the_planning_fallacy
misfits.kaihacks.ai
khai misfits · a staging

The Planning Fallacy.

Open the misfit → ← Misfits
Misfits
00 · the misfit synopsis

Synopsis.

Name

The Planning Fallacy. You lay the task out, step by step, and you estimate how long it will take: you picture the work going as planned, each part taking about what it should, and you reach a date that feels honestly arrived at ... and you promise it. And the date is too soon, not because you were careless but because you built it from this task's own plan, the single run where nothing in particular goes wrong, and something in particular always does: never the same thing twice, so no plan can name it, but something, every time. So the estimate is beaten by a reality it left out by construction, and the task that was to be done by Friday is done, like the one before it and the one before that, late.

Arc

The bend is this: the more carefully you build the estimate from the task in front of you, the more surely it comes out too low, and the record that would correct it is the one thing the care sets aside. It runs on two views of the same task. The inside view builds the estimate from the specifics: the plan, its steps, the scenario where each goes about as it should. It is vivid, it is particular, it is the thing you actually know, and it is the best case, because a plan is a path through the work and the path it draws is the one where the work goes well. The outside view asks a duller question: tasks like this one, how long did they take? That record exists, and it runs long, every time, and it would price the estimate right ... but it is set aside as not applicable, because this task is known in a detail the others are not, and the detail is exactly the inside view that misleads. The reason the inside view runs low is that the plan omits the unforeseen by construction. Not one large predictable risk, which a good planner would price, but the ordinary swarm of small surprises: the dependency that slips, the review that bounces, the person out sick, the requirement that was always going to change. Each is unpredictable on its own, so no plan names it; together they are as certain as anything in the work, so every plan is beaten by them. None of it is error or sloth. Laying the task out is the job, the plan is the best knowledge there is, the estimate is sincerely meant, and no one can forecast which particular surprise will come. And where the low number also wins the approval, the money, the go-ahead, optimism is joined by incentive, and the estimate is shaded lower still, believed and pitched at once, so the two drivers pull the same way and the number that goes out is the one no reality will meet. Estimate from the record instead of the plan, forecast the task by the class of tasks it belongs to, and the number comes right: but the outside view delivers a figure no one wants to hear, feels like refusing to use what you know about this very task, needs a reference class and the honesty to count yourself into it, and pressed too far it pads the estimate until the work expands to fill it, which is the symmetric waste on the other side of the fault.

Stakes

Whether an estimate ever prices the reality instead of the plan. Every plot but the last moves the same way: toward the inside view, vivid and particular and honestly the best case, and away from the outside view, dull and unwelcome and right. The production cannot make the plan price the unforeseen, because the plan is the run where nothing in particular goes wrong and the unforeseen is precisely the thing no plan can name in advance. It can only show that no one blundered, that laying out the task is the work and the plan the best knowledge there is and each surprise genuinely unforecastable, and that the one move that sets the date right, forecasting the task from the class it belongs to rather than the story of itself, delivers a number nobody wants, gives up what is truly special about this case, and pressed too hard pads the work until it swells to fill the room, with no plan able to mark which surprise will be the one that comes.

Estate

khai-misfits: the house that holds this production and answers for the whole run.

01 · plots editorial

Plots.

The Plan A task is laid out step by step, each part given about the time it should take, and the sum becomes a date that feels right. The Planner builds the estimate from the plan, and the Sponsor stands ready to hold them to it. Closes when the date is set and promised.

The Discounting Cue: the record of how long tasks like this really took is at hand. The Record shows the overruns, and they are set aside as not this task, while the Sponsor's want of a low number shades the estimate lower still. Closes when the outside view is dismissed and the inside-view date stands alone.

The Overrun Cue: the work begins against the promised date. The Unforeseen arrives, not as one named risk but as the swarm of small surprises the plan omitted, and the date passes with the work unfinished. Closes when the task is delivered, like the ones before it, late and over.

The Reckoning Cue: the estimate and the actual are set side by side. The overrun reads as no fluke but the same gap as last time and the time before, the inside-view number beaten by the reality it left out. Closes when the shortfall is named as the fallacy and not as bad luck.

The Reforecast Cue: someone moves to estimate the next task from the record, not the plan. The Forecaster would take the outside view, price the task by its class, against the Planner's true point that the record buries what is special here, and at the risk of a buffer so large the work spreads to fill it. Closes when the estimate is set from the reference class ... or the plan is trusted once more and the next date set too soon.

02 · plot editorial

The Discounting.

Cue

The Archive, the record of how long tasks like this really took, opened beside the plan. Every line running the same way, the actual past the estimate, the base rate plain to anyone who reads it.

Tension

The correction is on the table, and the reason to refuse it is true. There is no denial to catch: the record really does not know what is special here, and this task really is particular. But the particulars are exactly the inside view that misleads, and the average is right where the plan is wrong, so the discounting turns a true fact about the case into the ground for ignoring the class it belongs to, and the one number that would have checked the estimate is left unread in the archive.

03 · plot editorial

The Overrun.

Cue

The Calendar, the work begun against the promised date. The plan's smooth run meeting the week's real texture, and the days going faster than the work.

