Building an Investment Committee Memo That Decides

An investment committee memo puts one decision in front of the people authorised to take it, in a fixed order: what is being asked, what the mandate permits, what the arithmetic says, and what would have to be true for the answer to change. A memo that reports a result without splitting it has reported a number the committee cannot act on.

A committee has Rs 500 crore of somebody else's money, a document saying what may be done with it, and forty minutes of eight people's attention. The attention is the scarcest of the three, and the memo is what turns it into a recorded decision. So the memo is built as a form rather than an essay: a form has fields, the fields run in an order, and a field left blank is visible. Every failure below is a gap that an essay would have hidden and a form makes visible.

Everything below is the order the memo gets built in, worked end to end on the Anantara Multi-Asset Portfolio, an invented Rs 500 crore multi-asset mandate run for an invented charitable endowment. Faiz Ahmad Ansari runs the mandate and writes the paper; Rukmini Deshpande chairs the committee that reads it. Every figure belongs to one stated twelve month period.

A memo's job stops at the paper: what it must carry, in what order, and what a reader of it can check unaided. How a committee reaches a decision once the paper is in front of it, who may vote, and how the decision is recorded, is taken up separately later in this subject area.

Step one: what is the ask, and why must it be a question?

The first line of an investment committee memoA short paper prepared for the people authorised to take a portfolio decision, setting out what is being asked of them. states the decision being requested. Not the topic. Not the area under review. The decision, in one sentence, shaped so that the only available responses are yes, no, or a named condition.

A household sits down on a Sunday evening and somebody proposes a discussion about the car. Two hours later they have talked about the car and nobody has decided whether to sell it. Change the opening to whether the car is sold before the monsoon and the same two hours end with an answer. A paper headed with a subject gets discussed; a paper headed with a question gets decided.

For the annual review, Faiz Ahmad Ansari has two openings that describe exactly the same paper, and only one of them makes the meeting resolvable.

Nothing in the analysis changes between those two openings, only whether the room has been handed something it can answer. Writing the ask as a question also forces the writer to find out whether there is a decision at all. Sometimes there is not, and the honest paper says so in its first line: this item is for information and no decision is sought. An honest paper of that kind is a better memo than one quietly hoping a decision will emerge.

Step two: why does the constraint check sit above the argument?

The second section is the constraint checkThe written limits the mandate carries, each tested against where the portfolio sits on the day the paper is written., and it goes immediately below the ask, before a single line of reasoning. The placement looks like a layout preference and it is a structural rule.

A proposal that fails the mandate does not need to be argued. If the ask cannot be granted without breaching a written limit, every line beneath it is spent on a question the room is not permitted to answer, and the room will argue it anyway. People argue about whatever is in front of them.

A committee reading the reasoning before the permission will argue the reasoning, whatever the permission turns out to say. Placement is doing work here that no amount of emphasis inside the text can do.

Step three: what does a distance carry that a tick destroys?

Now the content of the check. Most constraint checks are a column of names and a column of ticks: every line satisfied, nothing breached, next item. The paper is accurate and it has told the committee almost nothing.

Think of a fuel gauge with no needle, only a light that comes on when the tank is empty. The light is accurate, and useless for the one thing a driver does with fuel information: deciding when to stop. A tick reports that nothing is breached today and destroys the one part a committee can act on early, the distance left to each limit.

So each line carries three things: where the portfolio sits, where the written limit sits, and the gap between them in two units. One unit is the mandate's own, points or rupees. The other is the movement a market would have to deliver before the gap closes on its own, and that second unit is what makes the check forward looking without making it a forecast.

Written as distances, the equity line has ten points of room; the holding cap has four tenths of a point; the unlisted line has none and never will; and no price can settle the credit standing line, so it carries no distance at all.

What do the four distances actually measure on this mandate?

Run the check on the Anantara Multi-Asset Portfolio. The mandate carries an equity range of 50 to 70 per cent, a cap of 5 per cent on any single holding, a prohibition on unlisted holdings, and a minimum credit standing on the fixed income sleeve written as a policy rather than a symbol. The total is Rs 500 crore, so one percentage point is Rs 5 crore.

Take the equity line. The policy weight is 60.0 per cent, or Rs 300 crore, with the lower edge at Rs 250 crore and the upper at Rs 350 crore, so there are ten points of room in each direction and each ten points is Rs 50 crore.

Rs 50 crore each way is the mandate's own unit, and it answers how much may be moved between sleeves before an edge is touched. The mandate's own unit is the right figure when the committee is authorising a shift and the wrong one when the committee asks whether an edge could be reached without anybody acting.

