Portfolio Construction Interview Questions: Sizing, Limits and Risk

A candidate walks into a long-only interview with three good pitches and delivers all three cleanly: thesis, catalyst, valuation, risk. The interviewer nods, then asks a different question entirely. How big would each position actually be, and why. What happens to the other twenty names already in the book once these three go in. Which one gets cut if the fund only has room for two.

That is the moment a strong stock-picker and a strong portfolio manager stop looking like the same candidate. Picking is one skill, defended in a pitch; building a portfolio out of the picks is a different one, and it gets tested badly by almost everyone preparing for this interview, because it rewards structure over enthusiasm and most preparation only builds enthusiasm.

2

skills tested

picking an idea, and sizing it

5% / 10% / 40%

a UCITS single-issuer limit

one real example of a constraint

Both

what concentration raises

expected active return AND tracking error

Why picking a good idea and building a good portfolio are different tests

Underneath the sizing question, the diversification question and the constraint question sits one thing: does the candidate understand that a portfolio is a single object with its own risk, return and limits, not a container that good ideas get poured into until it is full. A candidate who can hold that frame answers every version of this question the same way, because the frame does the work the specific facts cannot.

Where this shows up in the room

A portfolio construction question rarely announces itself. It usually arrives disguised as a follow-up to a pitch that just went well.

Diversification is about correlation, not count

The myth: a diversified portfolio just needs enough names in it

Ask a candidate what makes a portfolio diversified and a common answer is a number: twenty names, thirty names, "enough to spread the risk around." It is repeated often enough to sound like a rule, and it is wrong on its own terms. A portfolio of thirty stocks that all sit in the same crowded sector, moving together through the same cycle, is not diversified. It is thirty tickets on the same bet.

What looks diversified and what is

The question that catches this live

"How many names does a diversified portfolio need?" is a trap question precisely because it invites a number. The stronger version an interviewer actually wants answered is what those names would need to look like relative to each other, and a candidate who redirects the question toward correlation rather than count has already passed it.

Position sizing: what sets the weight

Conviction sets the case, not the number

Conviction answers one question: should this be in the portfolio at all. It does not answer how big it should be, and treating the two as the same decision is the single most common gap in an otherwise strong pitch. Two ideas a candidate believes in equally can, and often should, end up at very different weights.

The four things that decide the size

A worked example, with the arithmetic shown

Say a manager has equal conviction in two ideas, Position A and Position B, and sizes both at 4% of the portfolio because the case for each is equally strong. Position A has a tracking error of 2% against the benchmark; Position B, in a more volatile, less liquid corner of the market, has a tracking error of 6%.

Active risk from a position scales with both its size and its own tracking error, so Position A contributes roughly 0.08 (4% times 2%) and Position B contributes roughly 0.24 (4% times 6%) in the same units. Sized identically, Position B is contributing three times the active risk of Position A for the same stated conviction.

Position A (2% tracking error)

0.08

Position B (6% tracking error)

0.24

Both positions carry the manager's full conviction and the same 4% weight. Position B's higher tracking error means it is quietly running three times the active risk of Position A for an identical stated view.

Sizing both at 4% treats them as if they carried equal risk, and they do not. A manager who wants each idea to contribute a similar amount of active risk would size Position B down, or Position A up, until the two numbers converge, which is exactly the adjustment "size by conviction alone" skips.

The concentration trade-off, and why it is never a one-way street

Both, not one

A concentrated portfolio is often described as riskier, which is true, but the way it gets riskier is the part candidates miss. Concentration raises tracking error and expected active return together, because both are functions of the same thing: how far the portfolio's holdings sit from the benchmark's. A candidate who describes concentration as a straight trade of safety for return is answering a simpler question than the one being asked.

Why this is worth saying out loud in the room

Naming the trade-off correctly signals something an interviewer cares about more than the definition itself: real thought about what happens when a concentrated call goes wrong, not just when it goes right. Higher tracking error, run on purpose, is a deliberate choice with a cost attached.

Constraints are the design space, not the boring part of the job

The UCITS 5/10/40 rule, in plain terms

Here is one real constraint, stated exactly. Under the EU's UCITS rules, a fund cannot put more than 5% of its assets into securities from a single issuer. A regulator can raise that cap to 10% for any one issuer.

That relief comes with a condition attached. Every position the fund holds above 5% has to add up, across all of them, to no more than 40% of the fund's total assets. That aggregate ceiling is what stops the raised limit from becoming a loophole.

Why a great idea can be un-ownable

A single-issuer limit does not care how good the thesis is. A manager who is fully convinced a name deserves an 8% weight, inside a fund capped at 5%, simply cannot express that conviction at full size, no matter how the pitch goes.

A limit treated as a nuisance sounds like a student. One who treats the limit as part of the design problem, something to build a portfolio around rather than complain about, sounds like someone who has run money inside a mandate.