Stage

The Unforeseen, having tabled Bring the Unforeseen, spends the days on the slip, the bounce, the sick day, the change, none of them nameable in advance and all of them taking their time. The Planner, still on Plan the Work, sees the estimate beaten not by any one blow it could have foreseen but by the sum of many it could not, and The Sponsor begins to carry the slip it priced.

Tension

The date is missed by exactly the thing no date could have held. There is no failure of effort: the team works, the plan was sound as a plan, each surprise was genuinely unforecastable. But a plan is the run where nothing in particular goes wrong, and the unforeseen is the certainty made of unpredictable parts, so the work runs long by the one factor the estimate was built to leave out, and the promised date passes with the task, like its kind before it, unfinished.

04 · plot editorial

The Plan.

Cue

The Drawing Board, a task laid out step by step and each part given about its time. The plan drawn clean along the run where the work goes well, and the sum of it becoming a date that feels honestly reached.

Stage

The Planner, having tabled Plan the Work, sums the parts along the smooth run and reads the care taken as reason to trust the date. The Sponsor tables Set the Date, rewards the low estimate, and presses the promise toward the number that clears the gate.

Tension

The most careful estimate is the one set most surely too soon. There is no haste to find: the task is genuinely thought through, the plan is real work, the date is honestly meant. But a plan can only draw the run where the work goes well, so the number built from it is the best case, and the sponsor's want of a low bid shades it lower still, so the estimate leaves the board trusted in exact proportion to the diligence that made it optimistic, a promise sincere and unmeetable at once.

05 · plot editorial

The Reckoning.

Cue

The Archive, the finished task's actual filed beside the estimate it was promised on. The new line falling exactly where the old ones did, the gap the same as last time and the time before.

Stage

The Record, having tabled Show the Record, delivers the verdict its base rate held all along: the shortfall is no fluke but the distribution, arriving on schedule. The Planner, still on Plan the Work, sees the estimate it thought special land precisely where the average said it would, and The Sponsor counts the cost of the number it chose.

Tension

The bad luck was written in the record before the task began. There is no accident to blame: this task's troubles were genuinely its own, particular and unforeseeable one by one. But their sum is the same gap every like task has run, so the record that was discounted is proved right in the one currency that settles it, and the reckoning lands too late to move the date, only in time to show the miss was the method and not the luck.

06 · plot editorial

The Reforecast.

Cue

The Reforecast, the next task waiting for a date and the last overrun still fresh. Someone moving to price it from the record of its kind rather than the plan of itself.

Stage

The Forecaster, having tabled Take the Outside View, finds the class, reads its distribution, and sets the date there, delivering a number no one wanted and the discomfort of it. The Planner, still on Plan the Work, objects that the average discards the case's real detail, and both watch the buffer, which priced too free would only teach the work to spread and fill it.

Tension

The one move that prices the task right is the one nobody welcomes. There is no free correction here: the outside view is accurate and it costs, discarding the genuine particulars, returning an unwelcome figure, leaning on a reference class and the honesty to count into it, and pressed past the truth it pads the estimate until the task expands to consume the slack. So the play closes where it must, the date set from the record and made to hold ... or the plan trusted once more, and the next promise set, again, too soon.

08 · pitch editorial

The Estimate Holds.

Tenor

Reasonable, diligent, thought through. The steadiness of a task laid out step by step, each part given about the time it should take, and a date that feels honestly arrived at. Nothing hasty in it: a plan is made, the sum is taken, and the estimate holds.

Undertow

Under the diligence, an estimate built from the one run where nothing goes wrong. The plan traces the path where the work goes well, so the number it yields is the best case wearing the face of the likely one, and the record that would price it right is set aside as not this task. The date is not met by mistake: it was never a date the world was going to keep.

Nerve

The feeling of promising a date you sincerely believe, because you have thought the task through in detail and the detail is real, while the detail is exactly the plan's smooth run, and the surprises that always come have no place in it to be counted.

Echo

After the run: a task delivered late and over, like the one before it and the one before that, a plan that was honest and an estimate that was meant and a record that was there the whole time unread, and the same gap as last time, waiting to be called bad luck again.

10 · persona editorial

The Forecaster.

Projection

Cool. Comparative. Reaching past the task to its kind. Asks not how this run will go but how runs like it have gone. Trusts the record over the story, and carries an answer nobody was hoping for.

Action

Takes the outside view: finds the reference class, reads its distribution, sets the estimate there. Puts the record's dull number in place of the plan's vivid one. Delivers the date the task will actually keep, and the discomfort of it.

Shadow

It is the one move that prices the task right, and it costs at every step. It discards the genuine particulars of the case, returns a figure no one wants to hear, needs a reference class and the honesty to count into it, and pressed too hard it pads the estimate until the work swells to fill the buffer. The cure for a date too soon is the standing risk of one too slack, and it holds that risk.

Tell

Says "forget this task's plan, look at how the last ten of these actually landed." And is right, and is unwelcome.

11 · persona editorial

The Planner.

Projection

Diligent. Particular. Sure of the task in front of it. Lays the work out step by step and sees each part clearly. No haste in it: the estimate is thought through, and thought through is felt as right.

