Free OKR diagnostic

Diagnose your OKR.

Paste your OKR. Get a structured critique in seconds: what's broken, why, and three rewrites you can use immediately. Your API key never leaves your browser.

Analyze your OKR How the rubric works
No account No tracking No backend
Diagnostic result Sample OKR
Pasted as-is
Become the market leader in onboarding
KR1 Launch the new onboarding flow
KR2 Improve user activation

A real paste: ambitious words, no numbers, no timebox. The rubric reads it in seconds.

Key Result 1 Launch the new onboarding flow
Outcome form
A launch is an output. Score the change in behaviour it should cause.
Measurability
No baseline, no target. "Done" is the only signal this KR can send.
Other angle, adoption focus
Raise week-1 activation from 31% to 45% by end of Q3
Up to three rewrites per flagged Key Result.
4
fixes recommended
Ranked by impact, rewrites included.
Fix first
  1. KR1: score the adoption the launch should cause
  2. Give the set a timebox, for example "by end of Q3"
Already good
  • Two KRs is a focused, reviewable set

Instant, pre-scored. Or paste your own OKR below.

Analyze

Run your OKR through the rubric.

The pre-score runs locally, instantly, no key needed. The full AI report adds per-criterion scoring, fix-first triage and rewrites. Three free a day, or bring your own key for unlimited.

This is an What is the difference?

A stretch goal. Built to land at 60-80%, not 100%.

Your OKR

Diagnosing an OKR, not a KPI? See the difference.

No key set. Pre-score only.

Your scored report lands here. Load the sample, or paste your own and hit Analyze.

Copied to clipboard
Rule engine
Instant pre-score, no key required.

Runs locally before the LLM call. Output-verb detection, timebox regex, placeholder check. Instant pre-score, no key required.

Output-verb detection Timebox regex Placeholder check Runs locally
Three ways to work

Pick your mode.

One tool, three surfaces. Start with one OKR, run a batch, or build from scratch.

Diagnose
Deep critique

One OKR, a deep critique and three rewrites.

Batch
Score at scale

Paste many OKR sets, score them all at once.

Create
Build from scratch

Build a sound OKR from scratch, guided.

The rubric

Seven criteria. One honest score.

Every OKR is scored against seven named criteria. Each scores 0, 1 or 2, and the total normalises to 0-100. Pick a layer to see what it checks:

Where you are going. Three checks on the Objective itself:

1
Outcome direction
States a change in behaviour or state, not a position or activity
2
Ambition level
Stretches without being implausible
3
Clarity and scope
Unambiguous enough to guide prioritisation

How you will know. Two checks applied to every single Key Result:

4
Outcome orientation
Measures a result, not a task or deliverable
5
Measurability
Has a baseline, target, and unit of measurement

Does it hang together? Two checks across the Objective and all Key Results combined, the layer the report calls "the whole set":

6
Timebox
The set has an explicit time horizon
7
Internal alignment
KRs collectively cover the Objective without contradicting each other
Plus: the "So What?" test
If you hit every KR and nobody notices or cares, the Objective was wrong. This final check catches OKRs that pass the rubric but lack real stakes.
Anti-pattern catalogue

Twelve OKR anti-patterns, each with the fix.

The most common ways teams write OKRs that look fine but measure the wrong things.

A shipped box with a checkmark beside a chart whose line stays flat
01
Output-as-KR
KR flaw A KR that describes work the team does, not a change in the world.Fix: ask "if this ships and works, what changes for users?" and write that as the KR. A big balloon rises while a tiny buoy below stays unmoved
02
Vanity Metric
KR flaw A number that can grow while the thing you care about stays flat or gets worse.Fix: name the specific actor and the observable action they take. The orca steps over a bar set at ankle height
03
Planning Theatre
Set flaw A KR set at a level the team would reach anyway.Fix: if hitting 70% of this would genuinely satisfy you, raise the bar. A card holding only empty dashed outlines
04
Placeholder
KR flaw TBD, X%, or "to be confirmed" still sitting in the OKR.Fix: fill the field or strike the KR before locking the set. Do not commit drafts. An hourglass with no sand inside
05
Timebox Missing
Set flaw An Objective with no by-when.Fix: state the quarter or a specific date directly in the Objective text. A chain with its middle link missing
06
Alignment Gap
Set flaw No explicit reference to the parent strategy, pillar, or company OKR.Fix: name the parent component verbatim in the Objective or KR1. A flag planted on a tiny bump while a mountain stands unclimbed behind
07
Ambition Gap
Obj flaw A KR that requires no behaviour change to hit.Fix: find what a 10x outcome would require, then set the bar at 2-3x current. A gift box with a hammer sticking out
08
Solution Smuggling
Obj flaw An Objective that names a solution, foreclosing the alternatives.Fix: frame the problem at Objective level. Let the KRs reveal what changes. The orca swims a dashed circle, chasing its own tail
09
Activity-as-Outcome
KR flaw A KR that measures activity (touches, sessions, calls) instead of the result it should produce.Fix: ask "after you counted that, what changed?" and write the consequence. An arrow rises, curves over, and dives below the waterline
10
Inverse KR
KR flaw A KR that, if maximised, harms the goal it serves.Fix: write the KR for the underlying change, not the proxy. The orca looks back over its shoulder at its own wake
11
Lagging-Only
Set flaw Every KR measures something only visible after the quarter ends.Fix: add at least one KR observable by mid-quarter. A tiny raft overloaded with a teetering stack of boxes
12
Too-Many-KRs
Set flaw An Objective dragging more than five Key Results.Fix: cut to 3-5. The cut KRs become next-quarter candidates, not deletions.

Your key. Your data. Your browser.

OKR Orca has no backend. Your OKR text goes directly to OpenAI or Anthropic. Add your own API key, stored in your browser only.

Key stays local
Stored in localStorage. It never touches my server.
Direct to provider
Calls go from your browser straight to OpenAI or Anthropic. No proxy, no middleware.
Free tier available
3 diagnoses per IP per day without a key, via a rate-limited shared proxy.
The OKR Orca mark
Who built this
Frederik Metz
Agile coach Munich

I'm Frederik Metz, an agile coach in Munich. I've reviewed several hundred OKRs over the last few years, mostly from product teams trying to get sharper. OKR Orca is the rubric I use in those sessions, made available as a free tool.