In an analysis of 20,952 key results written by real teams, 52% described a task or a KPI in place of an outcome. Key results with an explicit deadline hit their target 67% of the time, against 51% without one. This guide covers each step of writing an OKR, from the Objective to the pre-cycle quality check, with data from real teams behind every recommendation.
The first time I was asked to write an OKR was one of the most frustrating tasks I've ever done. We had rolled out OKRs company-wide, and it was my turn to write one and submit it for approval. The guidelines were simple: the Objective should be ambitious and the Key Results measurable. It took me seven attempts, each one written, submitted, reviewed and rejected, and the whole exercise ate half a day. We were a team of 30+ people and I was considered one of the fastest, which meant writing OKRs for the entire organization took two to three weeks.
We were writing OKRs by the book, the literal book: Measure What Matters and the examples John Doerr gave to Google. The problem was that we weren't Google, and OKRs were a completely new framework for us, which came with its challenges. Today, AI can draft a reasonably good OKR in seconds.
What this guide adds is data from real teams and real OKRs: 20,952 key results from 876 organizations on the OKRs Tool platform, plus benchmark research across hundreds more. It's the guide I wish I'd had when I wrote my first OKR.
Why OKRs Fail Before the Cycle Starts
A large share of OKR failures are written into the goal on day one. In our analysis of 20,952 key results on the OKRs Tool platform, 52% were tasks or KPIs in disguise: they measured what a team did, or a number it already tracks, and said nothing about what changed.
The five verbs most common among underperforming key results were Complete, Launch, Conduct, Implement and Test. Each one describes work. A key result built on one of them can be finished in full while the business outcome it was meant to move stays exactly where it was, which is one of the main reasons OKRs fail in their first year.
One test catches most of the KPIs before the cycle starts. Ask of every key result: can I track this metric every week, forever, without it ever being done? If the answer is yes, it's a KPI, and it belongs on a KPI dashboard. A key result describes a time-bound change from a baseline to a target. Both matter, and the split between KPIs and OKRs keeps each one doing its own job.
The Two Parts of an OKR
An OKR has two parts with one job each. The Objective answers where the team is going, in words. The key results answer how the team will know it got there, in numbers, with two or three per Objective.
Step 1: Write a Strong Objective
An Objective answers one question: what does success look like at the end of this quarter? Projects, targets and metrics all come later, in the key results and the work underneath them. Strong Objectives share four traits.
- Qualitative. The Objective carries no numbers. "Grow revenue by 30%" is a key result, and the Objective above it describes why that growth matters.
- Inspiring. The best Objectives describe a changed state worth rallying a team around for twelve weeks.
- Time-bound. The quarterly cycle sets the deadline, but the wording should still feel urgent. "Improve customer satisfaction" could run for years; "Make onboarding so clear new users don't need support" points at a finish line.
- Outcome-oriented. "Run the Q3 product launch" describes activity. "Make the product launch the most successful in company history" describes what should be different when it's over.
The examples below show weak Objectives rewritten using those four traits, and the difference between objectives and goals goes deeper on the wording.
Step 2: Write Key Results With a Baseline, a Target and a Date
A key result is a specific, measurable change from a baseline to a target that proves the Objective was achieved. Tasks, project milestones and KPIs the team already tracks all fail that definition, and they make up the 52% of key results that measure activity.
The classic template looks like this:
Increase [outcome] for [segment] from [baseline] to [target] by [date]
When we broke 11,300 scored key results on the platform into the parts of that template, one part stood out from the rest.
Key results with an explicit timeframe hit their target 67% of the time, against 51% for those without one. A number on its own made no difference, and a percentage added only four points.
The baseline and target still matter, because without them nobody can score the key result at the end of the quarter. The deadline is the part that predicts a hit, so every key result needs a date someone can be held to, even when the quarter seems to imply one.
The Output Trap
The left column below shows tasks dressed up as key results. Each one on the right measures the change that task was meant to cause, and the task moves down a layer to become an initiative.
The KPI Test
A key result that holds a threshold is a KPI. "Maintain NPS above 40" describes a level to protect, and it passes the forever test: a team can track NPS above 40 every week indefinitely. "Increase NPS among enterprise accounts from 32 to 50 by end of Q3" describes a change with a start, an end and a deadline, which makes it a key result.
Step 3: Write Each Key Result So Anyone Could Score It
A short key result leaves its meaning in the owner's head. "Grow revenue" could mean new revenue, expansion revenue or total revenue, for one segment or all of them, and three people reading it at the end of the quarter could score it three different ways.