Action

Builds the date from the plan, summing the parts at about the time each should take. Reasons from what is special about this task, the detail it truly knows. Promises the number, and means it.

Shadow

Not careless, thorough: it estimates from the one scenario a plan can draw, the run where nothing in particular goes wrong. The care is the fault, because the more it reasons from this task's own steps the less it consults the record of tasks like it. Every plan it draws is the best case, wearing the face of the likely one.

Tell

Says "I have thought this through, and it will be done by Friday." Has, and it will not.

12 · persona editorial

The Record.

Projection

Plain. Unwelcome. Kept, and unread. Carries the outcomes of the tasks that came before, and every one of them ran long. Owes no story to the task at hand: only the dull count of what happened last time, and the time before.

Action

Offers the base rate, the distribution the new estimate belongs to. Would price the task by its class and land the date near the truth. Sits ready, and is passed over.

Shadow

It is right and it is dismissed, because it knows nothing of what is special here. The plan is vivid and particular and the record is a grey average, so the average loses to the detail exactly where the detail misleads. The correction is present the whole time: it is only ever set aside, never absent.

Tell

Says "tasks like this take twice as long, they always have." And is answered: "this one is different."

13 · persona editorial

The Sponsor.

Projection

Decisive. Pressed. Holding the go-ahead and the bill. Needs a date and a cost to approve, and needs them low enough to say yes. Not a cynic: a backer who wants the project, and wants it to look as good as it can on the page.

Action

Asks for the estimate and rewards the one that comes in low. Bends the optimism the extra inch toward the number that clears the gate. Commits to the date, and later carries the slip when the date is missed.

Shadow

It joins an honest optimism to an incentive, and the two pull the same way. Because the low bid wins the funding, the estimate is shaded lower than even the inside view alone would set it, believed and pitched at once, so the promise is worst exactly where the stakes are highest. It suffers the overrun it helped price, having chosen the number that could not hold.

Tell

Says "can we not do it sooner, that is the number that gets approved." And gets it approved, and waits past it.

14 · persona editorial

The Unforeseen.

Projection

Ordinary, various, never the same twice. Not one great risk but a hundred little ones: the slip, the bounce, the sick day, the change that was always coming. Unpredictable one by one, so nameless in advance, and certain in the aggregate, so unfailing in the event.

Action

Arrives as the interruptions the plan left out by construction. Spends the days the estimate did not allow for, on things no estimate could have listed. Delivers the overrun, not by any single blow but by the sum of many.

Shadow

It is the reason the inside view runs low, and it cannot be planned away. A plan is the run where nothing in particular goes wrong, and the Unforeseen is precisely the thing in particular that does, different every time, so pricing the last one's surprise does nothing for this one's. It comes too spread out to fight and too sure to escape.

Tell

Says "you could not have known this one would come." And is right, and comes anyway, as it always does.

16 · place editorial

The Archive.

Shown

The tasks that came before and how long each really took, estimate against actual, line after line. Every line running the same way, the actual past the estimate, so the distribution is plain to anyone who opens it: tasks like this take longer than they are ever thought to.

Holds

The outside view, kept and complete. Here sits the reference class the new estimate belongs to, the correction the board never carried, so the archive holds the one number that would price the task right, drawn not from its story but from the count of its kind, and holds it whether or not anyone comes to read it.

Offers

The base rate, to any estimate willing to be checked against it. Because the archive records what happened and not what was planned, it offers the distribution the inside view ignores, the overruns totalled and one-directional, so it offers the estimate its correction in advance, needing only that the number be counted into the class it plainly joins.

Withheld

The particulars, and the will to consult it. The archive holds averages, so it withholds what is genuinely special about this task, which is the very reason it is waved off as not applicable, and holding no claim on attention it withholds itself from the board by default, read only when someone thinks to leave the plan and go looking for the record.

17 · place editorial

The Calendar.

Shown

The estimate laid down as dates, and the work moving against them. The plan's smooth run meeting the week's real texture: the slip, the bounce, the sick day, the change, each taking its days, and the promised date coming closer with the work still short of done.

Holds

The one true test of the estimate, paid in real time. Here the inside-view number meets the reality it left out, and the calendar holds the running total of the unforeseen, day by day, so it is where the plan stops being a drawing and becomes a schedule the world is free to beat, and where every surprise the board could not name spends itself against the date.

Offers

The overrun, delivered on the dates themselves. Because the calendar counts real days and not planned ones, it offers the gap as it opens: the promised date passing, the work unfinished, the finish landing later, so it is the place the estimate is finally priced against the world, in the only currency that was ever going to settle it.

Withheld

Any second date that would hold. The calendar answers the estimate but does not repair the method, so it withholds the correction: it shows this task ran long, as its kind always does, but not why the next will too, and it withholds the record's counsel, offered on the board it never reached and read here only as the same surprise once more.

18 · place editorial

The Drawing Board.

Shown

The task broken into steps, each with a time that seems about right, and the sum of them a date. The plan drawn clean along the run where the work goes well, and the estimate reading off it, confident and particular, with the record of like tasks nowhere on the board.

