Peer Group vs Coverage Universe: What Fixes Each Set

A peer group and a coverage universe are different sets built for different reasons. A peer group is fixed by one question and changes when the question changes. A coverage universe is fixed by a decision to keep publishing and changes only when that decision does. A company can sit in one, in both or in neither, and the company that would reveal most is often in neither.

Two ideas stand behind this one. A coverage commitment obliges somebody to keep publishing on a company week after week, whether or not the company gives them anything to write about, and that obligation is set out under coverage commitments. How a comparison set gets chosen honestly, and how it quietly rots when it is not, is set out under comparison set selection. Both are assumed below.

A third job is left over, and those two ideas cannot do it separately. Two lists of company names sit in the same research file. The two lists overlap, and they are often printed in the same font in the same document. And they were assembled by two processes with almost nothing in common. Nearly every misleading relative number in published research comes from one of those two lists quietly doing the other one's work.

What exactly is a peer group?

A peer group is the set of companies placed beside a subject company in order to answer one stated question. One stated question is the whole of it. Not to answer questions in general. One question, named, written down before the set is built.

If the question is whether the gross margin gain at Sarvani Coatings Limited, an invented maker of decorative paints and industrial coatings, was shared across the field, the set required holds companies whose margins move for the same reasons. If the question is instead whether its decorative volumes are keeping pace, the set required holds companies selling decorative paint through the same kind of shelf. The two sets are not the same, and there is no reason on earth they should be. A peer group is not a description of who a company competes with; it is an instrument built to answer one question, and swapping the question changes the instrument.

Think of a household comparing its electricity bill. If the question is whether the bill is high for the season, the comparison is the same household last August. If the question is whether the new refrigerator is thirsty, the comparison is the neighbour with the same size home and an older machine. Same household, two questions, two completely different comparisons, and nobody finds that strange until the word "peer" gets attached to it.

What exactly is a coverage universe?

A coverage universe is the set of companies an analyst has undertaken to keep a current published view on. A coverage universe is not an opinion about which companies are interesting. A coverage universe is a standing commitment, and the commitment is the defining feature: every name on that list generates work on a schedule the company sets rather than the analyst.

Two things fix membership, and neither of them is about similarity. The first is capacityThe plain arithmetic of how many companies one analyst can keep genuinely current, given the hours there actually are in a working year.. There are only so many results seasons, transcripts, filings and follow up calls one person can absorb before the currency of the view starts to slip, and a list that ignores that arithmetic is a list of stale views. The second is a decision, taken by the deskThe team a research house organises around a group of sectors. It is the unit whose hiring decides which companies get followed at all. rather than by the analyst alone, that this particular company is worth carrying: that somebody wants to read about it, that the effort is justified, that it fits the kind of company the house is set up to follow.

Neither capacity nor a decision to keep publishing has anything whatever to do with whether two companies are alike, and the two lists drift apart for exactly that reason. The household version sits beside it. An electricity comparison set is chosen by the question. The list of bills actually opened every month is chosen by which accounts stand in the household's name. The two lists overlap and neither one contains the other.

How often does each one change, and what does a change cost?

Here the two sets stop merely differing and start behaving like different kinds of object.

A comparison set changes whenever the question changes, and inside one report the question changes constantly. Meghna Iyer, whose working file this sequence keeps returning to, writes a report on Sarvani Coatings Limited that asks three things in turn. Was the gross margin gain shared with the field? Are decorative volumes keeping pace? Is the shift towards industrial doing the work? Each of those needs a different set beside it, and she builds three, drawing on two names. Rebuilding cost her a line in the working file each time, and that line is the entire cost.

A coverage list does not work like that at all. Adding a name is an initiationThe published act of starting to carry a company on a coverage list, announced to readers rather than done quietly inside the desk.. An initiation is a document with a date on it that says: from now on, expect a current view here. Removing a name is a discontinuationThe published act of dropping a company from a coverage list, so a reader learns that the last view on it is no longer being kept up.. A discontinuation says the opposite, and it has to say it out loud. A reader who is not told will keep treating the last note as current for years. A comparison set is meant to be rebuilt and a coverage list is meant to hold still, and every change to the second one is an act somebody has to publish.

Notice what this does to the sizes. A short comparison set is a sign of a sharp question. A short coverage list is a sign of an honest promise. Both are short for reasons, and the reasons are unrelated.

What does each one actually oblige an analyst to do?

A comparison set obliges one thing and it is finite: publish the membership and the exclusions beside the comparison they were built for, so a reader can see what the number was measured against. Done once, in that document, the obligation is discharged. Nobody comes back next quarter asking for Set B to be refreshed.

A coverage list obliges something with no end date. Every name on it produces work whenever the company acts, and, more awkwardly, when it does not. Silence from an analyst reads as either agreement or absence, and a reader cannot tell which. So a quiet quarter still needs a line saying the view is unchanged and why.

One obligation is discharged by publishing a list once. The other renews itself every reporting period for as long as the name stays on. A coverage list is short and a research file holds many comparison sets for exactly that reason. A shopkeeper can price against whichever three stalls are relevant to today's question and owes nobody an explanation tomorrow. The moment she promises a customer a weekly quote on four specific items, those four items are hers until she says otherwise, out loud.

How do four invented names sort against both lists?

