"Tell Me About a Failure" 2026: No Firm Publishes It

No consulting interview guidance publishes a failure question

We have read the behavioral guidance of McKinsey, BCG, Bain, EY, Deloitte, PwC, KPMG, Accenture, Kearney, Roland Berger, Simon-Kucher, Oliver Wyman, L.E.K., EY-Parthenon and Strategy&, plus two banks. Not one publishes a question asking for a failure.

Accenture is the clearest test, because its behavioral interview post publishes six questions in full. They cover a challenging project, adopting something new, a difficult colleague, learning a skill, overcoming conflict, and handling pressure. Every one of them is a question you can answer with a success. None asks what went wrong and stayed wrong.

BCG goes further and removes the usual disguise. Its published behavioral article says candidates will not be asked about strengths and weaknesses in that style of interview, which is where the rehearsed near-failure normally lives.

So why prepare a failure answer at all? Because interviewers ask it whether or not their employer prints it, and because four of the lists above want evidence that only a failure supplies. Learning agility with no failure behind it is a claim about a course you took.

What the closest published consulting interview question asks

McKinsey publishes four named themes with a definition and one exact question each, inside the interviewing module embedded on its interviewing page. The fourth is Courageous Change, defined as: "Growing as a professional requires constant learning and openness to feedback and change." Its question: "Revisit a time when you had to rapidly learn something new to tackle a challenging situation."

Read it twice, because it is counterintuitive. The theme is called courage and the question is about speed of learning. It does not ask what you got wrong, and it does not ask how you felt. It asks what you did not know, how quickly you closed the gap, and what the situation demanded.

That reframing is the most useful thing on this page. A failure story built to McKinsey's published question has this shape: here is what I did not know, here is how I found out, here is how long it took, here is what it cost while I was finding out. The cost is where the failure lives, and it arrives as a fact rather than as a confession.

Two other published items point the same way. Simon-Kucher names learning agility as one of four things a behavioral example must demonstrate. Accenture names ability to learn among six behaviors and publishes two of its six questions about learning something new. So the sector publishes a great deal about learning and nothing about failure, and a candidate who understands that difference writes a better answer than one who prepares an apology.

Note the arithmetic McKinsey attaches to its themes: the firm suggests coming prepared with two personal examples for each of the four areas, which is eight slots, since 2 times 4 is 8. With five solid experiences each story has to serve 1.6 slots on average, because 8 divided by 5 is 1.6. A failure that also demonstrates rapid learning covers two of the eight by itself, which is the efficient use of a bad memory.

Why the honest version beats the disguised one

Deloitte publishes the mechanism on its interview tips page, in the plainest sentence any firm gives on this subject: "Note: it's important not to make stories up. Your interviewer will know the difference, especially when they ask follow-up questions."

Apply that to the three standard disguises and you can see each one fail.

The humble brag. "My failure was caring too much about quality." The follow-up is what it cost, in money or in time, and there is no answer because nothing was lost.

The safe failure. A missed deadline on a group project in first year. The follow-up is what changed afterwards, and a story chosen for being small has no consequence large enough to have changed anything.

The borrowed failure. Someone else's decision, narrated as though the speaker owned it. The follow-up is what you would have done differently, and the answer drifts into what they should have done.

BCG's published standard is the backstop: "Your interviewer will expect a real-life example in your answer." A real example has a date, a person who was affected, and a number. The disguised versions have none of those, which is why they die on the second question rather than the first.

The four parts a failure answer needs in a consulting interview

STAR gives a failure answer a shape and leaves out most of what a failure answer is for. Deloitte and BCG both publish four-part versions of STAR, and neither has a slot for any of the following.

1. The decision. What you chose, and what the alternative was. A failure with no choice in it is bad luck, and bad luck says nothing about you.
2. The cost, quantified. What it actually took away, in money, hours, or people. This is the part that makes an interviewer believe the rest.
3. The diagnosis. Why it happened, distinguished from who was involved. A cause you can name is the difference between a lesson and a grievance.
4. The change, at the level of a rule. What you now do differently as a standing practice rather than as a resolution. "I would communicate more" is not a rule. "Anything over a certain threshold now gets a written check" is.

Part four is where most answers stop short, and it is the one that carries the most weight, because it is the only part that predicts your behavior in a job you have not started.

One warning on scale. A failure large enough to be interesting and small enough to be yours is the target. Something that cost an employer a client is usually not yours alone; something that cost you a weekend is not interesting. In between sits the useful range.

One failure answer, with the arithmetic

The example below is ours. It is a demonstration rather than an answer to reuse, and Deloitte's published warning is the reason: an invented version survives the first sentence and dies on the second.

The answer. "I sold sponsorship for a student newspaper. I priced it at 400 dollars an issue and told the committee I would sell eight of the eight issues, so 3,200 dollars. I sold three. That is 1,200 dollars against a 3,200 target, about 38 percent, and the paper cut its color pages for the rest of the year.

The reason I got it wrong was a unit error in my own pitch. I had quoted a readership of 4,000, which was the print run. When I finally counted pickup across three issues, it averaged about 1,450 copies taken, so 36 percent of what I had been claiming. At 400 dollars that was about 276 dollars per thousand readers rather than the 100 I had told people, and two of the three buyers who said no told me the number looked wrong before I knew it was.

What I changed is a rule rather than a habit. I measured pickup for three issues, repriced at 180 dollars, and sold six of the remaining seven. And I now make anyone quoting an audience figure say which count it is: printed, delivered, or taken."

Recompute: 1,200 divided by 3,200 is 0.375, so about 38 percent. 1,450 divided by 4,000 is 0.3625, so about 36 percent. 400 divided by 1.45 thousand readers is about 276 dollars per thousand, against 400 divided by 4 thousand, which is 100. Six issues at 180 dollars each is 1,080 dollars.

All four missing parts are in it. The decision was the price and the promise. The cost is 2,000 dollars short and color pages lost. The diagnosis is a named unit error rather than a market that said no. The change is a rule about which audience count is being quoted, which a listener can imagine applying somewhere else.

The one published account of a real rejection

Firms do not publish stories of candidates they rejected, which is worth saying out loud, because it explains why every account of a consulting failure you read is written by someone who eventually succeeded.

One exception is on the record, from a trade title rather than from a firm. Consultancy.uk published an interview with a consultant who wrote: "I did not move forward after the second interview at A.T. Kearney in September 2016. I realised my approach in solving their interview case was a little too entrepreneurial and lacked the structure A.T. Kearney was looking for." He adds what he changed: "I decided to spend more time in refining my structure and practicing more cases with fellow students."

Hold it loosely. One candidate, one office, one year, under a firm name that changed in 2020, and nobody at the firm confirmed the reason. What makes it worth quoting is the shape rather than the content: a named cause and a named change, which is exactly the structure the four parts above ask for.

No firm publishes a score for this answer. The weights below are ours, applied by our own grader and read from the code that scores, and no firm publishes a marking scheme for a behavioral answer of any kind.