The other limits inside the same design space

A single-issuer rule is one constraint among several, and a mandate typically stacks more than one of them at once.

What else shapes what is ownable

None of these are obstacles to good investing. They are the shape of the job itself, decided by the client before a single trade is placed, and a portfolio manager's actual skill is building the best portfolio available inside that shape rather than the best portfolio that would exist without it.

Liquidity relative to fund size, and the limit nobody names unprompted

Daily dealing versus a holding that cannot be sold in a week

A retail-facing fund typically has to let investors redeem on a set, short cycle, often daily. That single fact quietly caps how illiquid any one holding can be, independent of conviction, independent of the single-issuer rule, and independent of every other limit already on the list.

A brilliant idea in a name that trades thinly can still be right and still be unownable at any meaningful size. Building the position, or unwinding it under redemption pressure, would move the price against the fund doing the buying or selling.

Why this is worth raising before anyone asks

Interviewers reach for liquidity questions constantly because almost nobody mentions liquidity unprompted. Raising it unprompted signals real experience rather than memorised theory, because it is the constraint that shows up first in practice and last in most preparation.

Rebalancing, and why doing nothing is also a decision

Drift is not neutral

Left alone, a portfolio's weights drift toward whatever has performed best, because that holding keeps compounding a larger share of the total while everything else stands still by comparison. That drift is not a flaw in the portfolio; it is what happens by default, and it quietly concentrates the fund exactly where it has already run the furthest, which is precisely the moment a reversal would hurt the most.

Calendar rules against threshold rules

Research on target-date fund rebalancing has found a threshold-based approach balances the cost of trading against the cost of drifting better than a fixed calendar does: letting an allocation drift by around 200 basis points, then rebalancing it back to within about 175 basis points of target.

There is no single frequency that always wins. The point worth making in the room is that leaving a portfolio untouched has a cost too, and it is the cost of quietly running more risk than was ever intended.

What a risk model tells you, and where judgment takes over

What a factor model is doing under the hood

A multi-factor risk model does not look at a portfolio stock by stock and guess what might go wrong. It decomposes the portfolio's risk into a much smaller set of common drivers, plus whatever is genuinely specific to each individual stock:

• Sector and industry — how much of the risk comes from being over- or underweight a given part of the market
• Country and currency — where the portfolio's geographic and currency exposures actually sit, not just where the listings are
• Broad investment style — factors such as value, quality or momentum that cut across sectors and countries alike
• Stock-specific risk — whatever is left once the common factors above are accounted for

That reduction is what lets a model produce a usable risk number instead of drowning in noisy historical correlations between thousands of individual names at once.

Where judgment has to take over

What the model gives you, and what it never will

The failure mode this produces has a name candidates rarely use correctly: managing to the model. A manager who treats every risk-model output as a hard instruction, rather than one input alongside their own judgment about what has changed in the market, has handed the decision to a tool that was never built to make it alone.

Unintended bets: the classic portfolio construction reveal

How a stock-picker ends up running a factor bet nobody chose

Picture a manager who builds a book of what they call "high quality" names, one sector at a time, each pick defensible entirely on its own. Run that same book through a factor risk model and it can reveal something the stock-by-stock process never surfaced: a large, single-direction tilt toward one investment style, built up almost by accident because quality names in different sectors tend to share the same underlying characteristics.

The manager never chose that tilt. It arrived as a side effect of picking good stocks one at a time, and it is exactly the kind of exposure a portfolio-construction question is designed to surface, because a candidate who has only ever pitched individual names has no reason to have gone looking for it.

Active share, and what it is measuring

Active share and tracking error get confused constantly, and they are not the same thing. Active share describes how different a portfolio's holdings and weights are from its benchmark, a structural measure of composition. Tracking error describes how much a portfolio's actual returns have deviated from the benchmark's, a volatility measure.

A portfolio can run a high active share with fairly ordinary tracking error, if its different holdings happen to behave similarly to the benchmark anyway, and the reverse is just as possible.

Active share

• How different the holdings and weights are from the benchmark
• A structural, composition-based measure
• Can be high even if returns barely diverge from the index

Tracking error

• How much returns actually deviate from the benchmark
• A volatility-based measure
• Can be modest even with meaningfully different holdings

The question bank: what a strong answer contains

Twelve questions a portfolio construction round genuinely asks, and what the interviewer is really checking each time. None of these have a single memorised sentence as the right answer; each one rewards a way of reasoning, not a script.

"How would you size a new position in the portfolio?"

Tests whether conviction gets treated as the whole answer or just the starting point. A strong answer names liquidity, benchmark weight, correlation with existing holdings and mandate limits as the things that turn a view into an actual number.

"What makes a portfolio diversified?"

