• Keep Objectives free of numbers and KRs free of adjectives — the Objective inspires, the KRs measure.
• Launch-coupled KRs (a KR whose number can only move after a dated mid-quarter dependency — a feature GA, migration, or partner launch): write the gating date into the KR and set a conditional re-baselining rule up front — if the dependency slips past , re-cut the target to the time actually left rather than scoring it 0.0. And don't concentrate multiple KRs on a single unproven dependency; if one launch underpins the whole set and it slips, the entire quarter scores zero.
• Review weekly, score at quarter-end. OKRs you only look at twice a year are theater.

---

Make it shareable: the visualize skill renders this as an Objective → Key-Result tree — a self-contained HTML visual you can screenshot into a deck or Slack.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Sidsaladi9
• Source: Sidsaladi9/persona-os
• License: MIT
• Homepage: https://sidsaladi.substack.com

Install and usage instructions live in the source repository linked above.