Answers
How to write a key result that actually measures something.
Most key results are tasks wearing a number. The test that separates the two is simple and unforgiving: on the last day of the quarter, can this be settled as true or false without anyone arguing about it? If the answer needs a conversation, it is not a key result yet. This page is about writing the ones that pass.
- Set up with your plan
- Unlimited users
- 30-day pilot
The three ways they fail
Almost every bad key result is one of these.
They look different on the page and they fail the same test, which is why fixing them is the same move each time.
Failure modes
A task in disguise
Launch the new website. Hire a sales manager. Roll out the training. These have a completion date and no outcome, so finishing them settles nothing about whether the objective moved. The giveaway is that you could do it badly and still tick it.
Activity instead of outcome
Make 200 calls. Publish 12 posts. Run 30 demos. These are honest and countable, and they can all be true while the objective is exactly where it started. Effort you control, standing in for a result you do not.
Unsettleable
Improve customer satisfaction. Strengthen the brand. Increase engagement. Two reasonable people will disagree in December about whether these happened, which means the key result cannot do the only job it has.
The rule underneath
One test, applied honestly.
Writing one
True or false, on the last day, without an argument
A key result is a claim about the world that will be either true or false at the end of the period. Not a plan, not an intention, not a body of work. A claim.
The clause that does the work is without an argument. Plenty of badly written key results can technically be settled, if someone is willing to litigate the definition in the review. If settling it requires interpreting what you meant, the review becomes a negotiation, and a measure that can be negotiated is not measuring anything.
- Name the number, the direction and the deadline, so there is nothing to interpret
- Say where the number comes from, because an undefined source is an argument waiting to happen
- Write it so that someone who was not in the room could score it
Outcome first, but activity is not always wrong
The usual advice is outcomes never activities, and it is right most of the time and wrong often enough to be worth qualifying. An activity target is legitimate when you genuinely do not yet know the conversion rate, and the quarter's real purpose is to find out.
The honest version of that is to say so on the key result. Run 30 discovery calls and publish the conversion rate is a different commitment from run 30 discovery calls, because the second could be satisfied by 30 bad calls and the first could not. If you are buying information, make the information the deliverable.
- Prefer the outcome whenever the link between activity and outcome is already known
- Where it is not known, make learning it the key result rather than pretending
- Never carry an activity target into a second quarter unchanged. By then you know
You cannot move a number you have never measured
Increase X from 40 to 60 requires that 40 is real and that someone can produce it again in twelve weeks the same way. A surprising number of key results are written against a baseline that was estimated once, in a meeting, by the person who wanted the target.
When the baseline does not exist, the first quarter's key result is to establish it. That is not a wasted quarter. It is the difference between a plan and a wish, and it is usually the cheapest quarter you will spend.
- Write down the baseline and the date it was taken, beside the target
- Name the system the number comes out of, not the person who fetches it
- If there is no baseline, make establishing it the key result and say so plainly
Worked examples
The same objective, written badly and then properly.
Four functions. In each, the first version fails the test and the second passes it, and the fix is almost always smaller than people expect.
Examples
Sales
Not a key result: grow the pipeline. It has no number, no baseline and no source. A key result: qualified pipeline for the second half rises from $1.4m at 30 June to $2.1m at 30 September, measured from the CRM on the last working day.
Operations
Not a key result: reduce delivery delays. Reduce by how much, from what, measured how? A key result: jobs delivered on or before the promised date rise from 78 per cent to 90 per cent, measured across every job closed in the quarter.
Finance
Not a key result: improve cash flow. Two people will score this differently in the review. A key result: average debtor days fall from 48 to 35, taken from the aged receivables report on the last day of the quarter.
People
Not a key result: strengthen the team. Unsettleable, and also not really about anything. A key result: every one of the nine roles on the accountability chart has a named owner and a written scorecard by 30 September.
A working order
Four steps, in this order.
Writing the number first is the most common mistake, because a number chosen before the outcome is a number chosen to be achievable.
The method
1. Name what changes
Not what you will do. What will be different about the business if the quarter goes well. One sentence, no verbs like improve or optimise.
- See OKRs vs KPIs
2. Find the measure
What number would move if that change were real? If nothing would move, the objective is probably not a change, and it may be a KPI you are already monitoring.
3. Get the baseline
Pull it from the system it lives in, write it down with the date, and name the report. If it does not exist, establishing it is this quarter.
4. Set the target and the source
From X to Y by this date, measured from that report. Then read it back and ask whether a stranger could score it without calling you.
Questions
Writing key results, answered
What makes a good key result?
It can be settled as true or false on the last day of the period without anyone arguing. In practice that means it names a measure, a baseline, a target, a deadline and the report the number comes from. If any of those five is missing, the review turns into a negotiation about what was meant.
What is the difference between a key result and a task?
A task is something you do and a key result is something that becomes true. Launch the new website is a task: you can complete it badly and still tick it, and completing it settles nothing about whether the objective moved. Sign-ups from the website rise from 40 a month to 100 a month is a key result, and it happens to require the website.
Can a key result be a yes or no rather than a number?
Yes, if the binary is genuinely unambiguous and the thing is genuinely a change rather than a task. Every role on the accountability chart has a named owner and a written scorecard by 30 September is binary, settleable and not a task list. Launch the new brand is binary and is a task. The test is the same one either way.
Should key results be activities or outcomes?
Outcomes, wherever the link between the activity and the outcome is already understood. Where it is not, an activity target is defensible for one quarter provided you make learning the conversion rate part of the commitment rather than a hope. Carrying the same activity target into a second quarter unchanged means you did not learn anything, which is its own finding.
What if we have no baseline for the measure?
Then the key result for this quarter is to establish it, stated plainly rather than disguised as a target. Writing increase from 40 to 60 when the 40 was guessed in a meeting produces a review where everybody argues about the 40. Measuring properly for one quarter is the cheapest quarter you will spend.
How many key results should an objective have?
Few enough that each one matters. The broader question of how many objectives and measures an organisation should carry at once is covered on the OKRs versus KPIs page, which puts it at three to five. OKRs vs KPIs.
Does this apply if we use Rocks or a scorecard rather than OKRs?
Yes, unchanged. The test is about whether a commitment can be settled, not about which framework the commitment sits in. A Rock that cannot be scored without an argument has the same problem as a key result that cannot, and Empiraa GPS is deliberately framework agnostic for that reason.
Still deciding? Ask ANI and get a straight answer.
See what a quarter looks like when every measure can be scored
GPS holds the plan, the owners and the measures in one place, and reports from live data rather than from a slide someone rebuilt the night before.