How to Write OKRs — Objectives That Inspire, Key Results That Measure

2026-07-11 · Norolu Frameworks · 4 min read

OKRs are simple on paper and easy to get wrong in practice. The two most common failures are writing Key Results that are really a to-do list, and setting an Objective that is just a number with no direction. This guide shows how to write an Objective and Key Results that work together, where they should come from, how to grade them, and the mistakes to avoid — with one concrete example throughout.

Norolu Frameworks SWOT analysis screen. A café example shows strengths, weaknesses, opportunities and threats, plus the Cross-SWOT strategies derived from them — all in one screen (actual screenshot).
Actual screen: 3C → SWOT → Cross-SWOT → MECE → OKR, all in one screen. Your strategy data stays on your device.
The strategy cascade: 3C (scan) → SWOT (organize) → TOWS (derive strategy) → MECE (break down) → OKR (measure) 3C SWOT TOWS MECE OKR
The strategy cascade: 3C (scan) → SWOT (organize) → TOWS (derive strategy) → MECE (break down) → OKR (measure)

What an OKR is — Objective vs. Key Results

An OKR has two parts. The Objective is a short, qualitative, ambitious statement of what you want to achieve — it should be memorable and give direction (“Become the go-to English course for busy professionals”). The Key Results are the measurable outcomes that prove you got there (“raise 30-day retention from 20% to 40%”).

The load-bearing rule: an Objective is qualitative and inspiring; a Key Result is quantitative and verifiable. If your Objective already contains the number, it is probably a Key Result in disguise; if your Key Result has no number, it is probably a task.

Where OKRs come from — strategy, not a wish list

Good OKRs are not brainstormed from scratch each quarter. They are the measurable end of a strategy you already chose. That is why the frameworks connect: a 3C analysis feeds a SWOT, Cross-SWOT (TOWS) derives the strategy, MECE breaks it into workstreams, and each workstream becomes an Objective with Key Results.

Coming from a decomposed strategy is what keeps your OKRs both complete (no important area missing) and focused (you are not inventing goals unrelated to your direction).

Step 1: Write the Objective

Keep it qualitative, ambitious, and time-bound — usually one quarter. Take an online language school as the running example:

Aim for one to three Objectives at a time. More than three, and nothing is really the priority.

Step 2: Write Key Results — outcomes, not tasks

Each Objective needs three to five Key Results. A Key Result describes a change in the world, not an activity you performed. The test: could you finish the task and still miss the goal? Then it is a task, not a Key Result.

Give every Key Result a starting value and a target (“from 20% to 40%”). Without the baseline you cannot tell how far you have moved.

Step 3: Grade, set a cadence, and hand off tracking

Score each Key Result from 0.0 to 1.0 at the end of the period. For ambitious (“stretch”) OKRs, one widely used convention treats around 0.7 as healthy — consistently scoring 1.0 can mean the targets were set too safe. For OKRs you have committed to fully deliver, aim for 1.0. Decide which kind each one is up front, so the score means something.

Set OKRs for a quarter and check in regularly (a quick weekly review of confidence). Setting and grading is a design activity; the day-to-day number tracking that feeds those scores is a different job. Our sister app Baton Board is built for that ongoing KPI tracking — this cockpit focuses on designing the OKR and hands the daily tracking off to it.

Our strategy cockpit runs the whole cascade — 3C → SWOT → Cross-SWOT → MECE → OKR — on one screen, with each step feeding the next. The OKR view nudges you to keep the Objective qualitative and every Key Result measurable. Everything you type is stored on your device only (nothing is sent to our servers), and the AI drafting assistant is free to try on SWOT — bring your own API key (BYOK); you pay only your provider's usage. Unlocking AI across every framework is a one-time $9.90.

FAQ

How many OKRs and Key Results should I have?

One to three Objectives at a time, with three to five Key Results each. Fewer, sharper OKRs beat a long list — the point of OKRs is to make priorities obvious, and too many defeats that.

What is the difference between a KPI and an OKR?

A KPI is an ongoing health metric you monitor continuously (e.g. monthly churn). An OKR is a time-boxed goal to change something over a quarter. A KPI can become a Key Result when you set a target to move it; otherwise you just watch it.

What score counts as good?

It depends on the type. For ambitious stretch OKRs, around 0.7 out of 1.0 is often considered healthy, and always hitting 1.0 may mean you aimed too low. For committed OKRs you intend to fully deliver, the target is 1.0. Agree the type before you start so the grade is meaningful.

What is the difference between a strong Objective and a weak one?

A strong Objective is qualitative, sets a direction, and moves people (e.g., "Become the default choice for busy professionals learning English"). A weak Objective is either too vague ("make the product a bit better") or already has a number baked in, blurring the line with a Key Result. If your Objective already contains a number, it may actually be a Key Result in disguise.

How do I tell a Key Result apart from a task?

Ask whether completing the work could still leave the goal unmet. If yes, it's a task, not a Key Result. "Ship the mobile app" is a task — shipping it doesn't guarantee the goal is hit. Measure the number that shipping it is supposed to move instead, like conversion rate.

Other guides

See all guides →

Build your OKRs in the browser — free, no sign-up

3C → SWOT → Cross-SWOT → MECE → OKR on one screen. The OKR view keeps Objectives qualitative and Key Results measurable. Strategy data stays on your device; the AI draft assistant is free on SWOT (with your own API key — BYOK).

Open the app →

AI agents such as Claude can write 3C, SWOT, MECE and OKR directly into your local data. AI drafts, you decide. AI agents (MCP) →