Research Output: The Note, the Report and the Update

Equity research comes out in three shapes. A note answers one question on one occasion. A report states the whole view with its assumptions, evidence, comparison set and risks. An update records what a new disclosure did to that view, including when it did nothing. The annual report and the investor presentation are inputs to all three rather than outputs of any.

Taking on a company commits an analyst to the obligations set out under the coverage commitment, and the heaviest of them is the duty to publish after every result whether or not there is anything new to say. The duty to publish on the calendar is the reason one of the three shapes below exists at all. An update is not something a researcher produces when moved to. An update is the required shape that a standing commitment forces out of an analyst four times a year.

How a research note is drafted, how a results release is taken apart and how an assumption set is built are settled elsewhere. The question left over is which container each of those goes into, and the container is not a formatting decision at the end. A long document silently promises the reader a full assumption set. A short one silently assumes the reader already holds it. Choosing the shape is therefore the first honest act in publishing. The wrong container makes a promise that will not be kept, before a single sentence has been written.

What is a research note, and what can it never do?

A note is short, it answers one question, and it exists because something happened. A supplier raised prices. A plant went down. A competitor said something on a call. The note takes that one occasion, asks the single question it opened, answers it, and stops.

The interesting property is what a note leaves out on purpose. A note does not restate why the view is held. A note assumes the view, the way a message to somebody spoken to every day assumes everything already said between them. The assumed view is what keeps a note to a single sheet, and the assumed view is also the whole of its limitation. The reasoning that would establish a view is not in a note. A note can never establish one, and no quantity of notes stacked together turns into that reasoning either.

Think of a shopkeeper who keeps a running book. The person who has been keeping that book all year knows what a change is a change from, so a single line reading that the tin supplier raised the drum rate this week is genuinely useful to them. Hand that one line to a stranger and it is close to meaningless. The stranger has no run of earlier lines to set the rise against, and cannot tell whether it is ordinary or alarming. A research note is that single line, written by somebody who is keeping the book, for readers who are reading along.

What is a Research Report, and what has to be in it?

A report is the other extreme, and it is defined by the reader it is written for rather than by its length. The reader of a report is a stranger. Somebody who has never read a word the analyst wrote, does not know the view, does not know the assumptions and has no run of earlier lines to set anything against.

The stranger sets the only test that matters: a document is a report when a reader who has read nothing else by the analyst can reconstruct the reasoning from it alone, and a document that fails that test is a note wearing a longer document's clothes. Length has never been the test. A twenty sheet document that restates a conclusion in six different ways and never puts a number on an assumption is a note with padding, and calling it a report does not change what the reader can do with it.

Six things have to be in it, and they are not a house style. Each one is there because a stranger cannot proceed without it. The view itself, stated plainly. The assumption set, with actual numbers on each assumption rather than adjectives. The evidence behind each of those assumptions, so the reader can disagree with a specific one rather than with the whole thing. The comparison set, with the names left out and the reason they were left out. The risks. And the named observations that would break the view, written down before they could be observed.

What is a Research Update, and what does it say when nothing moved?

The update is the shape the coverage commitment forces. A disclosure arrives, the calendar comes round, and the reader is owed a pass over the view whether or not the analyst feels there is anything to add. Two statements are required and neither is optional. Which assumptions were checked against the new disclosure. And what happened to each one.

Then comes the part that separates a run of updates worth keeping from a run of updates worth nothing, and it is a habit rather than an insight. A changed assumption is recorded as a dated move from one stated number to another stated number, never as a new number quietly sitting where the old one used to be. Write that the assumption for earnings before interest, tax, depreciation and amortisation, the reported EBITDAEarnings before interest, tax, depreciation and amortisation. The profit line taken before financing costs and before the cost of using long lived assets is charged against it. margin, moved from 16.04 per cent to 18.47 per cent over one year on the year three result, and give the date. Do not write 18.47 per cent where 16.04 per cent used to be and move on.

The dated move matters far more than its size suggests. The whole of the later examination of a call depends on it. A year from now somebody, quite possibly the analyst who wrote it, will want to know when a view changed and what changed it. If every version simply overwrote the last one, there is nothing to read backwards. The trail is the record, and the record is the only thing that lets anybody tell a research error from an unlucky outcome later.

And the update nobody wants to write is the one where nothing moved. A checked and unmoved assumption is a finding about the world, not an absence of work. An update announcing no change is therefore a complete update rather than a failure to produce one. A household that opens the electricity bill every month and finds it steady has not wasted the opening. The household now knows the bill is steady, and steadiness is exactly what it would need to know before deciding anything.

Annual Report vs Investor Presentation: which document was actually being read?

The two documents in this heading are not research output at all. Both are the company's own, both are inputs to all three shapes above, and mistaking one of them for the other does more model damage than any other confusion in research.

The two documents differ on four axes, and only the last of the four is usually noticed. Who wrote it and for whom. The obligation that produced it. The assuranceThe independent work an auditor performs before signing. Such work lets a reader treat a set of figures as checked by somebody outside the company rather than simply asserted by it. standing behind the figures in it. And how much selection went into deciding what appears at all.