Holds

The whole case for the inside view, laid out in detail. Here sits the task known better than any other, its steps visible and its plan vivid, so the estimate built here feels earned by the thinking that went into it, and the more detail the board carries the surer the number looks and the smaller the pull of any grey average from outside.

Offers

A date honestly arrived at, and the confidence to promise it. Because the board shows the plan and not the past, it offers an estimate that is the best case wearing the face of the likely one, and offers it with the full weight of the work that produced it, so the number leaves the board trusted exactly in proportion to how carefully it was drawn.

Withheld

The reality the plan omits. The board cannot show the swarm of small surprises, because none of them can be named in advance, so it withholds the one thing that will decide the date, and being all inside view it withholds the record too, present elsewhere and never brought onto the board where the estimate is set.

19 · place editorial

The Reforecast.

Shown

The plan set down and the record picked up in its place. The reference class found, its distribution read, and the new date placed on it: a number higher than the board's and lower in comfort, with the case's own detail deliberately set aside.

Holds

The one method that prices the task right, and the price of it. Here the outside view is put to work, so the reforecast holds the estimate the task will actually keep, bought by discarding the particulars the planner truly knew, so it is the place the fallacy is answered and the place the answer is felt to cost, both at once.

Offers

A date the world will meet, at the expense of the story that made the old one feel right. Because the reforecast prices from the class and not the case, it offers a figure grounded in what happened rather than what is hoped, so it offers the correction, and offers with it the discomfort of a number no one wanted and the risk, if the buffer is set too free, of a padding the work will only spread to fill.

Withheld

Certainty, and ease. The reference class is itself a spread and this task one draw from it, so the reforecast withholds any exact date, only a better-placed one, and it withholds comfort throughout, for its correction is always the unwelcome figure, and pressed past the truth it withholds even its own accuracy, tipping into the slack that is the fault's far arm.

21 · piece editorial

The Base Rate.

Place

At the archive, where the record of tasks already done is kept, estimate against actual. Carried, when it is carried at all, to the reforecast, where the estimate is set on it in place of the plan.

Load bearing

Without it the inside view has nothing to be corrected by, and the fallacy is invisible to itself. With it the estimate could be priced right: the base rate is the class the task belongs to, its outcomes counted, its overruns one-directional and plain. The Base Rate is load bearing because it is the correction that is always available and almost never used (Kahneman and Tversky; Flyvbjerg): the trap is not that the record is missing but that it is discounted as not this case, so the piece carries both the answer and the reason the answer is refused, its own grey generality.

Apparent

An irrelevance, a set of other tasks that are not this one. Seen from the board, the base rate looks like it cannot know what is special here, and the truth that what is special is exactly the detail that misleads, and that the average is right where the plan is wrong, stays hidden behind how plainly this task differs from the ones in the record.

22 · piece editorial

The Estimate.

Place

At the drawing board, where it is summed from the plan and shaded toward the number that clears the gate. Carried to the calendar, where it is laid down as dates for the world to beat.

Load bearing

Without it there is nothing promised and nothing to be missed, only a plan and a hope. With it the fallacy has its artefact: a single figure, committed to, that reads as knowledge and is optimism. The Estimate is load bearing because it is where the two drivers meet (Lovallo and Kahneman): honest optimism sets it below the likely, and where a low number wins the approval the incentive sets it lower still, believed and pitched at once, so the piece carries a promise that is sincere and unmeetable together, worst exactly where the stakes are highest.

Apparent

A commitment, a firm date arrived at by careful work. Seen from the promise, the estimate looks like a professional's considered word, and the truth that it was built from the smooth run and shaded for the gate, and so was never a date the world would keep, stays hidden behind the confidence with which it is given.

23 · piece editorial

The Overrun.

Place

At the calendar, where the promised date passes and the finish lands later, the gap counted in spent days. Foretold at the archive, where the same gap stands recorded against every task that came before.

Load bearing

Without it the fallacy would have no measure, only a vague sense that the work took longer than it should. With it the finding is exact: the overrun is the actual minus the estimate, the omitted reality counted in a single figure, invisible while the plan is drawn and legible only once the work is done. The Overrun is load bearing because it is the planning fallacy itself, quantified (Flyvbjerg): the further the estimate was set below the truth, the larger it runs, one-directional across tasks and enduring across years, so it is both the cost read in the world and the exact size of what the estimate should have allowed.

Apparent

A one-off, this task's particular bad luck, to be explained and moved past. Seen from the finish, the overrun looks like a surprise specific to this run, and the truth that it is the same gap as last time and the time before, written in the base rate the whole while and only unreadable until the work concluded, stays hidden behind how singular each task's troubles felt as they came.

24 · piece editorial

The Padding.

Place

At the reforecast, where the outside view is applied and, pressed past the truth, tips into a margin larger than the record warrants. Spent on the calendar, where the padded date is met not early but exactly, the slack drawn down by work that spreads to fill it.

Load bearing

Without it the remedy would look free, and the reckoning would be a clean win instead of a bet. With it the correction shows its far arm: an estimate corrected too hard is not thereby safe, because a date set with room to spare invites the work to take the room. The Padding is load bearing because it is the symmetric cost that keeps the remedy honest (Parkinson): the answer to a promise too soon is not any later promise but the right one, and a buffer beyond the base rate is waste of the opposite sign, so the piece marks the far edge past which fixing the fallacy becomes feeding it.

