When something is not working, the first idea that comes to mind is almost always one more thing. A checkpoint, a form, a mandatory field, a weekly meeting, a person to hold the pieces together. Removing hardly ever comes to mind, and when it does we treat it as giving something up: something lost, work already done and thrown away.
The two halves of that sentence have two different explanations and neither one is laziness. Faced with a problem, the changes that add present themselves more readily than the ones that take away, and anyone moving fast risks stopping at the first that arrives, having never considered subtraction at all.1 On top of that sits another mechanism: a loss weighs more than an equivalent gain, and what we have already spent holds us in place even when spending more is not worth it.2 The first explains why subtraction does not occur to us. The second explains why, when it does, we discard it.
That is not how it works. Subtracting is not throwing away, it is concentrating. The time, the money and the attention that were spread across everything end up on a few things, and those few start working better: not because they improved, but because somebody is finally looking at them.
The rest of this piece tries to explain why, knowing all this, we keep adding. It is that subtraction needs three conditions that organizations rarely have at the same time: a decision made at the right moment, something written down to rest it on, and a real chance to get your hands back into the place where work changes hands.
Every addition costs little now and a lot over time
It is worth looking closely at why adding is a problem, and not simply a mistaken preference. The reason is that the cost of an addition is spread across two different axes, and neither of them is the one anybody looks at when deciding.
On the first axis, time. An addition costs little once and then has to be kept standing every day that remains. One more mandatory field in a form takes five minutes to add and is then filled in ten thousand times, checked whenever someone fills it in badly, exported into every file format that from then on will have to carry it, and migrated every time the software changes. The decision lasts five minutes, the consequence lasts as long as the company.
On the second axis, people. The cost of adding is concentrated on whoever adds and is paid immediately, so they see it. The cost of keeping is split into tiny parts among everyone who passes through it, every day, and none of them pays enough to complain credibly. Thirty seconds each is not an argument. It is just a fact with nobody to state it.
The result is that the addition presents itself as a small and reversible decision when it is almost always a large and permanent one, and it presents itself that way precisely to the person with the power to make it. Whoever has to live with it was not in the room.
How long a decision gets does not depend on who signs it
How much time to spend on a choice should depend on how long it will last, on its reach, on its complexity and on what it would cost to undo. These are questions you can answer in a few minutes, and they give you at least the order of magnitude of the time worth spending.
In practice the time a decision receives depends on a different variable: on who has to sign it. It is an implicit rule and nobody would defend it out loud, but you can see it in the calendar of any organization. A choice that goes past the director gets a review, three meetings and a document. A choice that goes past nobody is made by whoever happens to be there at the time.
The trouble is that the two are not correlated. The choices that reach the desk of whoever signs are often the loud, visible ones with an obvious price tag: changing supplier, buying a piece of software, moving a person from one department to another. Many of them are reversible. If the new supplier turns out to be wrong you go back, and the cost is some annoyance and a few weeks.
The choices that reach nobody are the quiet ones. An integration — the piece of software that moves data from one system to another — is almost always of this kind: a technician sets it up one afternoon, there is nothing to sign, it appears in no meeting. It stays there for as long as the company exists, and in the meantime everything else learns to depend on it. By the time someone notices it exists, undoing it costs more than making it ever did.
The practical consequence is not “decide better”, which is advice nobody can act on. It is a different one: reversible choices should be made quickly and in quantity. Many small bets, each of them abandonable without ceremony. The cost of a small bet is not the bet, it is the ceremony we build around it. If abandoning it requires a closing meeting, a document and an explanation to whoever championed it, we will not abandon it, and it will stay alive for reasons that have nothing to do with the work.
Documentation is the condition, not the paperwork
There is a link here that usually goes unnoticed, and it is the heart of the matter.
What we cannot explain briefly, we do not write down anywhere. Not out of bad faith: because writing down something confused is enormously expensive, and whoever tries stops after half a page and tells themselves they will do it when they have time. And what is written nowhere cannot be removed, because to remove something you first have to know where the things passing through it end up.
It follows that documentation is not bureaucratic paperwork, and it is not an act of kindness towards whoever comes next either. It is the condition for the choice to exist at all. Without it there is no choice to make: there is only a thing that keeps working in a way nobody controls.
It also follows that there is a measure nobody ever uses. The length of a document does not tell you how diligent its author was. It tells you how disorderly the thing being described is. Something done well is explained briefly, and when the explanation gets longer it is usually not the explanation that has improved.
What documentation does not tell you
It is easy here to widen the claim until it becomes false.
