First of three pieces on investment committee architecture. The second piece (on pre-mortems for investment committees) and the third (on recording dissent) are forthcoming in the series.
An investment committee paper exists to do one thing: let a committee take a decision its members can defend afterwards. The defence is owed to a regulator, an auditor, a beneficiary, or the committee's own future selves, who will read the paper again when the outcome is known and ask whether the reasoning held. Most papers are written as if the only audience were the room on the day. That is the first mistake, and it shapes every other one.
The single test that separates a good paper from a bad one is reconstructibility. Can a reader who was not in the room rebuild the committee's reasoning from the paper and the minute alone. A good investment committee paper lets the committee reconstruct its own reasoning a year later; a bad one is a rubber stamp with numbers attached. A paper that passes the test has a defined decision, a named recommendation with an owner, a base case built against a reference class, an honest downside, a surfaced assumptions register, a risk register with owners and triggers, and an explicit statement of what would change the recommendation. A paper that fails it has numbers, a conclusion, and a committee reduced to a quorum.
The recurring failures are structural, not cosmetic, and they are predictable. They are also addressable, because each missing part of the anatomy maps to a discipline that already has a name and a citable lineage. This essay walks the anatomy part by part: not as a template to copy, but as the parts a defensible paper contains and the reason each one earns its place.
What is an IC paper actually for?
Consider a paper that read well and decided nothing. It ran to forty pages. It had a polished narrative, a confident base case, a chart pack that anticipated every obvious question, and a recommendation in the final line. The committee approved it inside fifteen minutes. A year later, when performance diverged from the modelled range, nobody could say which assumption had carried the decision, who had owned the risk that materialised, or what the committee had told itself would have to be true. The paper had communicated a conclusion. It had not recorded a decision.
That distinction is the whole subject. A paper that communicates a conclusion persuades the room on the day. A paper that records a decision survives the day, because it captures the basis on which the decision was taken in a form a later reader can test. The Myners Review put this in plain terms in 2001: institutional investment decisions should rest on clear written objectives, and decision-makers should be able to demonstrate the basis of a decision after the fact [1]. The demonstrability is not an administrative afterthought. It is the function of the paper.
The UK Corporate Governance Code 2024 pushes the same way. It sharpens governance reporting toward decisions and their outcomes, and it names group-think as a risk the boardroom must actively guard against [2]. A paper built only to persuade is a group-think delivery mechanism: it gives the committee a single confident narrative and no structured invitation to dissent from it. A paper built to be reconstructible does the opposite. It exposes the parts a reviewer would want to attack, which is exactly what an investment committee, or its sector counterpart in financial services, is there to do.
What is the committee actually being asked to decide?
The sentence most papers omit is the decision sentence. Not "here is an opportunity", but "the committee is asked to approve X, by date Y, on the recommendation of Z". Without that sentence the committee cannot tell whether it has decided anything. It can discuss, react, and nod, but a discussion is not a decision, and a nod is not a mandate.
The decision sentence does more than frame the paper. It binds the paper to the committee's terms of reference, which is the document that defines what the committee is empowered to decide in the first place [3]. A paper that asks for approval of something outside the committee's authority is not asking the committee to decide; it is asking it to overreach. A paper that asks for nothing specific is asking the committee to abdicate. The terms of reference and the decision sentence are the two ends of the same thread: the committee's mandate at one end, the specific exercise of it at the other.
The decision sentence also disciplines the analyst who writes it. To state the decision precisely, the analyst has to know the date by which it must be taken, the action it authorises, and the person standing behind it. Vagueness in the decision sentence is usually vagueness in the underlying work, surfaced. A paper that cannot name the decision is often a paper that has not finished thinking about what the decision is.
Who owns the recommendation?
A paper that presents options is not the same as a paper that recommends. Presenting three structures and inviting the committee to choose can look like neutrality. More often it is the analyst declining to own a view. The committee is owed a recommendation, and the recommendation is owed an owner.
Ownership here is precise. The responsibility assignment matrix, commonly rendered as RACI (responsible, accountable, consulted, informed), insists that every item has exactly one accountable owner [4]. Applied to the recommendation, that means one named person is accountable for the call, not the working group, not the function, not the deck. The accountable owner of the recommendation is not the same as the accountable owner of the work behind it. A team can build the analysis; one person recommends acting on it. The minute records who that was.
