What Are OKRs?

The method comes from Intel. Andy Grove, who later became Intel's CEO, described it in his book High Output Management through two questions: where do I want to go, and how will I pace myself to see if I am getting there? The first answer is the objective, the second gives the key results. John Doerr learned the method from Grove at Intel and took it to Google. According to Google's re:Work guide, he introduced it to Google's leadership in early 2000; Doerr's own company, What Matters, gives the year as 1999.

The objective is qualitative. It describes a change that matters, such as "the blog brings in customers, not just visits". Google asks for objectives that are ambitious and feel somewhat uncomfortable. Key results are quantitative. Each names a figure that can be checked at the end of the period, such as "from 40 to 70 demo requests a quarter that start on the blog". If nobody can say at the end whether it was achieved, it is not a key result.

Google's guide recommends a small number: three to five objectives for an organisation, with about three key results each. More objectives spread the team's effort thin. OKRs are public inside the company, so every team can see what the others are working on and align its own goals with the organisation's.

A KPI is an indicator that is watched continuously, such as a landing page's conversion rate or a campaign's ROI. It tells you how the business is doing. A key result sets how far that indicator should move in a given period. A KPI can measure a key result but does not replace it: "conversion rate" is a KPI, while "raise the demo conversion rate from 2% to 3% this quarter" is a key result.

A cycle lasts one quarter. At the start, leadership sets the organisation's OKRs and each team proposes its own. Google combines both directions: it starts from the company's priorities and lets teams say where they can contribute most. Halfway through the quarter, progress is reviewed, and an objective that is clearly not going to happen can be dropped.

At the end of the quarter, OKRs are graded. Google uses a scale from 0.0 to 1.0, where 1.0 means the objective was fully achieved. Each key result is graded, and the objective gets a rough average. Some key results can only score 0 or 1 (the campaign launched or it did not); others are graded in part. If the key result was to launch six features and three launched, the grade is 0.5.

Google's guide puts the sweet spot at an average of 0.6 to 0.7. A lower score may mean the organisation is not achieving as much as it could. A score that is always high means the goals were not ambitious enough. A low grade is not a penalty: Google treats it as data for the next set of OKRs and explicitly separates OKRs from individual performance reviews.

An invented marketing example: - Objective: the blog brings in customers, not just visits. - Key result 1: from 40 to 70 demo requests a quarter that start on the blog. - Key result 2: 25% of organic blog visits reach a product page. - Key result 3: the click-through rate on the calls to action in the five highest-traffic guides rises from 1% to 2%.

Google's guide is blunt about it: OKRs are not a checklist, nor a master list of everything the team will work on in the quarter. A team that treats them as a shared to-do list describes what it wants done rather than what it wants to achieve. Compare two versions of the same objective. With tasks: "publish 12 articles", "redesign the blog template", "send three newsletters". All three can be done by the end of the quarter without a single extra customer. With results: "from 40 to 70 demo requests from the blog". The tasks still exist, but the team chooses and changes them according to what moves the number. Google does allow yes-or-no key results such as launching a campaign; still, the more key results are tasks, the less the final grade says about the objective.

OKRs force a team to name the change it wants before deciding what to do. In digital marketing nearly everything can be measured, which makes it easy to measure activity alone: articles published, emails sent, ads running. A key result asks something else: what has changed for the business or for the customer?

They also sort priorities. With three public objectives, a team can turn down a request that moves none of them forward. And because the expected grade is 0.7 rather than 1.0, a team can set ambitious goals without a 0.7 counting as failure.

OKRs do not fit everything. For maintenance work that simply has to keep running, KPIs with a threshold are enough. Google's guide calls the alternative "business-as-usual OKRs" and lists them as a mistake: an objective written around what the team already does, without changing anything.

• Write the objective as a change that people outside the team will notice, not as an activity.
• Give every key result a starting value, a target value and a date.
• Keep the number down: few objectives and about three key results each.
• Review at mid-quarter and write down what the team learned.
• Grade honestly, and always by the same standard.

• Turning key results into a task list: "publish 12 articles" measures effort, not results.
• Setting ten objectives at once.
• Writing key results without a number that can be checked.
• Using OKR grades to assess individual people.
• Confusing a KPI you want to hold steady with an objective meant to change something.