For that second question the unit is a market move. A rise in equity lifts the portfolio total as well as the sleeve, so the move is not symmetric in the two directions. Holding the other Rs 200 crore still, a fall of 33.3 per cent in the equity sleeve reaches the lower edge and a rise of 55.6 per cent reaches the upper one. The self-lifting arithmetic is worked through under the investment mandate, and the memo carries its result.

Now the single holding cap. The largest position stands at 4.6 per cent, or Rs 23 crore, against a cap at 5 per cent, or Rs 25 crore. In the mandate's unit the room is 0.4 points and Rs 2 crore; in the market's unit that holding would have to rise 9.15 per cent on its own before the cap is reached. The memo carries the result of a derivation rather than repeating it.

The third line behaves differently. The mandate permits no unlisted holdings and the portfolio has none, so the distance is exactly zero and stays zero: no price move widens or narrows it. An eligibility lineA limit asking whether something may be held at all rather than how much of it may be held, so it is answered yes or no. is the only line in the check whose distance a market cannot touch, and the memo says so rather than leaving a reader to infer it from a zero.

The fourth line teaches the most. The mandate requires a minimum credit standing on the fixed income sleeve, Rs 150 crore or 30.0 per cent of the portfolio, stated as a policy rather than a rating symbol. No price resolves this line, so the check cannot report a distance for it. The memo records the test applied and the date it was last run.

Here the invented record runs out. The record locks the sleeve, the weight and the fact that the standing is written as a policy. Which test was applied, and when, it does not lock, so the memo field stays empty and says so.

A memo is read by people who cannot check every line themselves, so the fields left visibly empty are the only ones they can be sure were not quietly guessed. An honestly empty field is stronger evidence of care than a plausible number in the same slot.

Read the last row of that table. The nearest line to binding is the single holding cap, and nothing in a column of ticks would have said so.

Step four: why is a headline result not yet a finding?

The fourth section reports the year and does two things in one block: it states the result, then it splits it. Not in an appendix, not in a later paper. Adjacent, with the parts shown to add back to the headline.

Take the record over one stated twelve month period. The portfolio returned 14.2 per cent. The composite benchmark, being 60 per cent a broad equity index and 40 per cent a broad bond index, returned 12.6 per cent, so the excess returnThe difference between what a portfolio earned and what its stated comparison earned over the same period. is 1.6 percentage points gross of fees. The risk-free rate stood at 6.5 per cent throughout. Without that rate no risk-adjusted figure compares with anything, so it is stated beside every one of them.

Before that 1.6 becomes a finding the memo has to answer one question. BetaHow much a portfolio has tended to move when its comparison moved, taken here as a figure from the record. against that benchmark was 1.08 for the same period. Beta as a measure, and how it is estimated, is taken up separately under beta estimation. The memo uses it, and the split is the use.

Here is the arithmetic the memo prints. The benchmark returned 12.6 per cent against a risk-free rate of 6.5, so it earned 6.1 points above that rate. A portfolio carrying a beta of 1.08 and doing nothing else would be expected to return 6.5 plus 1.08 times 6.1, or 13.088 per cent. The portfolio returned 14.2, so the remainder is 1.112 percentage points.

The other part follows at once: with a remainder of 1.112 points against a gross headline of 1.6, the part explained by carrying more of the same market is 0.488 points. The same figure comes straight from the beta. A beta of 1.08 rather than 1.00 adds 0.08 of the benchmark's 6.1 point excess, and 0.08 times 6.1 is 0.488. The two parts add back to the headline exactly, and a split that does not add back has used a rounded beta or a different window somewhere in the working.

The memo states the finding in its own words. Of the 1.6 points of gross outperformance, 0.488 points, being 30.5 per cent of the headline, came from carrying more of the same market. The other 1.112 points is what the extra exposure does not explain, and the record names that remainder alphaWhat is left of a return once the return expected from the exposure carried has been subtracted. and rounds it to 1.11.

The memo's refusals matter as much. The paper has not said the manager was skilful and it has not said the mandate is being run well. Printing two numbers that add to a third leaves the conclusion to the people authorised to draw it.

At a beta of 1.2623 the remainder falls to zero and exposure explains the whole gross excess. Below 1.00 the exposure part turns negative and the remainder has to exceed the headline. The headline is a measured fact and the split is only an accounting of it, so the bar never changes length.

What does the same headline look like from two different years?

The split is compulsory rather than analytical decoration, and here is why. Two portfolios can report the same 1.6 points of gross excess for a stated year and be in completely different situations, and a memo reporting only the headline hands the committee no way to tell them apart.

The lower row is constructed for the comparison and belongs to no record. The headline of that constructed row is identical to the real one. A committee looking only at headlines would record the same sentence in both minutes, and in one case that sentence would describe something the mandate never authorised.

