Strategic, Custom and Peer Benchmarks: How They Differ

A benchmark is the portfolio a mandate would have held if nobody took a single decision. A strategic one is a fixed public composite. A custom one is built to match one mandate's own constraints. A peer one is what other portfolios did, and it is neither investable nor stable. The Anantara Multi-Asset Portfolio holds cash its composite does not contain.

Performance attribution showed that every attribution effect is measured against a benchmark and quietly inherits whatever that benchmark happens to be. A question then sits on the table, and almost nobody asks it out loud: where did the benchmark come from, and who decided it was the right one? The three kinds differ on two axes that actually matter, who builds each one and what each one rewards, and the mismatch between a mandate and its benchmark has a price in rupees.

The running example is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore run by Faiz Ahmad Ansari for an invented charitable endowment whose investment committee is chaired by Rukmini Deshpande. Its policy weights are equity 60.0 per cent at Rs 300 crore, fixed income 30.0 per cent at Rs 150 crore and cash 10.0 per cent at Rs 50 crore, summing to Rs 500 crore exactly. Its benchmarkThe alternative portfolio a result is measured against. It is stated in advance so that whatever the actual portfolio did can be read as a difference from something. is a composite benchmarkA benchmark made by blending two or more market measures in stated proportions, rather than using a single one on its own. of 60 per cent a broad equity measure and 40 per cent a broad bond measure. Every weight and return below belongs to one stated twelve month period.

Notice what has already gone wrong, before anybody has chosen a single holding. The mandate holds cash. The composite holds none. Ten points of the Anantara portfolio sit in a bucket the yardstick does not contain, and that gap was written into the pairing rather than chosen by anyone.

What job does a benchmark actually do?

The definition worth memorising is not the one most people carry. A benchmark is the alternative that needed no decisions. The alternative is what the mandate would have held if Faiz Ahmad Ansari had gone away for the year, taken no view on anything, and simply let the stated shape sit there. Everything he actually did then becomes a difference from that shape, and a difference is readable in a way a raw number never is.

Consider how a school report is judged. A mark of 68 says almost nothing on its own. Set 68 against the paper the whole class sat and the mark starts to say something. A brutal paper can leave everyone under 50, so the class average is neither a target the student was aiming for nor a standard of adequacy. The class average is simply the quantity subtracted. What is left is about this student rather than about the paper.

The consequence almost nobody states out loud is the one that governs everything else in this area. A benchmark is not a target and not a standard of adequacy, it is a subtraction, and every measure that follows in performance work is built out of that subtraction rather than out of the raw return. Excess return is a subtraction. Active risk is the volatility of a series of subtractions. An attribution effect is a subtraction sliced by bucket. Change the thing being subtracted and every one of those numbers changes, without a single holding in the Anantara Multi-Asset Portfolio moving by one rupee.

The choice of benchmark is therefore not administrative housekeeping. The choice reprices the entire record of the year, and it is taken once, usually quickly, and often by whoever had a spreadsheet open.

What is a strategic benchmark, and what does it buy?

A strategic benchmarkA fixed, published blend of broad market measures in stated proportions, settled when the mandate is written and then left alone for the life of it. is a fixed, published composite of broad market measures in stated proportions, settled when the mandate is written and then left alone. The Anantara portfolio's 60 and 40 composite is exactly this shape. Two proportions, two broad measures, written down once.

Its first strength is that it cannot be changed after the fact. Once 60 and 40 is in the document, nobody can look at a disappointing year and discover that 50 and 50 would have been the fairer comparison all along. Fixing the proportions sounds like a small procedural point. Adjusting the yardstick after seeing the result is not a rare temptation but the normal human response to a bad number, and resisting it is the whole reason the form exists.

Its second strength is that anybody can check it. A trustee with no access to the holdings, working from published proportions and published market measures, can reproduce the benchmark return independently and arrive at the same figure the manager reported. A strategic benchmark buys a number that survives being checked by somebody with no reason to trust the manager. Most measures in this business do not have that property.

