We hand over a document, and it has to hold up without us
Plenary's work ends in one written record of how a decision was made. It names the question, the evidence, the alternatives, the limits of confidence, and the moment the board chose. It is written for a reader who was not in the room and does not agree with the outcome. When the decision is reopened, that is the person holding the file.
A deck records the answer. It does not record the reasoning.
Slides carry the conclusion and drop what produced it. The option rejected early, the data that never arrived, the assumption holding the numbers up: none of it survives the deck. When a regulator or a new board reopens the decision, the deck is the wrong object to hand them. So we build the record instead, and we build it as the work goes, not at the end.
The parts of a recordEvery record has the same parts, even the ones that weaken our case
Every record carries the same parts. The order shifts with the file. Nothing is dropped because it turned out to be inconvenient.
The question as put
The decision in one sentence, in the form the board will vote on.
What was known, and on what date
Every piece of evidence carries the date it reached us.
What was not known
Each gap named, with the reason it stayed open.
The alternatives
Every option considered, including the ones set aside early, and the reason each was set aside.
Limits of confidence
Where the evidence is thin, where it is contested, where a single assumption carries the result.
What would make this wrong
The conditions under which the recommendation fails, written before the vote.
The recommendation
Short, and the reasoning sits behind it rather than inside it.
Dissent
Any position that did not carry, in the words of the person who held it.
Sources
Every document, interview, and dataset the analysis used, with dates.
The minute
What the board decided, on what basis, and what it asked for next.
The contents page of one record: energy, a tariff decision
Redacted from a live file. The client is not named and the figures are out, and the headings sit in the order the board read them.
Sector
Energy and utilities
File status
Reopened by the regulator.
Engagement
Decision framing, 2 weeks, then evidence review, 4 to 8 weeks, then board handover, 2 weeks.
Summary
A network operator set a tariff on evidence that was still moving, and the regulator reopened the file nineteen months later.
The board session ran the 40 minutes the record was written for, and most of it went on the first two headings and the eighth. 19 months later the regulator opened the same file at the fifth heading and read backwards.
Read the tariff caseHow a regulator reads a record
A regulator does not read for the recommendation. They already know what was decided. They read to find out whether the decision was defensible on what was known at the time, or only defensible afterwards.
So they go to the dated evidence first, then to the options, then to the gaps. One question sits behind every page they turn. Did this board claim to know something it could not have known, or ignore something it did? A record answers that in its structure, before anyone has to ask.
A court reads it the same way. So does a new board, and so does the executive who inherits the file. None of them were in the room. The record is what they get instead.
“The regulator asked for our reasoning. We had it in writing, and it was dated.”
Every record carries a section on what would make it wrong
It is written before the vote, while the outcome is still unknown. It names the assumptions carrying the result, and the conditions under which each one fails.
Boards sometimes ask us to soften it. We leave it as written, because it is the part that does the work when someone reopens the file.
A recommendation that was right and a recommendation that was lucky look the same in hindsight. The difference is visible only if someone wrote it down at the time.
Insight: When the evidence arrives late
Two of the four write it, and no juniors sit on the analysis
Partners cover energy and health systems. Principals cover financial services and public infrastructure. Nadia Fournier writes most of the firm's method, and the two on the file read each draft against it. A full cycle runs 8 to 12 weeks, and most of that time goes into the record rather than into meetings about the record. We cut any sentence the meaning survives without. Writing for the reader who will disagree tends to make a document shorter.
The file stays with the client, in a form they can edit
We hand over the working document in editable form, not a PDF. The client owns it and amends it after we leave. A record that cannot be corrected stops being a record the day the first fact changes. The live version is theirs. We keep an archive copy, which holds the file as it stood on the day of handover, and that is the copy a case cites.
Two copies exist
The live version. With the client, editable, theirs to amend.
The archive copy. Frozen on the day of handover. This is the one a case cites.
Who can open a record, and for how long
The engagement letter sets access at the start. It does not change later without the client saying so.
The client
Permanently, in editable form, from the day of handover.
The two named on the file
For the life of the engagement, and afterwards for citation only.
Anders Vik
He keeps the archive, and retrieval runs through him so every request for a file is logged against a name.
A regulator or a court
Through the client, never through us. We confirm what a record contains on the client's instruction and not otherwise.
Nobody else
The file itself goes no further than the names above. What we publish from a file is the reasoning, with the client unnamed and the figures redacted.
The archive opened with the first file, not once there was enough work to justify one.
112 files between 2019 and 2026, in sectors 9 regulators oversee
9 sector regulators whose tests files in the archive have been read against
2 names on every one of those files, one of them always a partner
8 to 12 weeks, the length of a full cycle
40 minutes, the board session a record has to survive
The counts come from the archive and can be checked against it. The cycle length and the board session are how the work is set up rather than things the archive measures. A count of clients would say nothing about whether the work held up.
A decision nobody could reconstruct
A water utility asked us to review an investment programme running 11 years. The previous contractor had left 400 slides and no record of the alternatives. Nobody still at the utility could say which options were considered at the start, or why any of them were dropped. We could not rebuild the reasoning, because the reasoning had never been written down. The record on that file begins in the present, and says so on its first page.
Read the water utility caseThe document opens with the question and closes with the minute
The record is open from the start. Each stage adds to it, and nothing is written twice.
Decision framing, 2 weeks
Narrow the question to one the board can actually answer. Name the evidence that would change the answer. The record opens with that question, in the words the board will use.
Evidence review, 4 to 8 weeks
Gather the evidence and check it, then record where confidence runs out and what cannot be known at all. Most of the record is written in this stage.
Board handover, 2 weeks
Put the record in the form a 40 minute board session needs, then run the session. The minute closes the file.
Two of the four are named on each file. Anders Vik keeps the record.
Mireille Osei
Partner, energy and utilities
Sat on an energy regulator's appeal panel for 4 years.
Tomas Reinholt
Partner, health systems
Built clinical trial protocols before Plenary.
Nadia Fournier
Principal, financial services
Writes most of the firm's method.
Anders Vik
Principal, public infrastructure
Keeps the decision archive, the one cited in cases.
Tell us what the board has to decide, and when
Send the decision rather than the brief. All four see what arrives. The person who replies is the one who covers your sector, and that person would be on the file. We will say in the first conversation whether a record is the right instrument, and we will say so when it is not.