Strategic planning: subtraction, the binding constraint and falsifiers written down early

Most plans are not wrong. They are unfalsifiable, which is worse.

A document full of directional statements about growing, improving, and strengthening cannot fail, because nothing in it could ever be shown not to have happened. Twelve months later everyone agrees it was broadly achieved, nobody can point to a decision it changed, and the next one gets written in the same shape.

A plan that is worth the days it costs does three things a wish list cannot: it says what you are not doing, it names the assumptions it rests on, and it states in advance what would prove it wrong.

What to take away

What a plan actually contains

Separate three things that normally arrive as one document, because they have different lifespans and different owners.

The bet. What you believe about the world that makes this the right thing to do. Which customers, which problem, why now, why you rather than anyone else. This is the part that can be wrong in an interesting way.

The constraint. What you are refusing. Which markets you will not enter, which requests you will decline, which work you will stop, which opportunities you will let pass. This is the part that makes the bet real.

The schedule of contact with reality. When you will look, at what, and what result would cause you to change course. This is the part that is almost always missing.

A document with the first and none of the others is an aspiration. Teams can tell the difference immediately, which is why plans that are all bet and no constraint produce visible cynicism.

Subtraction is the whole test

The reliable way to find out whether a plan has decided anything is to look for what got cut.

If every existing activity survives, and new priorities are added on top, no choice has been made. The organization has the same capacity it had last year, so the new work will be done with whatever is left over, which is nothing. In practice the plan will be quietly resolved by individuals deciding what to drop, one at a time, without coordination.

So make the subtraction explicit and specific:

That last one is the hardest and the most useful. Any real strategy trades something. If you cannot name the number you are prepared to see fall, you have not chosen a direction, you have listed everything good.

Start from the binding constraint

Planning usually starts from goals, which produces a list of desirable outcomes with no theory about why they are not already happening.

Start instead from the thing that is actually limiting you. Every organization has one at a time, and most of the effort spent elsewhere is wasted. The general argument, that a system's output is set in one place rather than everywhere, is the theory of constraints.

Ask what happens if demand doubles tomorrow. The answer names the constraint. If you could not deliver, the constraint is capacity or supply, and more marketing makes things worse. If you could deliver easily and nobody is asking, the constraint is demand or positioning, and hiring more delivery people is expensive idling. If you could sell and deliver but cannot afford the gap, the constraint is cash, and both of the others are premature.

The constraint moves once you relieve it, which is the part people forget. A plan built for last year's constraint is the most common form of stale strategy: still optimizing distribution when the problem became retention two quarters ago.

Name the constraint on the first page. If people disagree about what it is, you have found the real planning discussion, and it is worth more than the rest of the offsite.

Capacity is the plan

The gap between a plan and what happens is usually arithmetic that nobody did. The tendency to underestimate what a piece of work will take, even by people who have done similar work before, is consistent enough to have a name: the planning fallacy.

Before the plan is agreed, count the people who will do it, in the periods they are available, net of the work they already have. Not headcount: the actual hours of the specific people who can do the specific work, after support, maintenance, holidays, and the standing commitments that never appear in a plan because they are not projects.

Two failures show up immediately when you do this. The first is that most of the capacity is already committed to keeping existing things running, and the plan assumed a team with an empty diary. The second is that several initiatives depend on the same two people, so they cannot run in parallel however the timeline is drawn.

Neither is a reason not to plan. Both are reasons to sequence, and finding them in the planning room is much cheaper than finding them in month four.

Assumptions get names and owners

Every plan rests on things being true that might not be. Left implicit, they get treated as facts. Written down, they become the most productive part of the document.

Write each one as a claim that could be false, not as a background condition. Then attach three fields: who owns finding out, when they will report, and what happens if it turns out otherwise.

The value is not in the table. It is that assumptions with owners get tested, and assumptions without owners get defended. An assumption nobody owns will still be in the plan at the end of the year, unexamined, having been treated as settled since the day it was typed.

Write the falsifiers in advance

Decide, while nobody is invested, what result would mean this is not working.

This is the single highest-value habit in planning, and it is uncommon because it feels pessimistic in a room full of people who have just agreed on something. Do it anyway, and write it into the document: by this date, if this measure is not at least this, we stop, shrink, or change approach.

Two things make a falsifier real. It has to be measurable with data you will actually have, not data you would need to build a system to collect. And it has to be attached to an action, because "we will review it" is what everybody says and nobody does.

The reason to write them early is that the same evidence, encountered later, gets reinterpreted. Six months in, a disappointing number is explained by the launch timing, the market, the sample, the seasonality. Every explanation is plausible. Collectively they mean nothing is ever disconfirming, and the initiative continues by default until the money runs out.

Continuing by default is the most expensive decision most organizations make, and it is never made explicitly.

Sequence beats ambition

A plan with five parallel initiatives, in an organization with capacity for two, will deliver zero.

The reason is not effort. It is that partially-done work delivers nothing while consuming everything: attention, coordination, meetings, and the compounding cost of context switching. Two things finished are worth more than five things eighty percent complete, and eighty percent complete is where parallel initiatives go to sit.

Sequence by dependency first: what has to be true before the next thing can start. Then by information value: do the thing that resolves the biggest uncertainty earliest, because everything downstream of it is being planned in the dark until it lands. Then by reversibility: cheap and reversible moves before expensive and irreversible ones, since the cheap ones buy you the information to decide the expensive ones.

Say explicitly what is not being started yet. An initiative that is scheduled for later is not dropped, and saying so is what stops people from starting it quietly.

The annual cycle is the wrong cadence for most decisions

Planning once a year produces a document that is accurate for about six weeks and then becomes a thing people work around.

Different decisions have different natural frequencies, and pinning them all to the annual cycle either makes you slow or makes you ignore the plan.

Set those cadences deliberately rather than defaulting to whatever the budget calendar demands. The budget calendar exists for a financial reason and is not evidence about how fast your decisions need to move.

How plans die quietly

Watch for these, because none of them announce themselves:

Any one of these is worth stopping for. They are all cheap to fix in the first quarter and expensive in the fourth.

What to hold on to

The parts of a plan that do work are the unglamorous ones: the list of what you are not doing, the assumptions written as claims with owners, the falsifiers agreed before anyone was attached to the outcome, and an honest count of who is actually available. The vision paragraph is the part everyone remembers writing and the part that changes nothing.

Common questions

How long should a plan be?

Short enough that the people executing it have read all of it. If it needs a summary, the summary is the plan and the rest is supporting material. Length is usually a symptom of not having chosen.

What if the environment is too uncertain to plan?

Then plan the decisions rather than the outcomes. Name what you will do first, what you will learn from it, and what each result would cause you to do next. Uncertainty is an argument for shorter loops and explicit falsifiers, not for having no plan, and "we will stay flexible" is how organizations do five things badly.

Who should be in the room?

The people who will have to say no to something because of it. A plan agreed by people who face none of its constraints will not survive contact with the people who do. That connection between planning and daily judgment is part of the wider job described in management foundations.

How do we stop the plan being ignored the week after it is published?

Attach it to the things that already have force: what gets staffed, what gets funded, what gets declined, and what gets reviewed. A plan that changes none of those is a document, and people are correct to ignore it.