This matters most when the outcome is bad. A recommendation with no named owner diffuses across everyone who was in the room, which in practice means it lands on no one. The committee that approved it cannot interrogate the recommender's reasoning later, because there is no recommender on the record. Naming the owner is not about apportioning blame after the fact. It is about making the reasoning attributable in advance, so it can be questioned while the questioning still matters. This is the natural output of the firm's investment and commercial analysis service, where a board-grade case is built to carry exactly that kind of named, defensible recommendation.
What is the honest downside, not the token sensitivity?
The second near-universal failure is the downside. A paper flexes one input by a round number, labels the result the worst case, and moves on. That is a token sensitivity, not an honest downside, and an experienced committee can smell the difference. The honest downside is a coherent adverse state: a description of the conditions that produce a poor outcome and the way those conditions interact, rather than a single dial turned the wrong way. The distinction between a single-input sensitivity and a coherent scenario is itself a subject; here it is enough to say the committee needs the second, not the first.
The deeper problem is where the base case comes from. Kahneman and Lovallo drew the distinction between the inside view and the outside view: the inside view anchors on the specifics of this plan, while the outside view places the decision in a reference class of comparable past cases and reads the base-rate distribution [5]. Most base cases are pure inside view. They are the sponsor's plan, lightly haircut. An honest base case is built against a reference class first, then adjusted for what is specifically particular about this case. The outside view is usually less flattering, which is precisely why it is more useful.
The forward-looking version of the same discipline is the pre-mortem, in which the team assumes the decision has already failed and works backwards to the plausible causes [6]. It is named here and pursued properly in the next piece in this series; the point for the paper is structural. A paper that contains a coherent adverse state, a reference-class base case, and the seeds of a pre-mortem has done the downside honestly. A paper with a single-input flex has decorated the downside and skipped it. The failure modes of return metrics are part of the same terrain, which is why the note on where IRR misleads investment committees belongs alongside the downside section of any appraisal paper, and why the comparison of net present value and payback for infrastructure committees matters wherever the decision criterion itself is in question.
Where are the assumptions, and who owns the risks?
The load-bearing middle of a paper is two registers: the assumptions and the risks. Both fail in the same way, by being buried or generic, and both are fixed in the same way, by being surfaced and owned.
The assumptions register is the schedule that surfaces the material assumptions in one named place, each with its source and its sensitivity, rather than leaving them embedded in model cells or scattered through prose. A reviewer who can see the assumptions in one schedule can test the recommendation against them. A reviewer who has to reverse-engineer them from the model is not reviewing; they are excavating. The register is consistent with the Myners principle that the basis of a decision be demonstrable: the assumptions are the basis, made legible [1]. Where the assumptions derive from fund terms, the constraints are often set before the model opens, which is why the note on reading a fund LPA before opening the model is a useful companion to the assumptions register.
The risk register fails when it becomes a generic list. "Market risk, execution risk, regulatory risk" owns nothing and triggers nothing. A risk register that informs a decision carries two things for every line: one named accountable owner, and one defined trigger that moves the risk from watch to live. The owner is the A in RACI, the single person accountable for watching that risk [4]. The trigger is the observable event that says the risk has arrived, stated in advance so that nobody has to argue about it in the moment. A risk without a named owner is not owned; a risk without a trigger is not monitored. The generic three-line list satisfies neither test, which is why it can sit in a paper for years without ever doing any work. This owned, triggered posture is the same independent challenge to prevailing assumptions that runs through the firm's work in financial services and investment.
What would change the recommendation?
The section almost every paper lacks is falsification: a statement, in advance, of the evidence that would overturn the recommendation. The idea is Popperian in origin, but the practical move is simple. Name the observations that would change the call before the call is made.
A paper that cannot say what would change its recommendation has not made an argument; it has made an assertion. An argument is contestable by construction: it identifies the conditions under which it would be wrong. An assertion is not. It simply states the conclusion and trusts the reader to agree. Committees are routinely asked to approve assertions dressed as arguments, and the falsification section is the part of the anatomy that forces the distinction into the open.
Three disciplines sit adjacent to falsification and reinforce it. The pre-mortem surfaces the failure paths in advance [6]. The reference-class base case names the population the decision belongs to, and therefore the base rate against which the optimistic case must be judged [5]. The decision journal, the contemporaneous record of the decision, the reasoning, and the expectations on which it rested, lets the committee review the call against the outcome rather than against hindsight [7]. The paper plus the minute is the committee's decision journal. A falsification section is what makes that journal worth keeping, because it records not just what the committee believed but what would have proved it wrong.
