The Cost of Treating Every Decision Alike
Reversible and irreversible decisions carry different risk. Most organisations run them through one process, and pay for it at both ends.
Most decision processes are calibrated for the wrong risk.
Organisations tend to build one decision process and then apply it everywhere. It has a template, a forum, a set of approvers and a rhythm. It was usually designed after something went wrong, which means it was designed to prevent one particular failure from happening twice. The process then spreads, because a process that already exists is easier to reuse than a process that has to be argued for.
One gate, two unrelated risks
What that produces is a single gate standing in front of choices that have nothing in common. Renaming a product tier and withdrawing from a market arrive at the same committee, in the same format, against the same evidence bar. One of those can be undone in an afternoon. The other cannot be undone at all.
Running a heavy process over a light decision looks like diligence. It reads as care, and it is defensible in any review. What it actually produces is delay, and delay carries a price that nobody records, because the cost of a slow decision never appears as a line item.
A reversible decision costs more to deliberate than to undo.
The distinction worth drawing is not between large and small. Size is a poor proxy, and teams routinely spend three weeks on a cheap choice and twenty minutes on an expensive one. The distinction is whether the decision can be walked back, and at what price.
Size is the wrong axis
A reversible decision is one where the cost of being wrong is simply the cost of changing course: a checkout copy change, a pricing test on a single segment, an extra report in the weekly pack. Each can be run, watched and withdrawn. The information that would settle the argument in the room is usually cheaper to buy by acting than by discussing.
The failure mode is not the wasted hour
Reversible decisions are numerous and they arrive constantly. They consume the attention of the people best placed to judge the permanent ones, and they consume it first. By the time an irreversible decision reaches the table, the room is tired, the calendar is full, and the appetite for a long argument has already been spent somewhere cheaper.
There is an asymmetry underneath this that is worth stating plainly. Treating an irreversible decision as reversible is the more expensive error, and it is the one every governance process is built to prevent. Treating a reversible decision as irreversible is cheaper per instance, which is exactly why it goes uncorrected and accumulates. The first mistake produces a post-mortem. The second produces an organisation where nothing moves quickly and no single meeting is ever to blame for it.
The cost of a slow decision never appears as a line item, which is why it is the easiest cost to keep paying.
The irreversible ones rarely announce themselves.
Irreversible decisions are easy to name in hindsight and hard to see in advance. They seldom arrive labelled as consequential. They arrive as a technical choice, a first senior hire, a clause in a partner agreement, a default setting nobody expects anyone to change.
A few that reliably turn out to be one-way:
- The data model that the next four years of product arguments will be conducted inside.
- An exclusivity term agreed to close a deal in a quarter that needed closing.
- The first senior hire into a function, who then sets the bar for everyone hired after.
- A pricing structure customers anchor to within weeks and remember for years.
Routine on the agenda, permanent in effect
Each of these can be presented as routine and reviewed as routine. What makes them irreversible is not the act itself but everything that gets built on top of it before anyone thinks to revisit it.
The tell is dependency, not size
A decision is close to irreversible when other work will start depending on it before any results come back. Once that dependency forms, undoing the choice means undoing the work, and that bill is paid by people who were never in the room when it was taken.
There is a second reason these get missed. Review processes are good at examining the decision on the agenda and blind to the one that was settled in order to write the agenda. By the time a proposal is formally reviewed, the irreversible part is often already inside the framing: the architecture is assumed, the partner is assumed, the structure is assumed, and what remains to be approved is the last and most reversible layer.
Reversibility is a property of the system, not of the choice.
The move that changes the economics is to stop treating reversibility as a fact to be discovered and start treating it as a property to be designed. Two organisations facing an identical choice do not face identical risk, because one of them can undo it and the other cannot.
- A migration is irreversible if the old system is switched off on cutover day, and reversible if both run in parallel for a quarter.
- A price change is irreversible if it reaches the whole base at once, and reversible if it reaches new customers first.
- A senior hire is close to irreversible when the role is shaped around the person, and closer to reversible when the role is defined before the search opens.
What the room should argue about instead
This changes what the room should be arguing about. The question stops being whether everyone is sure and becomes what would make this undoable, and what that costs. Sometimes the answer is that reversibility is too expensive to buy, and the decision is correctly handled as one-way. That is a conclusion worth reaching deliberately, rather than a default arrived at by omission.
It is the same discipline as the preparation that makes a decision possible: most of the quality is set before anyone votes. The wider set of these patterns sits under decision architecture.
Sorting the two is where the thinking becomes work.
In practice this rarely fails at the level of principle. Most leadership teams agree immediately that one-way and two-way decisions are different. It fails at the level of the operating system, because nothing in the calendar, the template or the approval chain reflects the distinction, and so the default stays in force.
The distinction has to live in the calendar
The work is unglamorous and specific:
- Name which decisions in the current quarter are one-way, out loud, before they are taken.
- Give the reversible ones a shorter path, a named owner and a date on which the result is read.
- Buy reversibility where it is affordable, and say plainly when it is not.
Most of all it means protecting the scarce commodity, which is not time but the concentrated judgment of a small number of people. That is the shape of the work we do with commercial and product teams: not producing the decision, but building the conditions in which the right ones get the attention they need and the rest stop competing for it.