structured-thinking · part 6 of 7
Precise Enough to Be Wrong
Reading time: about 9 minutes.
The pricing decision is on Thursday. On Tuesday, someone is still rebuilding the cost model, because the version from last week allocated shared support time across product lines using headcount, and it turns out three of those people only work on one line. The new allocation is more defensible. It moves the unit cost by about two percent.
The decision on Thursday is whether to raise prices by five percent or ten. Nobody in the room can tell the difference between those two options at the level of precision the model just gained. The two days went into a number that was already good enough on Friday, while the question of what competitors would do, which actually separates five from ten, got a conversation in a hallway.
What This Article Answers
- What does 80/20 mean applied to analysis, rather than to sales figures?
- How do I know in advance which 20% of the work matters? (You don't. You work backwards from the decision instead.)
- How precise does this number actually need to be?
- Isn't "good enough" just a nice way of saying sloppy?
- When is being rigorous the right call, and how do I tell before I have spent the time finding out?
- How do I hand over an estimate without it being mistaken for a measurement?
TL;DR
Precision is not a virtue you apply evenly. It is a resource you spend, and the right amount is set by the decision the number feeds, not by how uncomfortable the number makes you. Decide how precise the answer needs to be before you start working on it, write that down, and stop when you get there.
- The rigor budget is a decision, not a discovery. Set it up front, in writing, or the deadline sets it for you at the worst possible moment.
- The threshold that matters is the one that changes the answer. If the recommendation is the same at plus or minus 20%, more accuracy than that buys nothing.
- Reversibility sets the budget. Cheap to undo means decide fast. Hard to undo earns the extra week.
- 80/20 has no measured constant behind it. It is a useful posture, not a law, and the man who named it said so himself.
- The unexamined 80% is deferred, not worthless. Say which parts you did not check, out loud. That is what separates a fast answer from a careless one.
Precision on the vertical axis, time on the horizontal. The decision threshold is a flat line partway up. Everything above the line is real work that changes nothing, and it is usually where the last two days went.
The Budget You Set Before You Start
The last part argued for committing to an answer on day one and spending the rest of the time trying to kill it. This one is about the other half of that discipline: how hard you should try, and how you decide that in advance rather than discovering it on the last afternoon.
Here is the move. Before you start the analysis, write down how precise the answer needs to be, and why that level and not a finer one. The rigor budget. Not "as accurate as I can get it in the time," which is not a budget, it is a description of working until you run out. Something you could be held to: this needs to be right to the nearest ten percent, because the decision it feeds is a yes or no on a threshold that sits well outside that range.
The reason to write it down is that precision has no natural stopping point. There is always another allocation to refine, another sample to widen, another assumption to test. Left unbounded, analysis expands until the deadline stops it, and the stopping point ends up set by the calendar rather than by anything about the problem. Everybody has watched a model get more correct in its last two days and less useful, because the correctness landed on a part of it nobody was going to act on.
Herbert Simon named the underlying behaviour satisficing, a blend of satisfy and suffice: searching the options until you find one that clears a threshold you set beforehand, rather than continuing until you have found the best one. His point was descriptive. Under real constraints of time and information, this is what capable people actually do. The failure is not that people satisfice. It is that most of them do it accidentally, at the end, under pressure, instead of deliberately, at the start.
Work Backwards From What Would Change
The question "how precise does this need to be" sounds unanswerable until you attach it to a decision, at which point it usually answers itself in about a minute.
Take the pricing example. The choice is five percent or ten. So the useful question is not "what is our true unit cost", it is "is there any plausible unit cost at which those two options swap places?" If the answer is no, and it often is, then the cost model needed to be roughly right, and the real work was always the competitive response. That is the whole calculation. Write the decision at the top of the page, list the options, and ask what would have to be true for the recommendation to flip. The distance to that flipping point is your budget, and anything finer is spending.
This is the useful reading of 80/20, and it is worth being precise about what that rule is and is not. Vilfredo Pareto observed in the 1890s that wealth concentrated in a few hands. He never stated the rule that carries his name. Joseph Juran generalised the idea into "the vital few and the trivial many" for quality control in the 1940s and attached Pareto's name to it, then in 1975 published a piece called "The Non-Pareto Principle; Mea Culpa" explaining that the name had been his own mistake. There is no measured constant. Treating 80/20 as arithmetic ("I'll do 20% of the work") is a misreading of a rule of thumb that was already a misattribution. The usable version is a posture: effects concentrate, so find the concentration before you spread effort evenly.