What belongs in the paper, what belongs in the pack, and where does dissent go?
A paper is not its evidence. The paper carries the reasoning; the supporting pack carries the evidence. Collapsing the two produces the forty-page object that nobody can navigate and the committee approves out of fatigue. The model, the data tables, the source documents, and the sensitivities all belong in the pack, indexed and retrievable. The paper references them and stands on its own at the level of argument. The discipline is not length for its own sake; it is the separation that lets a reader follow the reasoning without wading through the workings, and audit the workings without re-reading the reasoning. The firm's growth strategy and capital allocation work observes the same separation between the board paper and the analysis behind it.
The last part of the anatomy is the one most papers never consider: where dissent goes. A committee that approves unanimously every time is either remarkably well aligned or quietly suppressing disagreement, and the paper should leave room for the second possibility to be recorded. A good paper does not assume consensus; it makes space in the structure for a dissenting view to be minuted against the record, so that a later reader can see not only what was decided but what was argued. This is the subject of the third piece in this series, on recording dissent, and it is enough here to flag the requirement: the paper should leave room for dissent, and the minute should capture it.
Synthesis: the paper as a control
Pull the parts together and the spine is single. The reconstructibility test is the function the whole anatomy serves: a defined decision, an owned recommendation, an honest downside built against a reference class, surfaced assumptions, risks with owners and triggers, and a falsification condition all exist so that a reader who was not in the room can rebuild the reasoning from the paper and the minute alone. None of the underlying disciplines is proprietary. The Myners Review, the FRC Code 2024, the CFA Institute's framing of the fiduciary's staged duty [8], the pre-mortem, the outside view, and the decision journal are all citable, and an investment committee can ask for them by name.
The reason to insist on the anatomy is not documentation for its own sake. The structure of the paper is itself a control on the analysis behind it. A section that demands an owner for every risk forces the analyst to confront risks that a persuasive deck would let them leave abstract. A falsification condition forces the analyst to state what would prove the recommendation wrong, which is the question optimism most wants to avoid. The discipline of the paper disciplines the work. At the firm, we have come to treat the investment committee paper as the place where the discipline of the analysis is made visible, not the place where a decision already taken is dressed for approval.
That is the case for the anatomy, and it is also its limit. A well-built paper does not guarantee a good outcome; capital at risk does not work that way. What it does is make the reasoning attributable, contestable, and reviewable, which is the most a committee can ask of the artefact it decides on. The open question the next two pieces take up is whether a committee that never records dissent is structurally exposed, however well its papers are otherwise built. A paper can be reconstructible and still wrong. It cannot be defensible if the disagreement in the room left no trace.
Notes
1. Institutional Investment in the United Kingdom: A Review (the Myners Review), Paul Myners, HM Treasury, March 2001. https://uksif.org/wp-content/uploads/2012/12/MYNERS-P.-2001.-Institutional-Investment-in-the-United-Kingdom-A-Review.pdf
2. UK Corporate Governance Code 2024, Financial Reporting Council, January 2024 (in force for financial years from 1 January 2025; Provision 29 from 1 January 2026). https://www.frc.org.uk/library/standards-codes-policy/corporate-governance/uk-corporate-governance-code/
3. The Chartered Governance Institute UK and Ireland, guidance notes on board and committee practice, including terms of reference for board committees. https://www.cgi.org.uk/resources/factsheets
4. A Guide to the Project Management Body of Knowledge (PMBOK Guide), Project Management Institute: the responsibility assignment matrix (RAM), commonly rendered as RACI (responsible, accountable, consulted, informed). https://www.pmi.org/
5. Daniel Kahneman and Dan Lovallo, "Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking," Management Science 39(1): 17 to 31, 1993. https://ideas.repec.org/a/inm/ormnsc/v39y1993i1p17-31.html
6. Gary Klein, "Performing a Project Premortem," Harvard Business Review 85(9): 18 to 19, September 2007. https://hbr.org/2007/09/performing-a-project-premortem
7. J. Edward Russo and Paul J. H. Schoemaker, Decision Traps: Ten Barriers to Brilliant Decision-Making and How to Overcome Them, Doubleday/Currency, 1989. https://archive.org/details/decisiontrapsten00russ
8. Investment Governance for Fiduciaries, Michael E. Drew and Adam N. Walk, CFA Institute Research Foundation, April 2019. https://rpc.cfainstitute.org/research/foundation/2019/investment-governance-for-fiduciaries