Its weakness is single and serious. A strategic benchmark is built from broad market shapes, so it will not fit a mandate carrying unusual constraints. The Anantara mandate must hold cash. No broad composite of an equity measure and a bond measure contains a cash bucket, so the fit fails at exactly the point the mandate is most specific about itself.

What is a custom benchmark, and where is it exposed?

A custom benchmarkA benchmark assembled for one mandate, so that its buckets and proportions mirror what that mandate is actually permitted to hold. is assembled for one mandate. Its buckets mirror the mandate's own opportunity setThe complete set of things a mandate is permitted to hold, after every restriction on kind, credit standing, currency and listing has been applied., meaning the full set of things that mandate is actually allowed to hold once every restriction has been applied. For the Anantara mandate that would mean a composite with three buckets rather than two, at 60, 30 and 10. The cash the mandate is required to carry would then have a counterpart in the composite it is measured against.

The strength is exactly what the strategic form lacks. The custom form measures the manager against what the manager was allowed to do. If the mandate forbids unlisted holdings, imposes a minimum credit standing on the fixed income sleeve and requires a cash reserve, then a yardstick containing unlisted names, weaker credit and no cash is asking about a portfolio that was never available to be built. Nobody learns anything from a comparison with a portfolio nobody could have held.

The exposure is the mirror image of that strength, and it is uncomfortable enough that people skip past it. A custom benchmark is built by somebody, and if that somebody is close to the person being measured, it can be built to be easy without anyone ever writing down a dishonest sentence. A credit floor nudged a shade weaker in the benchmark than in the mandate. The more sluggish of two defensible measures chosen for one bucket. None of those choices is a lie. Each is arguable on its own. The result is a yardstick nobody outside the room can reproduce, quietly set a little lower than the thing it measures.

Strategic Benchmark vs Custom Benchmark: which trade is actually being made?

Set the two forms side by side on five questions and the shape of the choice becomes hard to miss. Who sets it. Who can change it. Can an outsider reproduce it. Which behaviour does each one reward. And where does each one fail.

Read the last row twice. The trade is the finding. The strategic form buys verifiability at the cost of fit, the custom form buys fit at the cost of verifiability, and there is no third form on the shelf that quietly collects both. Anyone claiming that an arrangement has solved this has either written the construction rules down and dated them, or has not thought about it. Writing the rules down and dating them converts a custom benchmark into something an outsider can reproduce, and that is the honest way through.

The everyday version is a driving test. A published test route anybody can inspect is verifiable and may not resemble the roads actually driven daily. A route built around one driver's own commute fits perfectly and can be laid out to avoid every junction that driver finds hard. The way out is not a cleverer route. The way out is publishing the route in advance. The fit is then real and the choosing is on the record.

Why is a peer benchmark the weakest of the three?

A peer benchmarkA comparison built from what other broadly similar portfolios reported over the same window, usually as a median or a quartile position rather than a single portfolio. is what comparable portfolios did over the same window, usually reported as a median or a quartile position. A peer benchmark is intuitive, a committee reaches for it first, and it carries three separate faults. The word separate is doing real work in that sentence.

The first fault: a peer median cannot be held. The median is computed at the end from a set that was not knowable at the beginning. No portfolio available at the start of the year would have delivered it, and a median nobody could have chosen fails the definition set out above.

The second fault: the population moves. Each year's median is computed from whichever portfolios reported that year, and the reporting group differs from one year to the next. Comparing two medians treats a moving population as a fixed one. The population was never fixed.

The third fault is survivorshipThe distortion that appears when a comparison is built only from the participants still present at the end, because the ones that stopped reporting are missing from it.. Portfolios that closed are not in the set at all. The closed portfolios were there at the start of the year and absent from the calculation at the end, so the survivors flatter the median without anybody intending it.

The three faults are separate faults and not one, and repairing any single one leaves the other two standing exactly as they were. Freeze the population and the median is still unheldable and still missing the closures. Reinstate the closures and the median is still unheldable and still recomputed on a shifting group. A peer comparison is useful as background colour and it is not a yardstick, and the difference between those two roles is the whole of the matter.

What has to be true before a benchmark can be used at all?

Four questions, and a usable benchmark needs a yes to all four. The four are worth running out loud in the room rather than assuming.

