Clarifying Questions 2026: The Test That Filters Them

Clarifying questions sit first on every prompt L.E.K. publishes

L.E.K. publishes three case prompts on its interview preparation page at lek.com, checked 30 September 2026: an airplane brainteaser, a brewery profit puzzle, and a pharmaceutical pricing prompt. Under each one, the firm prints the same list of approach questions, and the first item is identical every time: "What clarifying questions would you ask?"

Three prompts, three subjects, one opening move. The firm's numbered steps for the case say the same thing at step one: "Listen to the question, repeat or paraphrase to be sure you understand; state your assumptions and restate the facts."

Oliver Wyman puts it inside the first of its six method steps, under the heading Pinpoint the issue, at oliverwyman.com, checked 30 September 2026: "Ask clarifying questions." The next line in the same step reads "Summarize your understanding of the situation in your own words."

Bain publishes nineteen tips across three lists, and one of the nine general ones is to settle the objective before launching into an answer. McKinsey attaches its own version to the opening of its Diconsa case: "Develop an overall approach before diving into details."

Four firms, same placement. None of them says how many questions, and none says which.

The clarifying question test: does a different answer move a number

Here is the filter. Imagine the interviewer gives you answer A, then imagine they give you answer B. If both futures lead you to write the same first number, the question was conversation. If they lead to different numbers, ask it.

Applied to a sizing prompt, that test usually leaves you with three questions and rejects the rest.

The fourth row is the one candidates skip most often, and it is the cheapest of the four to ask. The third row is the one that costs most when it is skipped, and the rest of this page is about why.

Notice what the test does to the question you were tempted to ask about industry background. A summary of the market's history is interesting and it moves no number, so it waits. Ask it at the end, when the recommendation is already on the table and curiosity is free.

A published rate, and the unit hiding inside it

The Centers for Disease Control and Prevention publishes two figures for US marriages at cdc.gov, checked 30 September 2026, both provisional for 2023. There were 2,041,926 marriages, and the marriage rate was 6.1 per 1,000 total population.

Put those two together and you can recover the population they imply. Divide 2,041,926 by 6.1, which gives 334,742, then multiply by 1,000: about 334.7 million people. Check it the other way: 334,742,000 times 6.1 divided by 1,000 returns 2,041,926. The two published figures are consistent, and now you have a third number you did not have to look up.

The unit trap is inside the rate. Read "6.1 per 1,000 total population" quickly and it sounds like 6.1 people out of every thousand got married. It means 6.1 ceremonies per thousand people, and each ceremony involves two of them. The rate is per event, and the population is per person, and those are different denominators sitting in one sentence.

That is the single clarifying question worth asking before any market gets sized: what is one unit of demand here. A wedding is one unit. A person is half a unit. Anyone who does not settle that in the first thirty seconds will settle it in the last thirty, in front of the interviewer.

Sizing the same market twice, and why one is double

Take the US wedding market. The spend figure below is ours, declared as an assumption because the CDC publishes counts and nothing about money. Assume 30,000 dollars of spend per wedding.

1. Per event, which is the correct unit. 2,041,926 weddings times 30,000 dollars is 61,257,780,000 dollars, which rounds to about 61.3 billion dollars.
2. Per person, which is the error. Two people marry at each wedding, so 2,041,926 times 2 is 4,083,852 people. Multiply by 30,000 dollars and you get 122,515,560,000 dollars, about 122.5 billion.
3. The gap. 122.5 divided by 61.3 is exactly 2. The methodology was identical in both. One clarifying question separates them.

A factor of two is survivable if you catch it. What makes it dangerous is that nothing inside the arithmetic looks wrong. Every multiplication is correct in both columns. The error lives one level above the arithmetic, in what the multiplicand counts, which is precisely where a check on your own work never looks.

There is a fast defense. Say the unit out loud when you write it, in the form "weddings, at 30,000 dollars each". A unit spoken beside a number is a unit you cannot silently swap. Our page on market sizing questions applies the same discipline across other denominators, and case interview math tips covers the arithmetic hygiene beneath it.

A consultant's account of this error in his own case interview

This error is not hypothetical, and a firm published someone making it. Tom Adams, an associate in L.E.K.'s London office, describes his own case interview on the firm's careers blog at lek.com, checked 30 September 2026.

He was asked to size the wedding market. In his words: "I sat there a second and realized that I had market sized as if each person individually had a wedding instead of two people!" The interviewer's prompt, as he reports it, was "Well, that's great methodology, but are you forgetting something?"

Two things are worth taking from that account, and one thing is worth refusing. Take the fact that the methodology drew praise while the answer was twice too large, which tells you what the interviewer was weighing. Take the fact that the error surfaced at the end, at presentation, which is the most expensive moment for it to surface. He writes that the key was demonstrating a structured process rather than necessarily arriving at a set right answer.

Refuse the conclusion that the unit therefore does not matter. This is one person's account of one interview in one office, published by his employer, and it is a source for what he experienced rather than a rule about scoring. Nobody at the firm has said that a doubled market is free.

Clarifying questions that buy nothing, and what to do instead

Some openings fail the test predictably, and it helps to name them before an interview rather than during one.

1. Asking for information the prompt already gave. It reads as inattention, and the answer changes nothing.
2. Asking for the answer in disguise. "Do you think the price is too low?" invites the interviewer to do the case.
3. Asking five questions to look thorough. The opening is a fixed budget of about two minutes, and five questions consume it with nothing written down.
4. Asking a question whose answer you could assume out loud. Stating an assumption is faster and it keeps you in charge of the number.

The fourth is the most useful reframe on this page. Many questions become assumptions instead, which converts a request into a decision. Oliver Wyman publishes an instruction that supports exactly this, under its approach step: "Show your client why it's better than the alternatives. State your assumptions."

So the opening sequence is short. Paraphrase the prompt in one sentence, ask the two or three questions that move numbers, state the assumptions that replace the rest, then draw. Our page on the case interview opening statement covers the sentence itself in more detail.

Where the opening shows up in the scale

The weights below are ours, read from the grading code in src/lib/grading. No consulting firm publishes a marking scheme, which is one of the steadiest absences in our research files.

Drilling the clarifying questions in the first ninety seconds

The opening is the most drillable part of a case, because it is short and it has a checkable output.

1. Run the two futures test on ten prompts. For each candidate question, write the number it would move. Any question with a blank beside it gets cut.
2. Rebuild one sizing under a changed unit. Redo the wedding market per household instead of per wedding, holding spend at 30,000 dollars per event. Households do not change the event count, so the answer stays 61.3 billion dollars, which is the point: the right unit is the event, whatever the population is measured in.
3. Time it. Paraphrase, two questions, two assumptions, in under ninety seconds, out loud.

Our structuring drill times the opening and returns Coverage, Separation, Depth and Concision separately with the first correction named. The math drill covers the arithmetic that follows. Both are free and need no account, and neither one will decide what a unit of demand is in your prompt.

For prompts that hand you no equation at all, see unstructured cases. For what to do once the branches are drawn, see structured thinking.