The platform data shows a steady climb in hit rate as key results get more descriptive.
Key results of one to three words hit their target 35% of the time, rising to 69% at sixteen words or more. Teams that hit their goals write a median of 14 words per key result, against 10 for teams that miss.
The gap sits mostly between companies, so word count marks a team that thinks carefully about what it commits to, and padding a key result with extra words won't change its odds. The useful habit is the thinking itself: name the metric, the segment, how it will be measured and by when, until a new hire could score it without asking.
The free OKR Quality Grader checks a draft for missing baselines, vague metrics and task verbs before it goes live.
Step 4: Give Every Key Result One Named Owner
Every key result needs one named person before the cycle starts, accountable for tracking it, raising the alarm when it stalls and owning the score at the end. A team or a "leadership group" can do the work, but one person has to answer for it.
The OKR Benchmark Report found that teams requiring single ownership see 26% higher completion rates than those with shared or vague accountability. A key result with no clear owner is the likeliest to go silent, and the OKR Intelligence Report found that 7% of key results are abandoned mid-cycle with no revision or escalation.
On the OKRs Tool platform, the gap is wider at the extremes: top-quartile organizations leave a median of 8% of key results at zero all cycle, while bottom-quartile organizations leave 56%. The fix for key results with no owner starts at the writing stage, before anyone has a reason to assume someone else has it.
Step 5: Calibrate Ambition for the Type of Goal
OKR scoring usually runs on a 0.0–1.0 scale, and the familiar advice says to aim for 0.7–0.8. That band fits stretch goals, where a goal that lands at 100% every quarter probably wasn't stretching anyone. Committed goals, the ones a team has promised to deliver, are supposed to reach 100%, so decide which of the two types of OKRs each one is before setting the target.
Early finishes can come from well-set targets too. On the platform, 43% of the key results that hit their target reached their final value in the first half of the cycle and held it. Check the baseline and the update history before calling one sandbagged.
Ambition also gets easier to calibrate with practice.
Teams in their first two cycles average 51% completion, rising to 59% in cycles three and four and 79% from cycle five onward. Early quarters are where a team learns what a realistic stretch looks like for its own work, which is why the OKR maturity curve rewards shipping an imperfect first draft and improving it at the end-of-cycle reflection.
One calibration test helps in every cycle: if the team hits this at 70%, would that still be meaningful progress? If 70% would feel like a failure, the target belongs in the committed category.
Step 6: Choose the Right Number of OKRs
Teams running one or two Objectives per quarter are twice as likely to achieve them as teams running three or more, according to the OKR Benchmark Report. Each Objective works best with two or three key results, which puts a focused team at two to six key results for the quarter, in line with the benchmark on how many key results per objective.
The same limit applies at every level of a cascading OKR structure. Company OKRs set one or two priorities, departments interpret their contribution, and team key results name the specific outcomes. When the number is bounded at each level, the whole organization can see what matters most this quarter.
Step 7: Separate Initiatives From Key Results
The most common structural error in OKR writing is treating the work as the goal. Initiatives are the campaigns, sprints and experiments that move a key result. They belong in the OKR, underneath the key result they serve.
Attaching two or three initiatives to each key result in the first week of the cycle gives the owner a plan to report against at the first check-in, and keeps the key result itself focused on the outcome.
Step 8: Run the Pre-Cycle Quality Check
Before any goal goes live, run it through seven checks.
A key result that passes all seven is a real business commitment. One that fails more than two needs rewriting before the cycle starts, while there's still time to fix it without losing a week of the quarter.
OKR Writing Examples by Function
Each example below follows the same pattern: a qualitative Objective, key results with a baseline, a target and a date, and one named owner per key result.
