Engineering Consulting Cases 2026: The Second Door

The firm that says the assessment is different

Start with the sentence that settles the route question. McKinsey's interviewing page describes the standard path for client-facing work as "a personal experience interview, followed by a problem-solving interview". Then it carves out something else entirely: "If you're applying for a technical role, you'll undergo an assessment tailored to evaluate your specific technical skills. This may include coding challenges, problem-solving tasks, or other role-specific exercises designed to assess your proficiency in the technical areas most relevant to the position."

The same page carries a second version of the split, on the screening rather than the interview: a candidate may be asked "to complete a test or participate in a game, depending on the client-facing role you're applying for", and those assessments cover "problem-solving, coding abilities, and more."

So the firm publishes two assessment regimes and lets the role decide which applies. What it does not publish is which roles count as technical. The page never names one. Its analytics arm carries coding requirements on its own job pages and is not mentioned here at all. An engineer reading this page in good faith cannot tell which door they face.

The analytics arm answers the degree question directly, though. Its data scientist page states a requirement rather than a welcome: "You have an advanced degree in a quantitative field, such as computer science, engineering, mathematics, physics, or statistics". That is the only page in our reading where an engineering degree is a condition of entry instead of a tolerated origin. The interview that follows it is McKinsey's own, with one change, and that change is set out in the QuantumBlack interview guide.

The one published technical route, case interviews in both rounds

One firm publishes the whole technical path with its stages named. BCG X's Australian associate recruitment page runs four steps under the heading Selection Process, and two of them are described.

The first round is "Virtual technical interviews, including case studies and technical questions". The final round is "Additional case-based and technical interviews to assess your problem-solving and technical skills". Success produces "an offer to join BCG's Consulting team", and the target degrees are named: computer science, artificial intelligence, data science and adjacent fields.

Read that closely. The technical door contains the case and adds to it. Both rounds pair case work with technical questioning, so the preparation adds up rather than swapping over.

The firm's general interview page tells you nothing about this. The name of that technical unit does not appear on it. The sequence exists and is published, at a regional address a candidate finds only by hunting for it.

The technology practice alongside it publishes a surprise. Its preparation page gives case instructions that would fit any generalist guide: "Structure your approach by breaking the problem down into smaller parts", "Make assumptions and calculations explicit so the interviewer understands your thought process", "Take the time to conclude and refer back to the original question", and "Focus on demonstrating a logical thought process and don't worry about getting the 'right' answer." A technology practice giving generalist case advice is the clearest signal available that the case does not soften for an engineer.

Where engineering is named as a qualification

Firms mostly describe backgrounds in aggregate. One document names engineering in a requirements line, and it does so three times.

OC&C's 2024 graduate brochure has nine office sections. Five of them carry a line headed Qualifications, and three of those five name engineering by name.

Paris is the row that matters. An engineering school master's sits on the same footing as a business school one, which makes engineering a route in its own right. It is also the only office of the nine that asks for a master's at all.

Four office sections carry no qualification line at all: Warsaw, Australia, the Greater China offices and the United States. So within one firm, engineering qualifies in three places, goes unmentioned in two, and the subject is skipped in four.

The widest published figure on background belongs to another firm, and it has a limit. Roland Berger writes that "Half of our consultants have backgrounds in engineering, natural sciences, social sciences, or the humanities." That is four families in one number. The share held by engineers alone is not published by anyone, and we do not split the aggregate.

A published case, recomputed operation by operation

This section exists only because somebody redid the arithmetic. Porsche Consulting publishes two practice cases and invites candidates to "Test your capabilities and solve the two test cases to be perfectly prepared for your interview." For the second case it publishes a worked solution. Underneath it sits the derivation of the assumption that solution rests on. That derivation is where a reader who checks finds something.

The question: "How long would it take for the Berlin Airport BER to recoup its construction costs with passenger air traffic alone?" Ten minutes to work it. The firm supplies three inputs. Construction costs of about 5.1 billion euros. Potential passengers of about 26 million. Turnover per airline passenger of about 20 euros. Interest and operating costs are excluded in writing.

The published answer is three lines: "€20 x 26 million = €520 million", then "5.1 billion / 0.52 billion", then "= 9.8 years".

