Case Interview Tips & Techniques How to conduct a Case interview

In this trend nowadays, professional coaches are more and more expensives, and consulting jobs are more and more competitive, the skill of CONDUCTING a case interview is needed more than ever. This is my guide on how to conduct a good case interview. Make you sure you have good your mock interviewers and become one yourselves.

First thing first, you should be a decent case interviewee to become a good interviewer. So all the logical foundations, all the fundamentals, all the question types, all the tips and techniques, all the business intuition should be grasped very well. 

Now back to the main topic. There are 3 equally important phases in case conducting: Preparing the case, Giving the case, and Evaluation.

We also know that there are two extreme ends of the case spectrum: the Candidate-led and the Interviewer-led cases, which require totally different styles and techniques of conducting. Even though there are some universal principles, it’s helpful to talk about each in turn. In each type, we will loosely go through all three parts of case conducting.

Candidate-led Case

One of the most popular pitfalls of amateur mock interviewers is that they still conduct the case with the mindset of a “candidate”. Even though they come up with a case context and key question, amateur interviewers subconsciously try to solve that case themselves and somewhat form their own version of the ideal path to answers. All the data and facts they gather about this case is along this path and the poor mock candidate has no choice but to follow it. Then, what should have been an educational and natural problem solving becomes a guessing game of what’s on the mock interviewer mind. 

With coaches that have real consulting experience, the effect of this is less severe as the chosen “path” is indeed realistic and good. But for most of us trying to be mock interviewers, the subconscious paths on our minds are not necessarily optimum and forcing poor interviewee going along that can be destructive.

Pure candidate-led cases should be models of real projects where the candidate acts like the project team and the interviewer does NOT play the role of the Engagement Manager. The interviewer is only there to create and operate something I call the “case universe”. Let the candidate, a.k.a. the project team, freely explore and interact with that universe.

So what does a case universe contain?

  • A problem
  • One or a few root-causes
  • Lots of data regardless of whether it is used or not

A case universe should NOT contain:

  • Any fixed framework
  • Any issue tree
  • Any hypothesis
  • Any solutions

Just a case question, one or a few root-causes, and lots of surrounding facts and data. The candidate’s job should be to interact with that universe, identify the root-causes, and come up with solutions as systematically and quickly as possible. Along the way, he or she can be creative in using different issue trees, frameworks, and hypotheses … however needed to arrive to the root-causes.

Let’s look at an example. Let’s say I want to create a case universe of a local movie theater. A bad approach would be like this:


  • A movie theater facing falling revenue over past few months
  • Quantity times price.
  • No problem with price.
  • Quantity has actually been falling
  • Further break-down quantity
  • Using Customer Segments
  • Student segment got problems
  • Competitor just opened a big marketing campaign for students

This is how you solve a case, not how to give it. With this approach, it is extremely difficult for the interviewer to let the candidate freely explore the case in a natural and meaningful process.

If the interviewee rightfully decides to break the case using customer segmentation, the interviewer would likely direct him or her into the profitability issue tree.


Ideally, the case creation process is reversed and ever expanding. We start with a case question, one or a few root-causes and working upward from there. And the great news is: many times just by doing this, you realize your case universe is really creative and interesting. Continuing the above example:


A movie theater with falling revenue over past few months


A local competitor running a big marketing campaign for students


We lost … let say 70% of student traffic to the theater



Then imagine it in your head. You realize, okay 70% less students to the theater means less overall number of ticket sold, depending on the share of customer segment. Okay, that leads you to make up some cool data on customer segments.



What else? You can also imagine less students means less sales on food and beverage. This area should be hit even more severely as students eat and drink more than adults in theater. So that leads to you drawing up some really cool data on product distribution of the theater.


Even more, less student can also means increased average ticket prices even without the theater changing any listing price. That is because usually students get lower price. Less of them mean the overall weighted average ticket price goes down. From there, you can make up various price tables of various products and services for various customer segments.



That’s on the price side, how about the cost side? Now making up all sorts of possible costs and allocations to each segment. You can even see how profitable each one is.


