OKRs: What They Are, When They Work, and What It Takes to Adopt Them

OKR stands for Objectives and Key Results. The framework takes five minutes to explain. Getting it to stick in an organization takes about a year. This guide covers the structure and what usually goes wrong along the way.

The structure

An Objective is one sentence about what you want to achieve. It is qualitative, and people should be able to remember it without looking it up.

Key Results are two to five numbers that decide whether you got there. They are quantitative, and they measure outcomes rather than activity.

An example from customer success

Objective: Customers get an answer before they have time to get annoyed

• KR1: Median response time down from 4 hours to 30 minutes
• KR2: 95% of enquiries resolved without a handover
• KR3: NPS up from 45 to 65

The point of splitting it in two is that you can argue about the wording in January instead of about the result in April.

Where the framework comes from

Andy Grove developed the practice at Intel in the late 1960s, as a tighter version of Peter Drucker's Management by Objectives. Grove's additions were a quarterly cycle instead of an annual one, and a requirement that every objective carry measurable results. In High Output Management he describes the method as MBO. The term OKR came later.

John Doerr worked at Intel under Grove and carried the method with him as an investor. In 1999 he presented it to Google, which had around forty employees at the time. Google still uses it and has published parts of its thinking through re:Work. Doerr's 2017 book Measure What Matters made the framework known outside the technology industry.

A warning about the adopter lists

You will see lists of companies that "use OKRs", and they are unreliable. Several of the names that recur come from vendor blogs citing one another, and some of the companies have said the opposite about their own practice. Google and Intel are well documented. For most other names the documentation is thin. And that a large company uses a framework is a weak argument that it fits yours.

Why they work, and where the ground is thinner than claimed

The limit creates focus. Three to five Objectives per team per quarter. It is the limit itself that does the work. Twelve priorities is a wish list.

Openness creates coordination. When every goal is visible, conflicts between departments surface during planning rather than in March.

Short cycles create adaptability. With quarterly goals you are never more than twelve weeks from a fresh start, and that makes being wrong cheaper.

Now for the part that is usually oversold.

Stretch goals. OKR practice says goals should be set high enough that 70% attainment is a good result. The argument is that a goal you were sure of hitting was set too low. This is practice from Intel and Google. It is not a research finding, and you should know the difference before you defend it in a leadership meeting.

Locke and Latham's goal-setting theory (A Theory of Goal Setting and Task Performance, Prentice Hall, 1990) shows that specific and demanding goals produce better performance than vague or easy ones — provided people have committed to them and have the ability to reach them. That is a different claim from 70% being the optimal attainment rate. In Goals Gone Wild (Academy of Management Perspectives, 2009), Ordóñez, Schweitzer, Galinsky and Bazerman went through the documented side effects of aggressive goals, among them increased risk-taking, weakened cooperation and the temptation to massage the numbers. Both are worth having read before you settle on an ambition level.

In practice that means stretch goals suit new ground, where nobody knows what is possible. On a goal you have promised a customer, 70% attainment means the customer did not get what you promised. Use both kinds deliberately, and say which kind it is when the goal is set.

Writing Objectives

A good Objective is qualitative and short enough for people to remember. It should also sit inside what the team actually controls.

Weak

• "Increase revenue by 20%" — this is a Key Result. It is missing an Objective above it that explains why.
• "Improve things" — nobody can act on this.
• "Hope the market turns" — outside the team's control.
• "Dominate the enterprise market" — sounds ambitious, but nobody can say what would make it achieved.

Stronger

• "Become the platform that is fastest to get started with"
• "Make enterprise customers our largest segment before the year is out"
• "Remove the reasons new customers leave in their first month"

The second set is not written with more energy. It is written so that you can agree on which numbers to measure.

Writing Key Results

A good Key Result is a number that measures an outcome. It also needs a starting value.

Weak

• "Launch the new dashboard" — an activity. The launch can happen without anything getting better.
• "Improve customer satisfaction" — no number.
• "No regretted attrition" — the team does not control this, and the wording makes it impossible to score anything but zero or full marks.

Stronger

• "Daily active users as a share of monthly, up from 35% to 55%"
• "P1 defects down from 120 to 20"
• "Share of new customers who finish setup: up from 61% to 85%"

Test for the activity trap: ask "and then?". If the answer to "we launched the dashboard" is "and then usage went up", the second half is your Key Result.

Always write down the starting value. Without it nobody can score the goal, and you find out in week ten that nobody measured the baseline.

Examples by department

Sales

Objective: Make the enterprise segment our most important source of revenue

• Contracts above $100,000 in annual value: from 12 to 50
• New annual revenue from enterprise customers: $5 million
• Average deal size: from $75,000 to $125,000

Product

Objective: Make the mobile experience the reason people choose us

• Daily active users as a share of monthly: from 30% to 45%
• App store rating: 4.5 with at least 10,000 reviews
• Share of new users who finish setup on mobile: from 40% to 70%

Engineering

Objective: Stop reliability being something customers notice

• Uptime: from 99.7% to 99.95%
• P1 incidents: from 15 to 2 per month
• Page load time, 95th percentile: under 1.5 seconds

Marketing

Objective: Become where people go to understand goal management

• Monthly visitors to the guide pages: from 250,000 to 1 million
• Qualified leads from content: 15,000
• Talks at industry conferences: 50

People

Objective: Build an engineering environment people seek out

• Engineers hired: 25, at least 15 of them senior
• Time from posting to signature: from 65 to 35 days
• Share of candidates who accept an offer: from 60% to 85%

Note that every Key Result has a starting value. Without one you do not know whether 45% is a doubling or a decline.