One: is it specified in advanceWritten down and agreed before the period it will be used to measure, so that it cannot be selected after the result is known.? Written down before the period it measures, so nobody can choose it after seeing the answer. Two: is it investableSomething a holder could genuinely have held for the whole period, rather than a figure that can only be worked out afterwards.? Somebody could actually have held it for the whole period. Three: is it measurable from published data, and can the return be computed by someone who is not the manager. Four: does it match the constraints the mandate imposes?

Run the Anantara pairing through the four. Specified in advance: yes, the 60 and 40 composite sits in the mandate document. Investable: yes, a holder could have held that shape all year. Measurable: yes, from published market measures. Match: no. The mandate requires cash and the composite contains none.

A benchmark that fails only the fourth question keeps producing numbers every single month, and a failure that keeps producing numbers can survive for years without anybody noticing it. A benchmark failing the first is obviously broken. Nothing can be compared against until somebody picks one. Failing the second or third makes the arithmetic impossible or unverifiable, and impossibility gets noticed. Failing the fourth breaks nothing mechanical at all. The spreadsheet fills in. The report prints. The excess return has a sign and a magnitude. Only the meaning is wrong, and meaning does not throw an error.

Each of the four failures breaks something different, and only one of them breaks nothing visible.

How Benchmarks Shape Portfolio Decisions: what changes once the yardstick exists?

Everything, and this is the part of the subject that reads as slightly uncomfortable once it is seen. The moment a benchmark exists, every weight in the portfolio stops being a weight and becomes a difference from a weight. Sixty per cent in equity is no longer sixty per cent in equity. The composite also holds sixty, so the active weight is zero. The Anantara Multi-Asset Portfolio's Rs 300 crore equity sleeve, the single largest thing in the mandate, is invisible to every active measure.

And the reverse holds. The things not held become positions. If the composite carries a large weight in something the manager dislikes, holding none of it is not neutrality, it is the largest bet in the book. At its extreme in this pairing: the composite holds 40 per cent in bonds. A portfolio holding no bonds at all would be running an active position of minus 40 points. The mandate caps any single holding at 5 per cent of the portfolio, so the largest position available through buying anything is 5 points. The biggest position this mandate can take is not a purchase at all, it is an absence, and it is eight times the size of the largest holding the constraints permit.

Put those same three positions into rupees and the asymmetry stops being an abstraction about points.

The asymmetry produces a reflex worth naming without moralising about it. Distance from the benchmark gets questioned and closeness to it never does, so a manager who wants a quiet year is pushed toward the benchmark whatever their view. Hold what the composite holds and no month brings an awkward conversation. Hold something else and every month brings one, including the months where the difference was simply noise.

The reflex is a behavioural consequence of the measurement itself rather than a criticism of anybody, and it operates on careful, honest people exactly as it operates on anyone else. The everyday version is a school that grades on attendance. Nobody instructed the students to stop thinking, and within a term the whole class optimises for the register.

What does a mismatch between mandate and benchmark actually cost?

Two things, and the second is the one that stings. The first is a permanent difference in return, priced below as a drag. The second is a permanent contribution to active risk that no decision produced.

Active risk is the volatility of the difference series between the portfolio and its benchmark. Look closely at what feeds that series here. In every single period, 10 points of the Anantara portfolio earn a cash return while the corresponding 10 points of the composite earn a bond return. The cash return and the bond return are not the same in any period, so a difference lands in the series every period, forever, from a bucket the mandate ordered the manager to hold.

Active risk measures distance and carries no record whatsoever of what put the portfolio there, so the constraint's contribution is charged to the manager anyway. The arithmetic cannot distinguish a constraint from a conviction. Both are simply the portfolio not matching the composite, and both enter the calculation with the same sign and the same weight.

What does the Anantara pairing look like once a price is put on it?

Line the two up bucket by bucket and then price the gap, using the holder's own stated assumptions and labelling them as the holder's own wherever they appear. The Anantara Multi-Asset Portfolio assumes equity returns 12.0 per cent, fixed income 7.5 per cent and cash 6.0 per cent. The endowment chose those three assumptions. Assumptions are not forecasts, not market expectations and not anybody's published estimates, and a different set gives a different answer.