Juran also revised his own phrase, late on, from "the trivial many" to the useful many, because trivial implied the rest had no value. That correction matters more than the arithmetic, and it is the guardrail on this entire article. The 80% you did not examine is not worthless. It is deferred. Which means you owe someone a sentence about it.
The same estimate at three levels of effort. The recommendation is identical in all three, which is the signal that the extra effort was spent, not invested.
What Sets the Budget: How Hard It Is to Undo
If the budget is not a fixed number, something has to set it. The most reliable variable is not the size of the decision, it is the cost of being wrong, and that is mostly a question of how easily the decision reverses.
Jeff Bezos put this into Amazon's 2015 shareholder letter in a form that has survived because it is genuinely useful. Some decisions are, in his words, "consequential and irreversible or nearly irreversible, one-way doors, and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation." Most are not: "they are changeable, reversible, they're two-way doors," and those "can and should be made quickly by high judgment individuals or small groups."
The letter also carries a number, that most decisions should be made with about 70% of the information you wish you had. Take the shape of that and leave the figure: it is one executive's stated policy, not a measurement. What survives is the structure. The door type sets the budget. Repricing a contract you can reprice again next quarter is a two-way door and deserves a rough number fast. Committing to a platform, publishing a figure to a regulator, or telling a client something they will build a plan on, those are one-way doors, and the extra week is not indulgence, it is proportionate.
The mistake this catches runs in both directions. Teams agonise over reversible decisions because they feel important, and wave through irreversible ones because they arrived on a quiet Tuesday. Size and reversibility are different axes, and they get conflated constantly.
Four decisions scored on how big they feel and how hard they are to undo. The verdict column is set by the second question only, which is the point.
Where This Goes Wrong
The honest objection is that "good enough" is what someone says right before they ship a bad number, and the objection is fair. A rigor budget is not a licence, it is a constraint you set on yourself and then have to defend. Three ways it turns into an excuse.
Setting it after the fact. If the precision you declare is always the precision you happened to reach when time ran out, you do not have a budget, you have a rationalisation with a name. The test is whether it was written before the work started.
Hiding the gap. An estimate delivered as though it were a measurement is worse than an over-engineered model, because the reader cannot tell and will treat it as solid. This is where the useful-many correction earns its place: the deliverable needs one line saying what you did not examine. Support costs are allocated by headcount, which overstates the smaller line by roughly ten percent, and correcting it does not change the ranking. One sentence turns a corner cut into a disclosed assumption, and it is most of what separates fast senior work from fast junior work.
Treating it as fixed. If the rough number lands close to the flipping point, that is the signal to spend more, and it is the only one that reliably justifies extending. Revise the budget on evidence, not on anxiety.
One situation where none of this applies: when the number itself is the deliverable, an audited figure or a regulatory submission, precision is not serving a decision, it is the product. Do not put a rigor budget on a thing whose entire value is being exactly right.
The Version of This You Can Do Today
Take whatever you are working on and write two lines at the top of the document, before you touch the analysis again.
This answer needs to be right to within ______.
Because the decision it feeds is ______, and the recommendation only changes if the number crosses ______.
If you cannot fill in the second line, that is not a sign to work more carefully. It is a sign that you do not yet know which decision this feeds, which is a framing problem, and a rigor budget cannot fix it. Go and ask.
Then find the one assumption with the widest range and the most leverage on the answer, and test only that. Leave the rest written down as stated assumptions. Most analyses have exactly one input where the honest uncertainty is wide enough to matter, and it is rarely the one that is most satisfying to refine.
The Point
Rigor is not a personality trait or a measure of how seriously you take the work. It is a resource, and spending it evenly across a problem is the same mistake as spending a whole budget on the first line item.
The senior move is not being fast, and it is not being thorough. It is knowing which one this decision is asking for, deciding that before the work starts, and saying out loud what you skipped.
Nobody is impressed by a number that is accurate to a decimal place the decision cannot see. Precision that cannot change the answer is just expensive handwriting.
Coming Up in This Series
Setting a rigor budget assumes you already know which decision the work feeds, and that knowledge almost always comes from a conversation rather than a document. Someone talks for forty minutes, the important constraint arrives in the middle of a tangent, and a week later everyone who was in the room remembers a different version of what was agreed. Nothing you do afterwards recovers what was not captured while the person was still talking. Next: structured listening, and how to play a rambling brief back as three clean themes before the meeting ends.
Part of an ongoing series on turning messy problems into clear recommendations.