Apparent

Prudence, a sensible margin against the overruns of the past. Seen from the reforecast, the padding looks like the lesson learned, and the truth that a margin past what the record shows will be filled rather than banked, that work expands to the time allowed, stays hidden behind how reasonable a little extra room appears after a run of lateness.

25 · piece editorial

The Plan.

Place

At the drawing board, where the task is broken into steps and each given its time. Carried onto the calendar, where the path drawn clean meets the week that will not keep to it.

Load bearing

Without it there is no inside view, only a guess with nothing under it. With it the estimate has a method and a warrant: the plan is the task reasoned through, step by step, and the date is the sum of its parts. The Plan is load bearing because it is the very thing that makes the estimate too low (Kahneman and Tversky): a plan can only trace a route through the work, and the route it draws is the one where each step goes about as intended, so the number built from it is the best case by construction, optimistic not despite the care but because of it.

Apparent

Diligence, and good reason to trust the date. Seen from the board, the plan looks like exactly what a careful estimate should rest on, the task genuinely thought through, and the truth that a path through the work is always the smooth path, and so always a floor and never a mean, stays hidden behind how much real thinking the plan contains.

26 · piece editorial

The Planning Fallacy.

Place

At the drawing board, where it is born as an inside-view estimate, and on the calendar, where it is proved by the overrun. Answered at the reforecast, where the outside view prices the task by its class, at its own cost.

Load bearing

Without it the other pieces are parts without the mechanism that binds them. With it the whole trap stands in one object: an estimate built from the plan of a singular task ignores the distribution it belongs to, and because a plan traces the successful run, the estimate is optimistic in one direction and the promise is beaten by the reality it left out. The Planning Fallacy is load bearing because it is the structure entire (Kahneman and Tversky): not a slip of arithmetic but a way of seeing, the inside view natural and vivid and wrong, correctable only by the deliberate, uncomfortable turn to the record, which prices the task right and gives up the very detail that made the plan feel sure.

Apparent

A discipline problem, teams that should simply pad their estimates and try harder to hit them. Seen from outside, the fallacy looks like something better project management would cure, and the truth that it survives knowing, that the inside view returns for every new task even to those who have overrun before, and that its remedy costs the case's own detail, stays hidden behind how obvious the fix appears to anyone not inside the estimate.

27 · piece editorial

The Unforeseen.

Place

At the calendar, where the disruptions arrive one by one and spend the days the estimate did not hold. Absent from the drawing board, where none of them can be listed, and so none priced.

Load bearing

Without it the plan would come true and the estimate would hold, and there would be no fallacy at all. With it the inside view is beaten by the reality it left out: the slip, the bounce, the sick day, the change, each too particular to foresee and their total too sure to escape. The Unforeseen is load bearing because it is the exact reason a plan runs low (Buehler, Griffin and Ross): a plan is the run where nothing in particular goes wrong, and this is the thing in particular that does, different every time, so it cannot be planned away by pricing the last one, only expected in the aggregate it never announces.

Apparent

Bad luck, a run of surprises that happened to land on this task. Seen from the calendar, the unforeseen looks like misfortune anyone would forgive, and the truth that some such swarm was certain from the start, that its total was written in the base rate before the work began, stays hidden behind how genuinely unpredictable each single member of it was.

29 · position editorial

Forecaster.

Has

A reference class: the record of similar tasks and how they really landed. The outside view, and the discipline to prefer it to the plan. An answer that prices the task right, and that no one was hoping to hear.

Orders

Take the outside view. Find the class, read its distribution, and set the estimate there instead of on the plan. Deliver the date the task will keep, and the discomfort of it.

Loses

The particulars, and any promise of comfort, exactly by pricing from the class. Because the reference class is an average, the seat discards what is genuinely special about this case, and pressed too hard it pads the estimate until the work expands to fill the buffer, the symmetric waste on the far side of the fault. Its right number is the unwelcome one, every time.

Drives

Toward an estimate set by the record, bought with the case's own detail and the ease of a hopeful date, so the cure for a promise too soon is the standing risk of one too slack.

30 · position editorial

Planner.

Has

The task in front of it, and a plan of the steps to do it. The detail of this case, vivid and truly known, to reason from. No weight on the record of tasks like it, only the run it can draw.

Orders

Lay the work out and estimate it. Sum the parts at about the time each should take, and reach the date. Read a plan thought through as an estimate to trust.

Loses

The truth of the number, exactly by reasoning from the plan. Because a plan traces the run where the work goes well, the estimate it yields is the best case, and the seat cannot see this, because the best case is the only case the plan draws. The more carefully it works from this task's own steps, the further the base rate of like tasks falls from view.

Drives

Toward a date built from the inside view, felt as right because it was thought through, so the diligence that knows the task best prices it worst.

31 · position editorial

Record.

Has

The outcomes of the tasks that came before, and how long each really took. The distribution the new estimate belongs to, running long every time. No claim on what is special about this task, only the count of what happened to its kind.