Sales
Objective: Build the enterprise pipeline that funds next year's growth
- Increase enterprise MQL-to-SQL conversion from 18% to 32% by end of Q3 (Owner: Sales Manager)
- Reduce average sales cycle from 95 days to 65 by end of Q3 (Owner: AE Lead)
- Increase demo-to-proposal rate from 45% to 70% by end of Q3 (Owner: VP Sales)
Product
Objective: Make onboarding so clear new users don't need support
- Increase Day-7 activation from 34% to 52% by end of Q3 (Owner: Product Lead)
- Reduce time-to-first-value from 6 days to 2 by end of Q3 (Owner: CX Lead)
- Reduce onboarding support tickets from 220 to 130 per month by end of Q3 (Owner: PM)
Marketing
Objective: Make organic search a predictable acquisition channel
- Grow organic MQLs from 90 to 180 per month by end of Q3 (Owner: Content Lead)
- Increase top-3 keyword rankings from 14 to 35 by end of Q3 (Owner: SEO Lead)
- Grow organic share of total traffic from 28% to 40% by end of Q3 (Owner: Head of Marketing)
Engineering
Objective: Build a platform customers can depend on without thinking about it
- Reduce P1 incidents from 8 per month to 2 by end of Q3 (Owner: Engineering Lead)
- Reduce mean time to recovery from 45 minutes to under 10 by end of Q3 (Owner: SRE Lead)
- Reduce failed deployments from 18% to under 5% by end of Q3 (Owner: VP Eng)
The same structure scales to every team, and our library of key results examples applies it to 108 key results across nine functions, with deeper sets for sales teams and marketing teams.
How to Write OKRs With AI
AI has changed the first step of writing an OKR more than any other. A prompt with the company strategy and a team's remit returns a reasonable draft in seconds, which would have saved me most of that half day. The draft still needs the same checks as one written by hand, and AI tends to fail three of them.
- The baseline. AI doesn't know your current numbers, so it either leaves the baseline out or invents one. Fill it in from your own data.
- The deadline. Drafts often describe a direction with no date attached, and the date is the element most closely linked to a hit.
- The owner. AI can suggest a role, but only the team can name the person who will answer for the number.
Treat an AI-written OKR as a strong first draft and run it through the quality check above. An AI-powered OKR tool that flags task verbs and missing baselines as the draft is written shortens the review that used to take my team seven rounds.
The Most Common OKR Writing Mistakes
Writing Objectives with numbers. "Grow revenue by 30% this quarter" is a key result. The Objective above it sets the direction: "Build the revenue engine that funds next year's growth."
Confusing initiatives with key results. "Launch the new pricing page" is an initiative. "Increase trial-to-paid conversion from 11% to 18% by end of Q3" is the key result that launch is supposed to move.
No baseline. "Increase NPS to 50" can't be scored, because nobody knows how far it moved. "Increase NPS from 32 to 50" can.
No deadline. A key result with no date hits its target 51% of the time on the platform, against 67% for one with an explicit timeframe.
Shared ownership. "Sales and Marketing jointly own this key result" leaves each side assuming the other is watching it. Name one person before the cycle starts.
Too many Objectives. Teams with one or two Objectives are twice as likely to achieve them as teams with three or more, so cut the list before adding key results to it.
Copying last quarter's OKRs. A goal that appears unchanged for the third cycle running is either stuck, which needs a new approach, or a health metric that belongs on the KPI dashboard.
Video: Three Tools for Writing Better OKRs
The checks in this guide are easier to run when the software does part of the work. In the video below, we share three tools in OKRs Tool that our clients use most when they write a new cycle's OKRs.
- Templates. A library of ready-made OKR templates gives every team a starting point, so the first draft takes minutes and the time goes into adapting it to the team's own baseline and targets.
- AI Coach. Teams can ask questions about their own OKRs, such as whether any key results are KPIs in disguise, and get an answer based on what they wrote.
- OKR Analysis. Before a cycle goes live, OKR analysis reviews every key result in it, so weak ones get rewritten while there's still time to change them.
The video also covers how we encourage clients to use AI. It works best as an accelerator that gets a team to a good draft faster, while the decisions about what matters this quarter and who owns each key result stay with the team.
Ship the First Draft and Improve It Every Cycle
Seven attempts at one OKR felt like failure at the time. Looking back, it was the learning curve every team goes through, made slower by examples written for a company that looked nothing like ours. The data now points to a shorter list of things that matter at the writing stage: an outcome, a baseline, a date, a description precise enough to score and one named owner.
Writing is only the start of the quarter. Teams that keep a weekly check-in habit complete 43% more OKRs, according to the OKR Benchmark Report, and The ROI of OKRs Benchmark Report puts the return on a well-run program at 1:25 across 330 organizations. A platform that checks each key result for an owner, a baseline and a target before it goes live, then asks every owner for a weekly update, turns a well-written OKR into one that gets done.
Data: OKRs Tool platform data, The OKR Benchmark Report, The ROI of OKRs Benchmark Report and the OKR Intelligence Report. View all research.