Work it on the record rather than in the abstract. Meghna Iyer's coverage list holds four names. Sarvani Coatings itself is one of them, and an analyst covers the company she writes about. Nandivarman Paints Limited, the biggest of the four and selling next to nothing outside the decorative trade. Then Kesaria Surface Solutions Limited, the smallest, weighted towards industrial work, and moving through a cycle differently for that reason. Thottam Chemicals Limited is fourth, and it makes the resins and additives the other three buy, so it stands one step up the chainA company that sells into a field rather than competing inside it, such as a maker of the raw inputs that every company in the field has to buy. from every one of them.

Her comparison set is built for one question: was the gross margin gain at Sarvani Coatings shared with the field? The set holds Nandivarman Paints and Kesaria Surface Solutions, and nothing else. Here is the ladder that poses the question, taken straight from the record and printed to two places so that the subtractions close.

Two notes before the sorting, both of which matter more than they look. First, the record prints these as 43.0 and 46.0 per cent and calls the two year gain 3.0 points, and built work elsewhere carries those printed figures, so both are shown. The two place figures above are what the rupees actually give, and they are the ones that survive a reader with a calculator: gross profit of Rs 792 crore on revenue of Rs 1,840 crore is 43.04 per cent, Rs 1,111 crore on Rs 2,415 crore is 46.00 per cent, and 46.00 less 43.04 is 2.96, exactly as stated. Second, the cost of materials share falls by the same 2.96 points over the same two years, from 56.96 to 54.00. The matching fall is forced arithmetic and not a second finding: gross profit and cost of materials are the two halves of revenue, so their shares must move by equal amounts in opposite directions, and anybody reporting both as if they confirmed each other has counted one fact twice.

Now read the four combinations off the pair of lists. Nandivarman Paints and Kesaria Surface Solutions sit in both. The ordinary case needs no defending. Sarvani Coatings itself sits on the coverage list and is not in its own comparison set. A subject company is what gets compared rather than something compared against. Saying so is worth the line. Files do get built where the subject appears inside its own peer table.

Thottam Chemicals is on the coverage list and outside the comparison set, and this is where the distinction earns its keep. A supplier is not a comparable. Thottam Chemicals sells to the makers rather than alongside them, its customers are the companies in the field, and its costs move for reasons one link removed. Put it in a margin table beside Sarvani Coatings and the table means nothing. And yet its own margin over the same two years would do more than either comparable to separate a pricing environment across the whole field from Sarvani Coatings pricing on its own account, so the single most informative name in this analysis is deliberately outside the comparison set.

The fourth combination is the one people find hardest. Suppose a large privately held maker competes with Sarvani Coatings directly in decorative paints, on the same shelves, at the same price points. The privately held maker passes every comparabilityWhether two companies are alike on the one axis the question turns on. Two makers can be comparable for a margin question and not comparable for a growth one. test that can be put to it. And it cannot enter a coverage list at all. There is no published view to maintain, no periodic disclosure to read and nothing to be current about. The maker belongs in the comparison set in principle and is unusable in practice.

The honest handling is not to pretend it away. The maker goes into the exclusion record with its reason categoryA short labelled ground, picked from a fixed list, for why a candidate was left out, so exclusions can be counted later instead of argued one at a time. written as no public disclosure. One recorded line does real work: a reader who later wonders why the strongest competitor is missing gets an answer, and the line records that the comparison set is narrower than the field rather than equal to it. The rest of the field, nearly all of it, sits in neither list and always did.

Which set gets published, and to whom?

Both get published, and a reader who assumes one is internal has misread what each list is for.

A relative number cannot be checked without knowing its denominator, so the comparison set is published with the comparison, inside the document that used it. A margin sixty basis points above the set is a fact about the set. Withhold the membership and what has been published is a number nobody can audit. A number nobody can audit is a different thing from a number nobody agrees with.

The coverage list is published as a standing list, separate from any one document, and it answers a question no individual note can. A reader who finds a note from fourteen months ago needs to know whether a current view still exists behind it. The list answers that, and a discontinuation answers it in the other direction.

The two lists are published for different reasons and settle different reader questions: one lets a reader audit a number, the other lets a reader learn whether a view is still being kept up, and neither one substitutes for the other.

What breaks when one list does both jobs?

Two failures are possible and they are not equally dangerous.

Take the second way first. The second way is the easier one. Using the comparison set as the coverage list is a promise to keep publishing on companies picked for a single question. Nobody keeps that promise, and the reason is visible in the figure above: three questions produced three sets inside one report, so this route commits an analyst to standing behind whatever the last question happened to need. The promise collapses within a quarter, loudly, and gets fixed.

The other direction is where the real damage lives. Nothing collapses at all. Use the coverage list as the comparison set, and a comparison gets made against whoever the desk happens to follow. Every member is genuinely in the same field. The table looks entirely reasonable. The table is labelled peers. And what actually decided its membership was a hiring decision taken by somebody who was not thinking about margins.

The first error announces itself and the second never does. A set assembled by a desk always looks plausible in a document, and that is why the second error is far the more common of the two.

What gate keeps the two lists apart?

The fix is not vigilance. A two gate test applied to every candidate, with the outcome written down whichever way it goes, takes about a minute a name.

The questions the two gates do not ask matter just as much. Neither one asks whether the company is already followed. Neither asks whether following it would be convenient. The omission is the whole design: the moment a coverage question enters a comparability test, the failure drawn above has been let in through the side door.