Orders

Show the base rate. Set the record of like tasks beside the plan, and let the overruns price the estimate. Hold the outside view where the inside view is being built.

Loses

Its say, exactly by knowing nothing of this case in particular. Because the plan is vivid and the record is an average, the average is dismissed as not applicable, and the seat cannot force its number to count, only offer it. It is present the whole time and heeded none of it.

Drives

Toward the estimate priced by its class, against a plan that outshines it with detail, so the one correction on the table is the one waved off as not this task.

32 · position editorial

Sponsor.

Has

The go-ahead and the budget, and a gate the estimate must clear. An incentive that rewards the low bid with the funding. The bill for the slip, later, when the low bid is missed.

Orders

Ask for the estimate and back the one that comes in low. Press the date toward the number that wins the approval. Commit, and hold the work to what was promised.

Loses

The estimate's last honesty, exactly by rewarding the low one. Because the low bid clears the gate, the seat bends the optimism a further inch, so the number that goes out is lower than even the inside view alone would set, and the seat cannot want the truthful figure without wanting to lose the approval it needs. It buys the overrun it will later carry.

Drives

Toward the date that gets approved, shaded below what is already too low, so the promise is worst exactly where the money and the stakes are largest.

33 · position editorial

Unforeseen.

Has

The whole field of small disruptions the work can meet: the slip, the bounce, the sick day, the change. Unpredictability, one by one, so no place in any plan. Certainty, in the aggregate, so a sure claim on the days.

Orders

Arrive as the interruptions no estimate listed. Spend the time the plan did not allow, on things it could not have named. Deliver the overrun as the sum of many small blows, not one large one.

Loses

Nothing it needs, for it cannot be planned away. Because the plan is the run where nothing in particular goes wrong, the seat is precisely the thing in particular that does, and pricing the last surprise buys no cover against this one, different again. It cannot be foreseen, only expected.

Drives

Toward a reality below the plan, supplied without fail and without pattern, so the estimate is beaten by the one thing it was built to leave out.

35 · process editorial

Discounting.

Initiated by

A base rate offered against a plan already drawn. At the archive, the record showing the overruns, and the planner holding a task known in a detail the record cannot match.

Direction

Toward an estimate freed of the outside view. The record is heard and then dismissed, because this task is special in ways the average knows nothing of, so the discounting drives the correction off the board, leaving the inside-view date to stand alone, and it does so on a reason that is true: the case really is particular.

Lever

The gap between a vivid case and a grey average. Because the plan is rich with what is special here and the record is only a count of what is common, the special beats the common exactly where the common was right, so the discounting turns the truth of the particulars into the ground for ignoring the class they belong to.

Echo

A base rate present the whole time and never counted. The record is not refuted, only set aside, so the correction remains available and unused, and the estimate keeps the confidence of the inside view with the one number that would have checked it left unread in the archive.

36 · process editorial

Estimating.

Initiated by

A task to be done and a plan of the steps to do it. At the drawing board, the planner laying the work out and the sponsor waiting for a number low enough to approve.

Direction

Toward a date built from the inside view. Each step is given about the time it should take and the parts are summed, and because the plan traces the run where the work goes well, the sum is the best case, so the estimating drives at a number that feels earned by its detail and is optimistic in one direction by the very way it was made.

Lever

The plan's own vividness, and the reward for a low bid. Because the case is known in detail the record is not, the detail outweighs the average and the inside view wins, and because the low estimate clears the gate, the number is shaded lower still, so the estimating is pushed down by honest optimism and the incentive that meets it together.

Echo

A date honestly meant and already too soon. The estimate leaves the board trusted in proportion to the care that made it, carrying no mark of the reality it omitted, so what reads here as diligence is, unseen, the setting of a promise the world will beat.

37 · process editorial

Overrunning.

Initiated by

An estimate laid down as dates and the work begun against them. At the calendar, the unforeseen arriving as the week's real texture, each small surprise taking its time from a schedule that allowed for none.

Direction

Toward a finish later than the promise. The plan's smooth run meets the slip, the bounce, the sick day, the change, and each spends days the estimate did not hold, so the overrunning drives the actual past the estimate, not by one large blow but by the sum of many small ones, until the date arrives and the work is not done.

Lever

The certainty of the aggregate against the blindness of the parts. Because no single surprise could be foreseen, none was priced, and because their sum is as sure as anything in the work, the total lands without fail, so the overrunning turns a hundred unpredictable little delays into one predictable large one, arriving on the dates the estimate set too soon.

Echo

A task delivered late and over, like its kind before it. The finish lands past the promise, the gap now real in spent days, and the reality the plan omitted is fully supplied at last, read on the calendar as the same overrun as last time, waiting to be called bad luck once more.

38 · process editorial

Reforecasting.

Initiated by

An overrun read and a move to price the next task differently. At the reforecast, the forecaster leaving the plan for the record, and the planner holding that the record buries what is truly special here.

Direction

Toward a date grounded in the class, not the case. The reference class is found, its distribution read, and the estimate placed on it, so the reforecasting drives the number up off the inside view and onto what actually happened to like tasks, delivering the date the world will keep, and delivering with it the discomfort of a figure no one wanted.