We redid it. Twenty times twenty-six million is 520 million, which holds. Five point one billion divided by 520 million is 9.8077, so 9.8 years, which holds. The passenger figure itself checks out too: 19.5 million for one airport plus 6.5 million for the other is exactly 26 million.

Then the firm explains where the 20 euros comes from. It uses one full aircraft with 108 passengers, and that is where our recomputation diverges.

The passenger line is printed as 117.48 in the working and as 117.46 in the total. Neither figure is the product. The fee is 1.06 per passenger for 108 passengers. That gives 106.00 for the first hundred and 8.48 for the remaining eight, so 114.48 in all. The weight line is also printed twice, as 43.74 and then as 43.73.

The rest of the derivation holds. Passenger charge at 15.90 times 108 is 1,717.20. Security charge at 1.21 times 108 is 130.68. Stowage is a flat 90 euros. Adding those to the firm's own landing total of 202.89 gives 2,140.77, which is exactly what the page prints, and 2,140.77 divided by 108 is 19.8219, so 19.82 per passenger, also exactly as printed. The document is internally consistent with its own incorrect term.

Substitute the corrected landing total and the chain becomes 199.92 plus 1,717.20 plus 130.68 plus 90.00, which is 2,137.80. Divided by 108 that is 19.7944, so 19.79 per passenger.

The conclusion survives, and that is the point worth carrying away. Both 19.82 and 19.79 round to the 20 euros the solution uses, so the 9.8 years does not move. We publish both readings side by side rather than deciding for the firm. It owns its own case, and the gap changes nothing that matters.

The first of the two cases cannot be worked to a number at all. Pumps Inc. gives five pumps, three process steps, and three exclusion constraints, including that "For technological reasons, Pump L can only be produced on the small "mechanical production" machine." The product-process matrix the question turns on is announced and never supplied. So the case teaches structure and stops, and saying so beats inventing the missing table.

What that recomputation teaches an engineer in a case interview

An engineer will recognize the habit in the section above, because it is the habit of a lab notebook. Somebody checked a product against its factors. A three-euro gap turned up in a derivation nobody was auditing.

That habit is worth more in a case than most engineers expect, and less in one precise way. A check announced out loud earns credit. A check run in silence until the number is perfect loses it.

The reason is mechanical. A case runs about twenty-five minutes. An estimate cannot be better than the inputs it was built from. Carrying three decimal places through invented inputs buys precision those inputs cannot support. A rounded number said out loud, with its assumption named, is the same work, finished.

The second habit worth adjusting is order. Technical training rewards reaching the answer, then showing the derivation. A case rewards the reverse. The recommendation leads, the support follows, and you name the risk yourself. On our own synthesis scale, answer first carries 35 of 100. Those weights are ours, and no firm publishes one.

Neither adjustment fixes a deficit. Both reorder a skill you already have. That is why engineers tend to move fast once somebody tells them which way round to put it.

One named account of a technical process

Firms publish stages. They rarely publish durations. One named account fills part of that gap and it needs its label.

Constantin Weisser, who was to join a technical consulting arm after a doctorate in physics, statistics and data science, described his process to the MIT consulting club. Six stages, with times attached.

1. "a two-hour coding assessment that was similar to the tech firm challenges seen on LeetCode. It tests computational thinking, familiarity with scripting languages and knowledge of machine learning libraries."
2. "a 30-minute screening interview filled with rapid-fire technical data science questions."
3. "a 90-minute data science case study interviewer asks the interviewee to describe how they would tackle the experimentation and data science modeling of a real-world case."
4. A technical presentation about his own past work, which he describes as needing to be "clear and approachable".
5. A personal experience interview, which he reads as matching the firm's general format.
6. "a 60-minute technical experience interview with ML experts will probe the limits of your understanding of ML."

His own summary of what the sequence selects: "Overall, the interviews select people with technical expertise and proven interpersonal, presentation and leadership skills."

Label this correctly. One person. One doctoral cohort. One process that ran in 2020 or early 2021, published by a university club rather than by the firm. The firm publishes neither the stage count nor the durations, so the six stages are his. The only bridge he draws to the generalist process is the experience interview, and that bridge is his reading. This is an anecdote. What it adds is the only minute-level breakdown of a technical consulting process in our corpus.

Where the doctorate actually concentrates

One more figure belongs here, and it comes from a regulator rather than a careers page. It reframes which category wants a technical background most.