Step five: what does no return figure contain?

The fifth section carries the facts about the year that a return line structurally cannot hold. Those facts are not caveats and they do not belong in a footnote. A return is a single ratio of end to beginning, and whole categories of thing that happened inside the year cannot be expressed by such a ratio.

Two of them are in the record. TurnoverThe share of a portfolio replaced by trading over a period: how much changed hands, not how the value moved. ran at 34 per cent over the stated twelve months, and the worst drawdownThe largest fall from a high point to the low point that followed it, inside a stated window. inside the same window was a fall of 9.7 per cent before recovery, against 8.1 per cent for the composite benchmark.

Turnover of 34 per cent on its own is a number nobody in the room can use. Turnover becomes usable the moment it is scaled against something the committee already has a feel for. On a Rs 500 crore portfolio, 34 per cent is Rs 170 crore replaced across the year, and the whole cash sleeve is Rs 50 crore, giving 3.4 times the cash sleeve passing through in twelve months.

A figure with nothing beside it is a fact nobody in the room can use, and choosing what to put beside it is the writer's job rather than the reader's. What that trading cost in rupees is not in the record, so the field stays empty and marked, and the committee knows there is a cost it has not been shown.

A drawdown figure is meaningless without its window. The window is quoted every time the figure appears. A paper reporting a worst fall without saying inside what period has reported a number that cannot be checked against anything.

Which of these figures is gross, and which is net?

One more line belongs in the reporting block and it is the one most often left out. The same record can produce two opposite readings of one year, so every excess figure has a basis and the memo has to say which.

The 1.6 percentage points above is gross of fees. The invented commercial terms of this mandate, taken up later in this subject area rather than here, total Rs 9.40 crore for the stated year, being 1.88 per cent of the Rs 500 crore portfolio. Subtract that and 14.2 per cent becomes 12.32 per cent against a benchmark of 12.6, a shortfall of 0.28 points net. The same year on the same record is 1.6 points ahead gross and 0.28 points behind net, and a memo printing the figure without the word has printed something ambiguous between two opposite conclusions.

The memo rule follows from that: every excess figure in a committee paper carries its basis, its period and the comparison it was struck against, in the same sentence as the figure.

Step six: how does the memo state what would change the answer?

The sixth section is where most memos collapse into hedging. The phrase it depends is not a sentence in a memo. Neither is subject to market conditions, nor a range of outcomes is possible, nor any softening phrase whose only function is to make the writer unfalsifiable.

In their place goes a short list of observablesFigures somebody could go and measure, as distinct from conditions describable only in general terms.: the two or three figures the answer turns on, each with the direction it cuts. A named observable is the opposite of hedging. Naming one is a commitment: if the figure comes back at that value, the reading changes.

For this annual review the two observables are the ones the split leaves open. The first is the beta measured over a different window. The 1.08 belongs to one stated twelve month period, and another window would move the 0.488 and the 1.112 in opposite directions. The second is how the 1.112 points divides between decisions the mandate authorised and decisions it did not.

And the record is silent on both, so the memo says so rather than estimating them.

Step seven: who decides what the memo leaves out?

The last section names what has been excluded and who excluded it. Naming the exclusions feels like an odd thing to put in a paper that took a week to write, and it is the section that separates a memo from an argument. A memo is a selection: somebody chose which constraint lines to check, which figures to split, which comparison to put beside the turnover and which observables to name, and every one of those choices removed something. The selection is itself a decision, and a committee that cannot see what was excluded is reviewing an argument rather than a position.

Think of a household budget one person prepares for the rest to discuss. If the preparer quietly leaves out the annual insurance premium, none of the numbers shown is wrong. The sheet is simply not the household's budget any more but the preparer's view of it, and nobody in the room can tell the difference by looking.

Notice the fourth row. The memo carries no recommendation about what to do with the mandate, and says so explicitly rather than leaving the absence to be noticed. Faiz Ahmad Ansari writes the paper; Rukmini Deshpande and the committee take the decision.

What does the whole memo look like assembled?

Put the seven sections together and this is the paper. The paper is short, and that is the point: every section either carries a figure a reader can check or says plainly that no figure exists.

What must never be a step in this procedure?

Four things arrive looking like the next step and belong nowhere in the sequence. Each one is added by a writer trying to be helpful, so each one is worth naming.

The fourth is the one that gets made most often and looks least like a fault. Faiz Ahmad Ansari can mention the missing trading cost aloud when the item is called, and everybody in the room will nod. Six months later the minutes record a decision taken on a paper that never carried it. The minutes are written against the paper and not against the meeting, so anything the committee is expected to weigh has to be in the paper.