Rules
Part of Strategic planning: steps, examples and decisions for 2027
Strategic planning metrics: practical details and examples
Strategic planning measures chosen for their lag: naming the distortion at adoption, measuring decisions as well as results, ownership, and retiring a measure.
The measures attached to a plan are usually chosen for a reason nobody says out loud: they are the ones already being collected. That is why the same handful appears on every plan regardless of what the plan decided, and why the review meeting can be full of numbers and produce no decision.
A measure earns a place on a plan only if you can state three things about it before you publish it: when it will move, how it will be distorted, and who has to act when it does something unexpected. This page is about getting those three right.
What to take away
- State the lag before you choose the measure. A number that cannot move for nine months cannot tell you anything in month three, and it will be stared at anyway.
- Every measure with consequences gets pursued instead of the thing it stands for. Name the likely distortion in writing when you adopt it, not after you see it.
- Measure the decisions as well as the outcome. Whether an assumption was tested by its stated date is checkable next week; whether the bet was right is not.
Lag is the property everyone skips
A plan usually mixes four kinds of measure with wildly different response times, and putting them on one page without labeling the lag is how review meetings go wrong.
| Kind of measure | When it can move | What it is good for |
|---|---|---|
| Did the thing happen | Days | Confirming the plan is being executed at all |
| Did behavior change | Weeks | Early warning that the mechanism works or does not |
| Did the outcome change | Months to a year | Judging the bet, once |
| Did it last | A year or more | Deciding whether to repeat the approach |
Two failures follow from ignoring this. The first is impatience: an outcome measure is checked monthly, wobbles for reasons unrelated to the plan, and the initiative gets amended repeatedly on noise. Separating movement a measure produces anyway from movement that means something is what a control chart exists to do, and the reasoning transfers to a plan review even where the chart itself does not. The second is complacency: only slow measures are chosen, nothing is expected to move for two quarters, and by the time it does not move, the year is gone.
The fix is to carry one fast measure and one slow one, labeled, and to say plainly which question each can answer and when. Then agree in the room that the fast one is not evidence of success and the slow one is not available yet.
Name the distortion when you adopt the measure
Any measure that carries consequences will be optimized, including in ways that damage what it was standing in for. That is not cynicism about people; it is the reliable effect stated as Goodhart's law. This is not a reason to avoid measuring. It is a reason to write down, at the moment of adoption, the sentence "the cheapest way to make this number look good without doing the work is ...".
That sentence is usually easy to write and almost never written. Once it exists, two things follow. You know which second measure to watch alongside it, because the cheap route always degrades something. And when the number improves, you have a stated alternative explanation to rule out rather than a celebration to interrupt.
Pairing is the practical defense: a volume measure with a quality measure, a speed measure with a rework measure, a growth measure with a retention measure. The rule is not that both must improve. It is that you look at both, always, in the same breath. The quality management side of this is where the paired number usually lives, and it is often owned by a different team, which is exactly why it gets omitted.
Measure the decisions, not only the results
A plan's outcome is unavailable for months. Its execution is visible now, and it is what you can actually correct.
Things worth counting that are available immediately:
- Assumptions tested by their stated date. Every plan rests on claims that might be false. Whether the person who owns each one has reported by the date they agreed is a fact about this week.
- Decisions still sitting. A count of decisions raised and not made, with how long each has waited. Held decisions stop work while everyone waits, and they never appear in any outcome measure.
- Work stopped. If the plan named work that would end, has it ended? This is the single most informative number in the first quarter, and it is almost never tracked, because the plan treated the subtraction as an assumption rather than as a task.
- Reviews that produced a decision. Not reviews held. A standing review that has produced no decision in three sittings has become a status meeting.
None of these tell you whether the bet was right. All of them tell you whether the plan is being executed, and a plan that is not being executed cannot be evaluated at all.
When the measure you want does not exist yet
Frequently the right measure needs data you do not collect, and the honest options are worse than people admit.
You can build the collection, which takes months and delays the answer. You can pick a proxy, which is available now and is not the thing. Or you can go and look, which does not scale and is more informative than either.
If you take a proxy, write down what makes it a proxy, in the plan, next to the number. "Sign-ups stand in for adoption, and will overstate it if the trial is easy to start" is a sentence that saves an argument nine months later, when the proxy has moved and the real thing has not. Proxies drift from what they stand for over time, always in the flattering direction, because people learn what is counted.
Where the volume is too low for any number to mean anything, stop looking for a measure and go and examine the individual cases. Ten cases read properly beat a percentage calculated from twelve.
How many, and who owns each
Small numbers of measures are not a stylistic preference. A page with fifteen numbers has no priority on it, so the review discusses whichever moved most, which is usually the noisiest rather than the most important.
Two or three per initiative, each with a named person who is expected to explain it. Not to defend it: to explain it, which means saying what caused the movement and what they intend to do. A measure with no owner produces a discussion in which everyone offers a theory and nobody is accountable for the next step.
And decide in advance what number would trigger an action, and which action. "We will keep an eye on it" is what everyone says, and it means the number will be read as commentary for as long as the initiative lasts. The strategic planning pillar sets out how to write that trigger before anyone is attached to the outcome, which is the only time it can be written honestly.
Retiring a measure
Measures accumulate. Almost nothing is ever removed, so dashboards grow until nobody reads them, at which point the organization is measuring a great deal and noticing nothing.
Three good reasons to retire one. The decision it informed no longer gets made. It has been flat for long enough that it carries no information, which usually means it is measuring something stable rather than something managed. Or it has been thoroughly gamed and the behavior is now more damaging than the ignorance would be.
One bad reason: it is embarrassing. A measure that is telling you something you do not like is doing its job, and removing it is the most common way a plan stops being falsifiable.
Put a review date on each measure when you adopt it, the same way you would with the rest of the plan. The practice of rewriting plan lines so they can be acted on, covered in strategic planning examples, applies to the measurement lines too: a number with no stated trigger and no owner is another sentence nobody can act on.
Where a plan measure ends up feeding an individual's evaluation, treat that as a separate decision with separate consequences. Numbers designed to steer an organization behave differently once someone's standing depends on them, as performance management sets out, and anything with formal consequences for a person should go through your HR function rather than through a dashboard.
Common questions
Should plan measures be the same as operational measures?
Rarely. Operational measures answer "is the work running normally," and they are chosen for stability. Plan measures answer "is this change working," and they should be chosen for sensitivity, which usually makes them noisier and shorter-lived. Using one set for both purposes gives you numbers too slow to steer the plan and too jumpy to run the work.
What if the honest answer is that we cannot measure this?
Say it in the plan. Then name what you will look at instead, whether that is a small number of cases examined in detail or a scheduled conversation with the people closest to the work. An acknowledged gap is manageable. A number invented to fill the gap will be quoted for years.
How do we stop the review turning into a defense of the numbers?
Ask for the explanation before the judgment, and separate the two meetings if you can. When the same session decides both what the number means and what it implies for the person presenting it, you will reliably get a well-argued explanation and no useful information.