CRA International's annual report for the year ended 3 January 2026 states its headcount plainly: "we employed 959 consultants, consisting of 164 officers, 563 other senior staff and 232 junior staff." Then it states composition: "Approximately 79% of our senior staff have advanced degrees, with 51% of the advanced degrees being doctorate degrees".

Multiply the two shares. The product of 0.79 and 0.51 is 0.4029, so roughly 40 percent of senior staff at that firm hold a doctorate. We publish that percentage and no headcount, on purpose. The same paragraph separates officers from other senior staff, and the sentence does not say which group it means. The base is either 563 people or 727. A percentage does not depend on which; a count does.

This is economic consulting rather than strategy consulting, and it is the only firm in our reading that files a number like this with a regulator. A quantitative background comes closest to being the majority credential in the very category most readers skip.

Four things to settle before choosing a route

In order, because the first decides what you prepare.

1. Establish which door your target role sits behind. One firm publishes two assessment regimes and never names which roles are technical. Ask the recruiter directly, because the answer changes whether you rehearse a coding exercise.
2. Assume the technical door adds the case rather than replacing it. Both published rounds of the one complete technical sequence pair case work with technical questioning.
3. Check whether your target office names engineering at all. Three office sections of one firm do, one requires a master's from a business or engineering school, and four say nothing about qualifications.
4. Decide what the business gap costs you. One firm publishes that its two-week induction covers business basics for those without a business degree. That is an answer to the gap after you are hired, and no firm publishes what the gap costs during an interview.

One route sits outside all four. A technology practice hires engineers against a published process of its own, with a different artifact and a different assessment, and the comparison between that route and the strategy route is set out on tech consulting.

The weights our own drills apply, and what each rewards

No firm publishes a rubric. The scale below is ours, set by us and read from our own code, and the right-hand column is our reading of an engineering background rather than any firm's finding.

Our math rubric weights Answer 40, Working 25, Units 15 and Interpretation 20. Units is the one an engineer takes for free and most candidates lose, and Interpretation is the one a technical training tends to under-serve, because it asks what the number means for the client rather than whether it is right.

The structure drill and the synthesis drill both run in the browser without an account, and how an answer is assessed sets out every weight with the reasoning behind it. Neither drill teaches business content, because the published technical assessments test tools we do not simulate.

What no firm publishes about engineering backgrounds

Five gaps, each left open rather than filled.

Nobody publishes the share of engineers. One firm gives half its consultants across four aggregated families, and nobody separates them. We do not split that number.

Nobody defines a technical role. The firm that publishes a separate technical assessment never lists the roles it applies to.

Nobody publishes whether the two doors lead to the same title. The complete technical sequence promises an offer to join a consulting team and does not name the level.

Nobody dates the published case. The airport derivation cites an aeronautical information package without a year, so the fees cannot be checked against a source, which is a second reason the recomputation above publishes both readings.

And nobody publishes the matrix that the pump case turns on, so that case ends at structure by construction.

For the syllabus question rather than the route question, cases without a business degree carries what firms publish about non-business origins. For the doctorate specifically, what a PhD is worth at an MBB firm counts the published advanced degree routes, and PhD and postdoc cases carries the application mechanics.

What changed on this engineering page, and when

1 October 2026. This replaces a 233-word page built entirely on assertion. It claimed engineers arrive strong on one scored dimension and weak on another, that precision works against them, and that they solve before framing, and it cited nothing for any of the three.

Two of those claims survive in a different form and are now labeled as our reasoning rather than as findings, which is what they always were. The third, about a consistent pattern across engineers, is gone: we have not measured it, so §6.8 of our own process puts it out of bounds.

The old page also had no route question in it at all, which was the real omission. An engineer's first decision is which door to apply through, two firms publish separate assessment regimes for those doors, and the previous version did not mention that any choice existed.

Added in this version: the published technical assessment carve-out and the absence of a role list to go with it, the one complete technical sequence and the fact that both its rounds contain cases, the three office qualification lines that name engineering, a firm-published case recomputed operation by operation with a three-euro discrepancy shown and the conclusion demonstrated to survive it, one named six-stage account with its durations and its limits, the roughly 40 percent doctorate share computed from a regulatory filing with its base ambiguity stated, and five gaps nobody fills.