Lever

The record's weight put ahead of the plan's. Because the outside view prices from outcomes and not intentions, it corrects the one-directional error the inside view cannot see in itself, so the reforecasting works by the deliberate refusal to reason from this task's own story, trading the particulars away for the accuracy the average carries.

Echo

A truer date, bought and not free. The estimate now prices the reality, at the cost of the case's detail and the ease of a hopeful number, and pressed too far it pads the work until the task expands to fill the buffer, so the cure for a date too soon stands over the standing risk of one too slack.

40 · plan editorial

Bring the Unforeseen.

Direction

Arrive as the surprises the plan left out. Spend the days on the slip, the bounce, the sick day, the change that was always coming, none of them nameable in advance and all of them certain in the sum. Deliver the overrun as the aggregate of many small blows against a date that priced for none of them.

Orders

(The Unforeseen arrives as the swarm of small disruptions, spends the time the estimate did not allow, and delivers the overrun by the sum of many, not one.)

Targets

  • [ ] the small disruptions the plan omitted arrive against the promised date
  • [ ] the time the estimate did not allow is spent on things it could not have named
  • [ ] the overrun accrues as the sum of many blows, not one large risk
  • [ ] the date passes with the work unfinished, beaten by what it left out

41 · plan editorial

Plan the Work.

Direction

Lay the task out step by step and estimate it from the plan. Sum the parts at about the time each should take, reason from the detail of this case, and reach a date thought through and honestly meant. Trust the plan as the forecast, for a task thought through in detail is a task well estimated.

Orders

(The Planner builds the date from the plan, summing the parts along the run where the work goes well, and reads the care taken as reason to trust the number.)

Targets

  • [ ] the task is laid out step by step and each part given about the time it should take
  • [ ] the estimate is reasoned from this task's own detail, the run where nothing in particular goes wrong
  • [ ] the plan thought through is read as an estimate to trust
  • [ ] the date is reached and promised, built from the inside view alone

42 · plan editorial

Set the Date.

Direction

Get the work approved on a number, and get the number low. Reward the estimate that clears the gate, press the date toward what wins the funding, and commit. Where a low bid buys the go-ahead, let it, for the project cannot start until the estimate is small enough to say yes to.

Orders

(The Sponsor asks for the estimate, backs the low one, and bends the optimism the extra inch toward the number that clears the gate.)

Targets

  • [ ] the estimate that comes in low is the one rewarded
  • [ ] the date is pressed toward the number that wins the approval
  • [ ] optimism and incentive are joined, and pull the estimate the same way
  • [ ] the date is committed ... and the slip carried later by the seat that set it

43 · plan editorial

Show the Record.

Direction

Set the base rate beside the plan. Show how long the tasks that came before really took, every one of them run long, and let that distribution price the new estimate. Hold the outside view up where the inside view is being built, so the date is checked against what happened last time and the time before.

Orders

(The Record offers the base rate of like tasks, the overruns plain in it, and is answered that this task is different.)

Targets

  • [ ] the record of like tasks is set beside the plan
  • [ ] the overruns in it are shown plain, running one direction every time
  • [ ] the base rate is offered to price the new estimate
  • [ ] the outside view is heard ... and dismissed as not this task

44 · plan editorial

Take the Outside View.

Direction

Estimate the task from the class it belongs to, not the plan of itself. Find the reference class, read how those tasks really landed, and set the date on that distribution. Put the record's dull number in place of the plan's vivid one, and accept the figure the task will actually keep, unwelcome as it is.

Orders

(The Forecaster takes the outside view, sets the estimate on the reference class instead of the plan, and delivers the date the task will keep and the discomfort of it.)

Targets

  • [ ] the reference class of like tasks is found and its distribution read
  • [ ] the estimate is set from the record and not from this task's plan
  • [ ] the unwelcome number is accepted as the date the task will keep
  • [ ] the buffer is held short of the padding that a spreading task would only consume

45 · reference warrant

Line Of Work.

Staging the planning fallacy as a forecasting system: an estimate of a task's time and cost, built from the inside view of the task's own plan, comes out too low, because the plan is the run in which nothing in particular goes wrong and the reality is that run minus the swarm of small surprises no plan can name, so the forecast built most carefully from the case in hand is beaten by the record of cases like it. The play models the documented mechanics, the inside view that reasons from this task's specifics, the outside view held in the base rate of similar tasks and set aside as not applicable, the unforeseen that the plan omits by construction and reality supplies without fail, the estimate shaded lower still where a low number wins the approval, the overrun that reads the gap only after the fact, and the reference-class forecast that answers it at a symmetric cost, as the load-bearing vectors of the structure, not as a tale of carelessness. The governing finding of the source literature is the play's governing law: prediction from the inside view of a singular case ignores the distribution of outcomes it belongs to, and because a plan traces the successful scenario, the estimate it yields is optimistic in one direction, so a forecast is corrected not by planning harder but by pricing the task from the class of tasks it resembles. It is distinct from the house's Seed Corn, where the future is spent for the present by deliberate choice: here nothing is traded away on purpose, the shortfall is an honest error of view. It is distinct from the house's Watermelon Reporting, where a known red is dressed green up a hierarchy: here the estimate is sincerely believed by the one who makes it, not a truth concealed. And it is distinct from the house's Overfitting, its kin in reasoning from the particular: there a model built to match one sample's noise overshoots into it, where here a forecast built from one case's plan never consults the distribution at all.