Zooming out a bit, what’s up with the competitor? Let’s say there are 2 local theaters in this small town. Why do they attack the student segment? You can imagine that their location is further away from the local high school and our theater is traditionally better with students. From there, you can think about what drivers governing students’ decision on which theater to go to. You can also make up some details about a new college about to open in town.


And so on and so on … You just let your creativity take you to wherever it goes. Your case universe, starting out as a simple problem plus root-cause, now keeps expanding more and more with so many possible twists and turns, ready to challenge even the best interviewee. This gives you so much flexibility and freedom in the next stage: Giving the case


Giving a candidate-led case


In a real consulting project, the difficulty level not only depends on the nature of the problem, but also on how smooth the data gathering processes go. In a perfect world, for every single hypothesis one makes, there is a piece of data readily available to prove or disprove that. And for every data we request, the response also include further information that can hint us on “where to go next”, or in consulting term: “what is the next hypothesis”!

But in the real world, it’s never that easy. Data often comes with some lead-time and mostly in unusable formats. Consultants spend at least about 1/3 of their time hunting, sorting, and analyzing data; making them ready to use.

In case interviews, the interviewer can govern that difficulty level by adjusting the amount and detail level of data giving out. 

Let’s say, for the Movie theater universe above, the interviewee has just set up a basic issue tree of: Price X Quantity. Any normal interviewee would something like: “Has there been any change to the pricing recently?”


  • To make the case really difficult, the interviewer would just say: “the quoted pricing has been the same for the past several month”. Technically, this is not wrong. Even though the average price has gone up, the quote prices are the same.
  • To make the case really easy, the interviewer can just give our a pricing table, detailing the different price level for different customer groups. That is a strong sign, telling the interviewee to look into customer segmentation.
  • To make the case moderate, the interviewer can say: “the quoted pricing has been the same but the average price has gone up”. This also gives the interviewee a hint, but he or she has to work to get it. Something interesting is going on there, but gotta dive in to find out why.


What happen when your interviewee goes outside of your designed universe?


The lazy way is to direct them back into your designed area with lame reasons like “we don’t have that data”. The better way is to quickly expand your universe to that area while still keeping the whole system work. The idea is to still give the interviewee a fair chance to solve the case regardless of directions, as long as the methodology logically works.

Continuing the above example: let’s say an interviewee draw a cool issue tree of Internal vs External. Within Internal, he starts gathering data on company’s structure along with each department’s roles and activities.

I honestly think that, even with this issue tree, there is still a path leading to the root-cause. We have to give the interviewee a chance to find that path. So, play along and have fun! 

  • “How the company is organized, structurally?”
  • Now expand your universe as you speak: “It’s a fairly small and simple business with 2 Big Departments: 
    • Operation: in charged of making sure products and services are up to the highest standard. That includes but not limited to theater maintenance, quality control, show scheduling, etc.
    • Business: in charged of maximizing profits. That includes but not limited to: strategy, marketing, pricing, etc.”
    • Then drop a hint on the marketing: “The business department has been changing the marketing coordinator 3 times over the past 2 months”
  • A good interviewee would catch that hint right away and dive right into the marketing area. And couple more steps from there, the root-cause will show itself.

That’s how a good interviewer conduct the case. Don’t shut the interviewee down just because he / she use a different issue tree or go outside of the defined zone.




Generally, there are 2 ways: Give it during or give it in the end. Some people prefer one over the other.

To me, this greatly depends on the level of the interviewee. In the beginning where he or she is still learning and making too many mistakes, it’s much better to give feedback during the case. Both you and the interviewee remember much better what just happened.

When you have an experienced interviewee who doesn’t make as many mistakes, it’s then better to catch up in the end. You will not destroy the flow of the case and allow the interviewee to have the most realistic experience.

What to talk about in feedback sessions?

No.1: your feedbacks on specific points or moments of the case (if you haven’t done so during the case)

No.2: overall grading on various criteria. Go to our website to download the free evaluation sheet many McKinsey interviewers use.

No.3: next step! Give the interviewee your advice on what to notice, what to work on, and what exercise to do to improve.

All of the examples in my Case Interview End-to-end program follow this guideline quite closely.



Most cases nowadays are mixed between Candidate-led and Interviewer-led. But for the purpose of this video, we talk about each extreme end.