An annual report is produced because a listed company must produce one, on a fixed rhythm, in a shape it does not choose, and its core financial statements carry an audit behind them. An investor presentation is produced because the company has something to communicate, on a rhythm it chooses, in a shape it designs, and nothing in it is audited as a presentation. Both are filed at the exchanges. Only one of them was written under an obligation to include the parts nobody wants to look at.

The presentation is not a dishonest document and is not trying to be one: it is a document with an audience, and selection is precisely what it is for. Nobody expects a shop to lead its window display with the slow moving stock. The error never belongs to the company that made the selection. The error belongs to the reader who forgets that somebody chose the contents, and then treats a chosen figure as though it arrived by itself.

Which document may be relied on, for which figure?

Cautions get nodded at and rules get followed. Reliance is therefore best stated as a routing rule rather than as a caution. Three kinds of figure, three destinations, and the routing does not depend on how anybody feels about the company.

A figure that has to tie goes back to the audited statements. Revenue, the cost of materials, the profit lines. A figure that explains why something moved goes to the notes to the accountsThe numbered sheets behind the statements, where each headline total is broken out into the individual items that make it up.. The items that did not make the face of the statements live there. A figure describing management's own view of performance can be taken from the presentation, and it is evidence of exactly one thing: what management chose to emphasise. A presentation figure is real evidence about the emphasis and no evidence at all about the business. The third routing is the one that costs money.

One year at Sarvani Coatings, three EBITDA figures, three documents

Sarvani Coatings Limited, an invented listed maker of decorative paints and industrial coatings, closes its books on 31 March. Take its year three, ended on that date, and watch a single completed year produce three defensible EBITDA figures and then three different documents.

Revenue for the year is Rs 2,415 crore. The presentation leads with an adjusted EBITDA of Rs 452 crore, a margin of 18.72 per cent, arrived at by adding back a restructuring chargeA cost booked for reorganising part of a business, such as closing a line or reshaping a workforce. The company does not expect a cost of that kind in the same shape next year. of Rs 6 crore. The face of the annual report carries the reported EBITDA of Rs 446 crore, a margin of 18.47 per cent. And the notes carry two things: that same Rs 6 crore charge, and a provision write-backA reversal of money set aside in an earlier year for a cost that then turned out smaller. The reversal lands as a reduction in this year's expenses. of Rs 4 crore that went the other way and was left in place.

Adding the charge back and removing the write-back applies the same test in both directions, and gives a third figure of Rs 448 crore at 18.55 per cent. Three figures, one year, one denominator, all correct arithmetic.

Work the gaps properly. Working them properly means working the rupee absolutes rather than subtracting one rounded rate from another. All three margins sit over the identical denominator of Rs 24,15,00,00,000/-, so every margin gap is simply the rupee gap over that same denominator. The step from reported to the two way figure is Rs 2 crore, or 0.08 of a margin point. The step from the two way figure to what management presents is Rs 4 crore, or 0.17 of a point. The second span is exactly twice the first because Rs 4 crore is exactly twice Rs 2 crore, and that agreement is forced arithmetic rather than a check that happened to pass. Together they make the 0.25 of a point separating the reported figure from the presented one, and because both are printed to two places, subtracting the printed rates gives the same answer the unrounded arithmetic gives.

Size the Rs 4 crore honestly and give it its denominator every time. Against the Rs 452 crore management presents it is 0.88 per cent. Against the Rs 448 crore two way figure it is 0.89 per cent. Both round to about 0.9 per cent of EBITDA at one decimal place, and both are correct answers to slightly different questions. The denominator has to travel with the number.

Now sort the output, using nothing that is not already above. The update after the results states that the margin assumption was checked against the disclosure, records all three EBITDA figures and shows the reconciliationA line by line bridge showing exactly which items were added or removed to get from one figure to another figure. between them. The note, written the same week, takes the single question of whether a write-back of that kind is likely to recur, answers it from the notes, and assumes the reader already holds the view. The report, written at initiationThe first full report a researcher publishes on a company after starting to follow it, written for a reader who holds nothing yet., carries all three figures, the reconciliation, the reason the two way test is the one run here, and the named observation that would change it.

The same disclosure produced three documents of quite different lengths, and not one of them needed a figure that was not already in the annual report and its notes.

What does every research output carry, however short it is?

Four things, and they are the same four whether the document runs to four lines or forty printed sheets. The date. The assumption set, or an explicit pointer to where it lives. The named observation that would change the view. And the disclosures the writer is required to make.

The first two do the heavy lifting and are the two most often left off. On the day of writing they feel like paperwork rather than content. A document missing its date and its assumption pointer is unusable within a quarter, however good it was on the day it was written. Not wrong. Unusable, and unusable is worse. A wrong document can at least be argued with. Nobody can tell what was known when an undated note was written, or what view it was sitting inside. An undated note asserting that the write-back is unlikely to recur cannot be argued with by anybody.

Test it against a household letter. A note left on a kitchen table saying the milk was paid for is complete if it is dated and says which milk. Undated and unattributed, it is a sentence that used to mean something to somebody.

One line on the order the parts of a document sit in, and it is an attribution rather than a lesson. The conclusion goes first and the support follows. Conclusion before support is the Pyramid Principle and belongs to Barbara Minto, 1978. How a document is actually drafted is covered separately under research writing.