Tests whether a candidate reaches for a count of names or for correlation between them. A strong answer explains that low correlation, not headcount, is what actually spreads risk, and can give an example of a large, undiversified portfolio to prove the point.

"Would you rather run a concentrated or a diversified book, and why?"

Tests whether the trade-off is understood as symmetric. A strong answer states plainly that concentration raises both expected active return and tracking error together, and picks a side based on the mandate and the client, not on a general preference for boldness or caution.

"What single-issuer or concentration limit would you expect a long-only fund to run under?"

Tests whether constraints are treated as real. A strong answer does not need the exact regulatory figure memorised; it needs to state that such a limit exists, that it shapes what is ownable regardless of conviction, and ideally a real example such as the UCITS 5/10/40 regime.

"A position you love cannot be sized the way your model wants. What do you do?"

Tests whether liquidity is understood as a hard constraint rather than a preference. A strong answer accepts a smaller position, a longer entry timeline, or passing on the idea entirely, rather than assuming the constraint can simply be argued around.

"When would you rebalance a portfolio, and when would you leave it alone?"

Tests whether doing nothing is understood as a choice with a cost. A strong answer names drift explicitly, describes a rule (calendar or threshold-based) rather than a gut feeling, and acknowledges that trading has its own cost too.

"What does a factor risk model tell you that stock-by-stock analysis does not?"

Tests whether the candidate has thought about aggregation, not just individual names. A strong answer describes how a book of defensible individual picks can still add up to one large, undiversified factor bet nobody chose on purpose.

"What is the difference between active share and tracking error?"

Tests precision under a question that sounds definitional but is really about whether two related ideas have been kept separate in the candidate's head. A strong answer states that one measures composition and the other measures return volatility, and gives an example of when they diverge.

"How would a pension fund and a university endowment build different portfolios from the same starting assumptions?"

Tests whether a candidate reasons from the client's liabilities rather than from generic asset-allocation theory. A strong answer starts with what each institution owes or expects to spend, then builds the portfolio backward from that obligation.

"Describe a time constraints changed how you would have otherwise built a portfolio."

Tests real experience rather than theory, even from a candidate with no direct portfolio management background; a personal investing example works if framed honestly. A strong answer names a specific constraint, not a vague one, and explains the actual trade-off it forced.

"What is the biggest risk in a portfolio you cannot see just by looking at the position list?"

Tests whether unintended, aggregate exposures are on the candidate's radar at all. A strong answer names a factor, sector, or currency tilt that only shows up once positions are looked at together rather than individually.

"Why would a long-only mandate hold anything outside its core equity and fixed income sleeve at all?"

Tests whether a candidate understands correlation as the reason, not return in isolation. A strong answer focuses on what happens to the whole portfolio in a downturn if the extra holding does not fall the same way, rather than comparing expected returns asset class by asset class.

The mistakes that end this part of the interview early

• Sizing every idea the same way. Treating conviction as the only input ignores liquidity, correlation and the mandate's own limits, and it is the single fastest tell that a candidate has only ever picked names one at a time.
• Describing concentration as a one-way trade. Higher tracking error and higher expected active return arrive together; naming only one signals the trade-off was never worked through.
• Treating constraints as scenery. A single-issuer limit or a liquidity term is not a footnote to mention if asked; it is part of how the portfolio gets built, and it is worth raising before the interviewer has to ask.
• Assuming the risk model has the final word. A model reports what is there. Deciding whether that exposure is worth keeping is still a judgment call, and answering as though the model made the decision skips the part of the job that is hard.

How to prepare, in order

1. Pick a portfolio you actually know, even a personal one, and size every position out loud with a reason for each weight. If the only reason is conviction, that is the gap to close first.
2. Learn one real constraint cold, such as the UCITS 5/10/40 rule, well enough to explain why it exists, not just what it says.
3. Practice describing the concentration trade-off in one breath: both expected active return and tracking error rise together, not one at the expense of the other.
4. Have a rebalancing answer ready that names a rule, calendar or threshold-based, rather than "when it feels necessary."
5. Be ready to say what a risk model would and would not tell you about a portfolio you describe, including where you would still need your own judgment on top of its output.
6. Prepare one example of an unintended exposure, real or hypothetical, where a set of individually good decisions added up to a bet nobody chose on purpose.

The constraints are the job

A pitch proves a candidate can find one good idea and defend it. This part of the interview proves something narrower and, for most candidates, harder: that the same person understands what happens once that idea has to share a portfolio with everything else already held, against a benchmark, inside a mandate that was written before the idea ever showed up.

Size by more than conviction, know that concentration cuts both ways, treat the constraints as the actual job rather than the paperwork around it, and be honest about what a risk model can and cannot decide for you. That is most of what a portfolio construction round is listening for, and it is the part a candidate who only prepared a stock story tends to skip.