Cadence

Company annual goals are set in the fourth quarter for the year after. Three to five Objectives, owned by the leadership team.

Team quarterly goals are set in the last week of the preceding quarter. Three to five per team.

Monthly review, 30 to 60 minutes per team. Progress and the team's confidence in reaching the goals are updated here, and blockers are escalated. The goals themselves are not changed here. Change the goals every month and you have a log rather than a goal.

Quarterly assessment in the last week. Every Key Result is scored, and the team talks about what it learned. Half a day for the leadership team, two hours per team.

Week	What happens
1	Start. Each team explains how its work connects to the goals
4–5	First review, first estimate of whether the goals will be met
8–9	Second review. This is where the decision to abandon something is made
12	Final push and assessment
13	Retrospective and next quarter's goals

Put the most weight on the week 8 review. It is the last point at which you can still move people and money and change how the quarter ends.

Scoring

Key Results are scored from 0.0 to 1.0. For numeric goals the score is the share of the distance you covered:

(actual − start) / (target − start)

A goal of going from 500,000 to 750,000, where you ended at 625,000, gives 0.5.

For binary goals it is 0 or 1. Use them sparingly, since binary Key Results are usually activities in disguise. The Objective score is the average of its Key Results.

How to read the score. Google aims for an average around 0.6 to 0.7. A steady average of 1.0 suggests the goals are being set too low. If the average sits below 0.4 over time, the cause may be unrealistic goals or problems with execution, and the two call for completely different responses. The average alone does not tell you which. Look at whether it is the same teams missing every quarter, and on which kinds of goals.

A score is not a performance review

Tie OKR scores to pay or appraisal and people will set goals they know they will hit. That is a rational response to the incentives they are given, and it destroys the framework. Keep the two processes apart, and say so out loud every quarter.

OKRs or SMART goals

They solve different problems, and most organizations need both.

	SMART goals	OKRs
Used for	Concrete deliverables and operations	Strategic direction and change
Level	Individual and project	Company, department, team
Ambition	Meant to be fully met	70–80% is a good result
Horizon	Flexible	Quarter or year
Form	One goal statement	One Objective, several Key Results

Used together: the company OKR is "Make enterprise customers our largest segment". A SMART goal for an individual is "Complete ISO 27001 certification by 15 March", which is one of the things that has to happen on the way there.

The seven mistakes

Too many goals. Ten Objectives per team means no prioritization. Set the limit at five and enforce it. The uncomfortable choice about what drops out is the whole value.

Key Results that are activities. "Launch the feature" and "Hire five engineers" are tasks. Ask "and then?" until you reach the outcome.

Safe goals. Teams set goals they know they will hit. Asking them not to does not help, because the cause sits in the incentives. Disconnect the goals from the appraisal system.

Set and forget. The goals are written in planning week and read again in week thirteen. The remedy is that the reviews are in the calendar before the quarter starts, and that leadership meetings open with goal status.

Goals that hang off nothing. A department goal that cannot point to which company goal it serves should be connected or struck. If it cannot be connected and still feels important, something is missing from the company goals.

Operations dressed as goals. "Maintain 99% uptime" is an operating commitment that belongs somewhere else. An improvement goal, like going from 99.7 to 99.95, does belong here.

The demand for 100%. A manager who expects full attainment has made stretch goals impossible, whatever the handbook says. This cannot be solved anywhere other than in the leadership team.

The first quarter

• Four to six weeks before: the leadership team learns the framework. Pick one or two teams to try it first. Decide where the goals will live.
• Two to three weeks before: the leadership team drafts three to five company goals. Department heads read and comment. The goals are presented to everyone.
• The last week: each department writes its own goals. Then a joint review where departments read each other's and look for conflicts. Then everything is entered and made visible.
• During: start in week 1, reviews in weeks 4 and 8, assessment in week 12.

What to expect. The first quarter will be poor. There will be too many goals, half the Key Results will turn out to be activities, and a team or two will discover late that they never measured the baseline. This is normal, and more training up front does not prevent it. The retrospective after the first quarter is where the learning actually happens. Expect three to four quarters before this feels natural.

What decides it. If the leadership team does not set its own goals, show them openly and score itself as strictly as everyone else, OKRs become a reporting exercise for the people below. An all-hands does not fix that.

Setting OKRs up in Markviss

The structure in Markviss follows the framework directly:

Markviss	OKR
Account	Company or team
Category	Objective
Metric	Key Result
KPI with a formula	Attainment rate

1. Account structure. One team gets by with one account. For a whole company: a parent account with sub-accounts per department, connected so numbers roll up.
2. Categories. Create one category per Objective. Use the description field for the wording itself.
3. Metrics. One metric per Key Result, with a starting value and a target value.
4. Scenarios per quarter. Create "Q1 2027", "Q2 2027" and so on. Enter planned values, update actuals monthly, and switch scenario to see history.
5. A KPI for attainment. Formula: (actual − start) / (target − start)
6. Dashboard. Put the categories and KPIs on a dashboard everyone can reach. The colour coding gives status without anyone having to open anything.

In short

OKRs give you a structure for agreeing on few enough priorities, making them visible, and measuring them in a way nobody can argue about afterwards. They do not decide what the company should bet on, and they do not make a team faster.

Whether they stick depends on whether the leadership team can sit through three quarters of mediocre attempts, and whether they can avoid tying the scores to pay.

Start with one team and one quarter. Write down the starting values before you begin.

Related Resources

Learn how OKRs fit into your organizational structure in our Goal Hierarchy Guide, or see the complete framework comparison in Metrics, KPIs, and OKRs.