Read that last row carefully. The total row hides a base trap of exactly the kind this area exists to teach. The absolute differences add to 20 points, but that counts the same money twice, once in the bucket it left and once in the bucket it landed in. The amount of the portfolio actually sitting somewhere the composite does not go is 10 points, or Rs 50 crore on Rs 500 crore. The gap is a structural differenceA gap between a portfolio and its benchmark that comes from the way the two were defined rather than from anything a manager chose to do. of 10 points of the whole portfolio, and it exists because the mandate requires it rather than because anybody decided anything.

Now price it. Holding 10 points in cash rather than in fixed income means 10 points earning an assumed 6.0 per cent instead of an assumed 7.5 per cent, a difference of 1.5 points on a tenth of the portfolio. The arithmetic is 0.10 times 1.5, or 0.15 percentage points a year. On Rs 500 crore that is Rs 75,00,000/-.

There is a second route to the same number, and it is worth running because it confirms the first. On the holder's own assumptions the policy portfolio at 60, 30 and 10 is expected to return 10.05 per cent. The composite's shape at 60 and 40, priced with those same assumptions, comes to 0.60 times 12.0 plus 0.40 times 7.5, or 10.20 per cent. The difference is 0.15 points, arrived at from a completely different direction.

The 0.15 points is a drag on the recorded excess return that arrives before the manager does anything at all, and against the year's 1.6 points of gross excess it is roughly a tenth of the whole result. Nobody chose it. The drag cannot be traded away without breaching the mandate. And it is sitting inside every performance figure struck against this pairing.

Then there is the harder consequence, the one that surfaces the moment anybody runs attribution on this pairing. The cash bucket has no counterpart in the composite at all. Computing an allocation effect requires a benchmark return for every bucket, and for cash there is not one, so somebody has to adopt a convention: price the missing bucket at the composite's overall return, or at the bond return, or at the cash return itself. Each is defensible. Each gives a different allocation effect. The convention is a choice, it is usually made once by somebody junior, and it is almost never written down.

Close on the honest position rather than a verdict. The Anantara pairing is usable. The composite is specified in advance, investable and measurable. The composite fails the fit question, so every figure struck against it carries a known structural difference that has to be stated rather than quietly removed.

Who should build the yardstick, and why does it matter?

The answer is short and not complicated. Where the person being measured also chooses the measure, the measure stops being evidence. Not because anybody cheated, but because a number that could have been set differently after the fact carries no information about the thing it was supposed to test.

A custom benchmark is written into the mandate at the outset and dated for exactly that reason, and its construction rules then exist on paper before the period they will judge. A dated construction rule converts the custom form from something only its author can reproduce into something any outsider with the document can rebuild. Dating does not remove the fit advantage. Dating removes the exposure.

How does a committee actually use this on a Tuesday morning?

One number, computed once, written on the opening sheet of the review pack. Everything above reduces to that single line.

An investment committee like Rukmini Deshpande's asks for the structural difference between the mandate and its composite to be computed before the meeting, in weights and in expected points, and printed at the top of the performance summary rather than in an appendix. For the Anantara Multi-Asset Portfolio that line reads: 10 points of cash the composite does not hold, Rs 50 crore, an expected drag of 0.15 points a year on the holder's own assumptions. The discussion of the manager's year then starts from what is left, the part anybody could actually have influenced.

A return figure struck against a benchmark that fails the fit test is a number about the pairing rather than a number about the manager, so a consultant reviewing an arrangement runs the four questions first and the returns second. A lender or a trustee reading a report asks one question that costs nothing: was this yardstick written before or after the period it is judging, and where is the date?

A household does the identical work without any of the vocabulary. Someone comparing their savings against a cousin's is running a peer comparison, with all three faults intact: the cousin's result was not available to be chosen, the cousin they compare against changes with the year, and the relatives who did badly are not talking about it. Someone comparing their savings against a deposit rate they could genuinely have taken is running a proper benchmark. The second comparison is worth having precisely because it was available in advance, and the first one feels more informative while saying almost nothing.

The subtraction definition stands against the two readings it displaces.