Documented does not mean understood. A document can be complete, up to date, and leave everyone none the wiser: it describes the steps without saying why they are in that order. And understood does not mean ready to be removed. Something can be perfectly clear and still be needed, in which case it stays where it is.
Documentation establishes one thing only: that the thing is assessable. It is a necessary and not a sufficient condition. Documentation does not tell us what to remove, it tells us what we would be capable of removing. Whether it is needed or not is a different question, and that one is answered by whoever knows the trade, by looking at what happens when it is gone.
The distinction matters, because otherwise “simplifying” becomes just another word for “cutting”. And cutting without knowing what you are cutting is not subtraction: it is loss.
The damage is in the handovers
So far one might conclude that the problem is complexity, and that the cure is therefore to simplify. It is a convenient and wrong conclusion, and it leads to taking apart things that were working.
The complexity that does damage is not the complexity of the trade. The trade is complicated by necessity, and rightly so: someone working in pharmaceuticals has complicated traceability because the problem is complicated. That complexity, when it is compartmented and automated, is perfectly bearable for years. Inside every compartment there is somebody who knows what they are doing, who notices when something goes wrong and who, if needed, can explain it.
It helps to draw the work as a graph, a network of points joined by lines. The people are in the points. The damage is in the lines: the spreadsheet office A prepares for office B with the columns B asked for years ago, of which it now uses two; the shared folder where something gets dropped in the hope that somebody picks it up; the connection between two programs that somebody set up once and nobody has touched since.
Inside a point there is a person. In between there is nobody, and things with no owner have no description either.
The problem concentrates precisely in the handovers between compartments. Nobody can say what happens if that spreadsheet stops being produced, because nobody knows who actually reads it, and when in doubt it keeps being produced. The cost of continuing is low and spread across everyone. The cost of stopping is concentrated on whoever proposes stopping.
We almost always automate the wrong part
From here you can see something usually discovered halfway through a project already under way. We happily automate inside the points, where the work is visible, where somebody can describe it and where the result can be shown to whoever signed off. On the lines we almost never automate, because there is nobody to ask for it.
The paradox is that the lines are also the part best suited to automation. Automating inside a compartment means encoding a trade, which is the irreducible part, the one where the exceptions are the substance of the work and not a defect in it. It is why so many automation projects run aground on a list of special cases that grows faster than anyone can shorten it.
A line, on the other hand, does not require knowing the trade. It requires knowing three things only: what passes, from where to where, in what form.
Which are exactly the three things needed to be able to decide to remove it. Automating a handover and documenting it are not two different jobs with the same price: they are the same job, and at the end of it you have bought both. Whoever did it automated a piece of work that belonged to nobody, and got in exchange the possibility of removing that piece, which they did not have before.
Why things stay
At this point you can see why certain things stay in a company far longer than would be reasonable.
Documentation does not decide whether something is needed. It decides whether we can afford to ask the question. It is an earlier and more modest step, and that is why it always gets skipped: it solves nothing, it only opens the possibility of solving.
What nobody can explain does not even make it onto the list of things that can be decided about. It is not rejected, it is not defended, it is not discussed. It was never on the table. It stays there by exclusion, and by exclusion it stays forever, because every year that passes something else learns to depend on it and the list of things we can still decide about gets one item shorter without anyone noticing.
It is why certain organizations are not complicated: they are immobile. It is not that they chose to keep everything. They never had the chance to choose.
How many of the things we have could we not remove even if we wanted to?
Further reading
Some of the links below are affiliate links: the price is the same for whoever buys, and a small share comes to me.
-
Subtract: The Untapped Science of Less, Leidy Klotz. Flatiron Books, 2021.
-
Essentialism: The Disciplined Pursuit of Less, Greg McKeown. Crown Business, 2014.
-
The Goal: A Process of Ongoing Improvement, Eliyahu M. Goldratt and Jeff Cox. North River Press, 1984.
-
The Checklist Manifesto: How to Get Things Right, Atul Gawande. Metropolitan Books, 2009.
Bibliography
Footnotes
-
G. S. Adams, B. A. Converse, A. H. Hales, L. E. Klotz, People systematically overlook subtractive changes, Nature 592, 258–261, 2021. https://doi.org/10.1038/s41586-021-03380-y ↩
-
D. Kahneman, A. Tversky, Prospect Theory: An Analysis of Decision under Risk, Econometrica 47(2), 263–292, 1979 (https://doi.org/10.2307/1914185). Sul costo affondato: H. R. Arkes, C. Blumer, The Psychology of Sunk Cost, Organizational Behavior and Human Decision Processes 35(1), 124–140, 1985 (https://doi.org/10.1016/0749-5978(85)90049-4). ↩