When I sold my company, the buyer's diligence team asked for the reasoning behind a pricing change we had made three years earlier. We had the board deck. We had the email announcing it. What we did not have was anything that said what we had expected to happen, who had agreed, and whether it had gone the way we thought. The change had worked, as it happened. I could not prove that we had known why.
That is the difference between a decision log and a decision record, and most companies keep the first while believing they keep the second.
A log is not a record
A decision log is a list: date, decision, owner. It answers "what did we decide?" and it is better than nothing, which is why templates for it are everywhere. What it cannot answer is the question a board member, an acquirer, a new executive or a future version of you actually asks: why did we think this would work, and were we right?
A decision record answers that. It is longer than a log entry and shorter than a memo, and it has a fixed shape so that the reader knows where to look.
The eight parts
The question, and why now. What is being decided, what triggered it, and what it costs to wait. One paragraph. If the trigger is not written down, the reader cannot tell a considered move from a reaction.
The hard lines. What the decision must not do, whatever path is chosen. The budget it cannot exceed, the customer it cannot lose, the value it cannot violate. These are the constraints the paths were judged against.
The paths considered. At least two, each with an honest case. Nutt's 1999 study found that seriously weighing more than one option was the strongest single predictor of a decision surviving, and the record is where that weighing is visible. A record with one path is an announcement.
The conditions. For the chosen path, the handful of things that would have to be true for it to work, written so they could be proven false. "Mid-market customers will accept a 15 percent increase" is a condition. "Pricing is an opportunity" is a hope. Each condition carries what was known about it on the day and where that knowledge came from.
The people asked, and what they said. Who was asked for their read, on which conditions, when, and what they answered, before they had seen anyone else's answer. Silence is recorded too, as silence, and kept separate from the people who were never asked. Kahneman, Sibony and Sunstein call this the hygiene of the record: the judgments are kept independent and kept.
The commitment. One name, one date, and the confidence that person had at the moment they committed. Rogers and Blenko's 2006 work on decision roles found that the single most common cause of decisions that stall or drift is that nobody has the D: several people are consulted, several approve, and no one is the person who decided. The record forces the name.
The tripwires. The signals, set on the day, that mean the decision needs another look, each with a date. Annie Duke's version is the kill criterion, a state and a date that mean stop. Without them, the decision has no way of telling you it is failing.
The look-back. The date, booked at commitment, when the record is reopened and two questions are asked separately: was the call sound on what was known, and how did the outcome compare to what was expected. This is the part that turns a company's history into something it can learn from, and it is the part that never happens unless the date was written down first.
Who the record is for
Write it for the reader who was not in the room. That reader might be a board member reading it next quarter, a new head of sales reading it in a year, an acquirer's lawyer reading it in three, or, most often, you, reading it in eighteen months and unable to remember why the obvious choice was obvious.
Two things follow from that. The record has a public face and a private one. The question, the paths, the chosen path, the conditions, the tripwires and the dates can be read by anyone the decision affects. The drafting notes, the individual answers and the dissent belong to the people who gave them, and the reader who was not in the room does not need them to understand what was decided.
And the record has to be written in the words of the decision, not the words of the tool that holds it. If it needs a legend to read, it will not be read.
Where Midfire fits
Midfire's record is built from these parts, in this order, and it keeps the two faces apart: the readout that a folder or a board can read, and the drafting record with the individual answers, which stays with the people on the decision. The commitment carries one name and one confidence, the tripwires carry dates, and the look-back is booked when the name goes on. The eight parts are older than the product. The product's contribution is that they are there every time, without anyone having to remember the list.
The record I wish I had
For the pricing change, it would have been a page. The question and the trigger (a competitor's move). Two paths (raise across the board, or raise on the new tier only). Five conditions, two of which the head of sales could have checked in a week. Four people's reads. My name, at 70 percent. A tripwire on churn at ninety days. A look-back at six months. The buyer's diligence team would have had their answer in five minutes, and so would I, three years earlier, when I could still have used it.