Last time I wrote about the seventh hat, the one nobody wants to wear, because committing to a decision without a record means carrying the accountability alone. But there is a cost on the other side of that too, one that the company pays whether or not anybody wears the hat. Every decision that gets made without a proper record is a small loan the company takes from its own future. And like all loans, it has interest and it has principal, and nobody is keeping track of either.
I have started calling this Decision Debt, because the parallel to technical debt is almost exact and the name makes visible something that has been invisible for as long as companies have existed.
How the debt gets taken on
It happens the same way every time. A decision needs to be made. Somebody says yes. The yes is real, it moves things, money flows, people are hired, vendors are switched, rules are bent. But the yes lives in a WhatsApp message, or a nod in a corridor, or a reply buried on the eleventh screen of a forwarded thread. The decision happened. The record did not.
That gap between the decision and its record is the debt. One gap is a small loan. A hundred gaps, accumulated over a year, across a company of two hundred people, is a balance sheet nobody has ever looked at because nobody knows it exists.
The debt grows with headcount, because more people means more decisions, and more decisions means more gaps. It compounds with every resignation, because in most companies the record is the person, and when the person leaves the record leaves with them. And it is invisible in the way that only compounding things can be, where each individual gap is too small to notice and the accumulated total is too large to see.
The interest
The interest on Decision Debt is paid every day, by many people, in small amounts, and it never appears on any report.
It is paid by the person who spends twenty minutes looking for a decision that was already made, and by the colleague who spends an hour helping them look. It is paid in the meeting where a question that was settled last quarter gets reopened and worked through again, because nobody in the room knew it had been settled, because the settlement lived in somebody's sent folder and that person is in a different meeting. It is paid by the manager who cannot explain the rule she is enforcing, and by the person on the receiving end of that rule who decides the system cannot be trusted.
It is paid by the procurement executive who makes his request smaller than it needs to be, because a clear yes is not available and a smaller ask is a smaller thing to be exposed on later. It is paid by the HR head who keeps a private folder of written confirmations, because the company's own records cannot be relied on, and whose diligence is invisible to everyone including herself.
It is paid every time somebody joins and spends four months rebuilding an understanding of why things are the way they are, most of which existed in somebody's head and could have been handed over in an afternoon if it had ever been written next to the decision it belonged to.
None of that has a cost code. Nobody submits a claim for any of it. It is absorbed into salaries that were going to be paid anyway, which is exactly why it can run for a decade without anyone objecting.
The principal
The principal on Decision Debt is paid rarely. When it comes, it comes all at once, at a moment you did not choose.
An investor asks who approved your largest supplier switch. The room goes quiet. Nobody can produce the decision. The investor adds a condition to the term sheet, and that condition sits in the closing documents permanently, so every future diligence sees that somebody once had to require your company to prove it could approve things.
Or a dispute. Two people disagree about what was agreed, and there is no record to resolve it against, so the dispute runs on memory and seniority and politics rather than on facts, and whichever way it settles, the relationship is damaged.
Or a grievance. Two employees compare notes and find they were treated differently for the same situation. The basis for each decision was never recorded, so there is no way to show that the two situations were actually different. Maybe they were. Maybe they were not. Nobody can tell, because neither decision exists as anything more than a fact about what happened.
The principal is dramatic and everybody fears it. But the principal is not where most of the money is. The interest is larger, because it is paid every day, by everyone, and it compounds.
Where the debt lives
You can get a rough picture of how much Decision Debt a company is carrying by asking one question: where does a typical consequential decision actually live right now?
In some companies, the answer is in people's heads. The decision was verbal. A nod. A phone call in a car between meetings. The company's decision history is distributed across the memories of its employees, unwritten and unsearchable. These companies are carrying the most Decision Debt and feeling it the least, which is exactly what makes it dangerous. Everything works beautifully until somebody leaves, or an outsider asks a question, or the company crosses about a hundred and fifty people and the founder's memory stops scaling.
In other companies, the answer is in email. The decision is written, timestamped, real. And it is buried in an inbox, tangled in a thread, forked across three different versions of the same conversation, indistinguishable from the forty non-decisions around it. These companies believe they have a record, because they do, technically. But a record you cannot find in under an hour is not a record. It is an archive nobody visits. The daily interest here is chronic: latency, chasing, retrieval cost, and the ever-present risk that the thread you need sits in a mailbox that was deactivated sixty days after somebody resigned.
In a few companies, the answer is in tools. A ticket in Jira, a workflow in the HRMS, an approval in the procurement system. These companies believe they have reduced the debt, and in a narrow sense they have. The ticket exists. It has a status. It says "approved."
But look at what the ticket actually contains. It tells you that somebody moved a status to approved. It does not tell you whether that person had the authority to approve it. It does not show you who else was involved, in what order, or whether the commercial check happened before or after the sign-off. The reasoning, if it exists at all, is scattered across comments that may or may not have been written at the time and that anybody with edit access can change after the fact. There is no sequence, no staged authority, no attached basis, and no way to prove the record has not been quietly altered since.
A status field is not an approval. It is a label that says "approved" in the same way a sticky note on a file says "done." On the day somebody disputes the decision, or a diligence process asks you to demonstrate how it was made, "it says approved in Jira" is worth exactly as much as "I remember it being approved." Both are claims. Neither is proof. And the difference between a claim and a proof is exactly the gap where Decision Debt accumulates fastest, because the company believes the debt has been paid when it has only been relabelled.
The tools were also built for other jobs, and each one covers only its slice, and the decisions that cross boundaries or fit no tool are still sitting back in email or in heads. The highest-stakes decisions, the exceptions, the reorganisations, the policy changes, are almost always the ones with no tool, because they are by definition the ones the tools were not designed for.
Most companies, honestly, are not cleanly at one level. They are jagged. Their purchase orders are in a system. Their policy exceptions are on WhatsApp. Their structural decisions are in somebody's memory. And the pattern is almost always the same: the riskiest decisions sit at the lowest level. The ones that most need a record are the ones least likely to have one.
Why naming it matters
I am not interested in making anyone feel bad about where their company sits. Every company arrived where it is for sensible reasons. Email is free and universal. Verbal trust genuinely works at small scale. The tools that exist were built for the jobs they were built for, and expecting them to also function as a decision record is unfair to the tools and to the people using them.
The reason to name the debt is simpler. A debt you cannot see, you cannot repay. And a debt you do not name, you cannot see.
Once you call it Decision Debt, three things happen that did not happen before. The daily cost stops being "how things are" and starts being an interest payment on a specific balance. The dramatic failure stops being bad luck and starts being a principal call on a loan you forgot you took. And the question of what to do about it stops being "we should document things better," which is vague and guilt-producing and never works, and starts being "which decisions are generating the most interest, and what would it cost to start recording them properly," which is specific and answerable.
The next thing I want to look at is how the debt is distributed, because it turns out that different kinds of decisions carry different kinds of debt, and a company that treats all its decisions the same will always overpay on the wrong ones.
Part of The Other Books, an ongoing series on the decisions companies forget.