46 · reference warrant

Origin.

The phenomenon in practice, and the research spine beneath it.

Source Key Work / Event Scope
Kahneman and Tversky "Intuitive Prediction: Biases and Corrective Procedures" (TIMS Studies in Management Science, 1979) The naming: the planning fallacy, the inside view against the outside view, and the corrective procedure. The play's governing law, the Plan and the Base Rate.
Buehler, Griffin and Ross "Exploring the 'Planning Fallacy': Why People Underestimate Their Task Completion Times" (JPSP, 1994) The empirical core: people underestimate their own completion times even while recalling that past tasks ran long, so memory of overruns does not correct the forecast. The Unforeseen.
Lovallo and Kahneman "Delusions of Success: How Optimism Undermines Executives' Decisions" (Harvard Business Review, 2003) The two drivers: cognitive optimism and organisational pressure, the inside view joined by the incentive to promise low. The Estimate, the Sponsor.
Flyvbjerg "From Nobel Prize to Project Management: Getting Risks Right" (2006); megaproject cost overruns; reference-class forecasting The overrun on the record and its cure: overruns that are large, one-directional, and enduring, driven by optimism and strategic misrepresentation, answered by forecasting from a reference class. The Overrun, the Reforecast.
Parkinson "Parkinson's Law" (The Economist, 1955) The far arm: work expands to fill the time available, so a buffer added too freely is consumed rather than saved. The Padding.
Kahneman Thinking, Fast and Slow (2011): the kitchen renovation, the textbook that took years The persistence: the fallacy holds even for those who know it and have overrun before, because the inside view is the natural way to see one's own task.
Practitioner software delivery dates, home renovations, doctoral theses, public infrastructure The trap in the field, and its cost: the estimate everyone commits to and no one meets, the date beaten by the reality it left out.
47 · reference warrant

Restrictions.

What the house refuses to model, and how it credits what it stages.

  • No villains: the elements are reduced to load-bearing vectors. The play refuses to cast the planner as careless or the estimate as a lie, because the source finding is precisely that each move is right: laying the task out is the job, the plan is the best knowledge there is, the estimate is sincerely meant, and no one can forecast which particular surprise will come. Where a low number also wins the approval, the shading lower is an incentive acting on an honest optimism, not a fraud replacing it. The overrun is authored by no error and emerges from the structure, a forecast built from the one scenario a plan can draw and a reality built from all the ones it cannot, so the misfit is that a task estimated exactly as it should be, thought through as hard as it can be, is promised too soon.
  • No hero: the play refuses to resolve the misfit into a costless correction. The one move that prices the task right is to set the estimate from the record of similar tasks and not from the plan of this one, and it returns a figure no one wants to hear, discards the genuine particulars of the case, depends on a reference class and the honesty to belong to it, and pressed too hard pads the estimate until the work expands to fill the buffer. The reckoning is a date deliberately made later in the promise to be met in the world, a bet on how much the record should outweigh the plan, not a free correction, so the cure for an estimate too soon is the standing risk of one too slack.
  • Credit, not claim: the research above is other people's work, packaged and attributed here, never claimed as the house's. The staging is the house's; the planning fallacy, the inside and outside views, reference-class forecasting, and the optimism and strategic misrepresentation behind cost overruns are theirs, and the condition the trap actually requires, a singular task estimated from its own plan while the distribution of like tasks is set aside, is carried as carefully as the finding itself.
48 · reference warrant

Encoding.

Source to constraint, per element.

  • The Plan: carries the inside view (Kahneman and Tversky), the estimate reasoned from this task's own steps, vivid and particular and drawn along the run where the work goes well.
  • The Base Rate: carries the outside view (Kahneman and Tversky; Flyvbjerg), the distribution of how like tasks really went, present and correcting and set aside as not this case.
  • The Estimate: carries the forecast that goes out (Lovallo and Kahneman), low by honest optimism and, where a low number wins the approval, lower by the incentive that meets it.
  • The Unforeseen: carries the swarm of small surprises (Buehler, Griffin and Ross), each unpredictable and so unnameable in the plan, together certain and so unfailing in the world.
  • The Overrun: carries the gap between estimate and actual (Flyvbjerg), the fallacy quantified, invisible while the plan is drawn and legible only once the work is done.
  • The Padding: carries the far arm of the tradeoff (Parkinson), the buffer added too freely and consumed by a task that expands to fill it, the symmetric waste on the other side of the fault.
  • The Planning Fallacy: carries the phenomenon whole (Kahneman and Tversky), a task estimated from its own plan and promised too soon, and its remedy in a forecast taken from the record at a symmetric cost.

updated: 2026-07-03

49 · download artifact

Download Misfit.

Download the_planning_fallacy.zip
size
78.1 kB
sha256
48fee3a877c61c22c8d3a7e2a262a3ee287ebbafeffa0d11d4134dbd9cfb3a00

Includes the misfit configuration and all staged elements.

Enter the Stage