In candidate-led cases, the interviewee represents the “whole project team” and the interviewer represents the “case universe”.

Interviewer-led is a very different animal. The context is much more simple and the “case universe” is much less important. The interviewer’s role is not just about the “case universe” anymore. Now, he / she is also the Engagement Manager and the interviewee becomes a poor entry-level consultant. The interviewer will run the universe and also looks for weaknesses to exploit and attack. A good interviewee is the one who can stand the ground and successfully defends his or her position using sound logics and consulting methodology.



So how do we prepare to conduct an Interviewer-led case?


Draft a simple context with or without a well thought-out case universe. You can even start a case with a piece of data or chart. Then prepare a few satellite questions surrounding that context. Usually, there are always follow-up questions you can come up with during the case, depending on the response of the interviewee. Unlike the candidate-led cases, you should have some clear idea of the solutions for each question so you can be ready to attack. There is a wide variety of question types you can choose from and it’s quite easy to come up with a few. Please see the article on our website (link below) or watch a future video on the Interviewer-led cases for a full list of possible question types.

For example, let’s prepare a case interview with the same movie theater.



Context: A local movie theater with recent revenue drop. There are two theaters in that relatively small town with 100,000 people.


Question 1 – Value Proposition: What factors determine a movie fan’s decision on which theater to go to?

An average answer would just list out factors like: location, price, quality, showtime, etc.

A bad answer would also list out some of those factors, but not all.

A really good one would break down customer into different segments and have a different answer for each group.



Question 2 – Market-sizing: What is the market size of movie theaters in that town?

A bad answer would dive straight into the estimation.

An average answer would ask for clarifications of units and definition before jumping into the calculations.

A great answer would take into consideration all qualitative insights already given in the case and even attempt to gather more. Then, breaking down the market-sizing problem where each branch can be easily estimated.

For example: 

  • Insights from Question 1 already tell us that customer can be broken down into various group. 
  • Insights newly acquired here in Question 2 can tell us that there are more than one way a movie Theater makes money, … hence a breakdown by product type: tickets, food & Beverage, and advertisement.

Then a mini BCG matrix can be drawn and so on.

Question 3: Calculation problem. Give the interviewee a set of data and ask for some math numbers.

Question 4: Solution. Given all insights of the problem and root-causes coming up throughout the case, what can be done to improve sales?


Exploiting and Attacking the interviewee

This is the very fun part. Now you have a chance to be an EM. Good or bad, depending on how well you grasp the consulting culture and methodology. You should attack as much as you can, regardless of the interviewee’s level. Here in interviewer-led cases, you don’t have to worry about case flow or hurting the interviewee’s morale. In fact, the more you attack, the more rhythm a case gets and the better your interviewee will learn. In many situations, I don’t even have to pause the case to give feedback in between; attacking is more than enough to send the message: “you ain’t doing it right, and here is why”! 

Here are a few common pitfalls I see many interviewee make that you can capitalize upon.

  • Why do you need that data?When the interviewee ask for a data which doesn’t serve the case direction or the running hypothesis.
  • How do you get that data?When the interviewee ask for an unrealistic data.
  • What does this mean? What implications can you draw from this?When the interviewee come up with a cool insight with but fail to link it back to the big case problem.
  • Does that number make sense?When the interviewee come up with a wrong calculation and doesn’t do a sanity check himself or herself.
  • What conditions or assumption do you also need to make that conclusion?When the interviewee make an unsound conclusion.
  • Is that a hypothesis or a conclusion?When the interviewee make a questionable statement without careful wordings.
  • Will that solution work? Is there any side effect of deploying that?When you see a questionable solution.
  • Why this? Why that? – When you hear any item you don’t agree with.
  • What else? – When you think the interviewee is missing some items.
  • And ETC..
  • And even many more depending on the specific context.

If you do this right, you don’t even have to pause the case to give feedback. And in the end, do an overall evaluation similar to the candidate-led one.

Each of us should have a case prepared, ready to mock interview others. So one really cool exercise for you is to write your cases out. If you comment yours in the comment section below, I personally will take a look at it and give you some pointers on how to make it even better.

Share on facebook
Share on google
Share on twitter
Share on linkedin

Leave a Reply

Notify of