"How many takeaway coffees are sold every year in a mid-sized city?" This kind of question opens or punctuates many consulting interviews, and it throws off candidates who are otherwise very comfortable with numbers. The reason is simple: nobody knows the answer, and the interviewer knows it. What they are looking at is how you build an estimate, how you justify it and how you check it. This article walks through a five-step method, a complete example built on fictitious assumptions, and the mistakes that cost the most points.
What the interviewer is really assessing in a market sizing
The firms are explicit about the spirit of the exercise. Bain describes the case interview as a structured conversation in which you work through a real business problem, without looking for a single correct answer: what matters is how you think, structure the problem and build on ideas. McKinsey, for its part, presents its problem-solving interview as a way to observe how you structure difficult, ambiguous questions, identify the important issues and formulate conclusions. Market sizing is a miniature version of that.
In practice, the interviewer scores four things.
- Structure: have you broken the problem into simple building blocks whose product gives the answer?
- Realistic assumptions: are a penetration rate, a purchase frequency or an average basket plausible, and are they presented as assumptions?
- Comfort with numbers: can you handle orders of magnitude without a calculator, rounding intelligently?
- Perspective: can you check your result through a second route, and say what you would need to know to refine it?
The final figure matters far less than these four points. A wrong but well-built estimate, corrected out loud after a cross-check, scores better than a correct result reached by luck that you cannot defend.
Starting from supply or from demand
There are two main ways to build an estimate. Bain presents them in its teaching case about a coffee shop chain: the top-down approach starts from a large number, such as the relevant population, and narrows it step by step; the bottom-up approach starts from a small quantity, such as one person's consumption, and scales it up.
The demand approach
It is the most natural one when the product is aimed at individuals. You start from the population, filter out those who can buy, estimate the purchase frequency, then the average price. Demand suits consumer goods, everyday services and subscriptions.
The supply approach
It starts from the players who sell: number of outlets, machines, vehicles or contracts, multiplied by what each one produces or bills. It is often more natural for a B2B market, an infrastructure or a piece of equipment: you count the sites, then their capacity, then their utilization rate.
Choose, then announce
Before calculating, say in one sentence which approach you are using and why. For example: "I will start from demand, because individuals are the buyers; I will then cross-check from supply." That sentence alone reassures the interviewer that you are in control of the thread. If you hesitate between the two, choose the one whose orders of magnitude you know best, and keep the other for the check.
Segmenting without drowning
A good segmentation separates groups that behave differently. A bad one multiplies categories until the calculation becomes unmanageable.
A few practical guidelines:
- Two to four segments are enough. Beyond that, you lose time and the interviewer loses the thread. If you have six, group them.
- Segments must be distinct and cover the whole scope. Working people, students, retirees and children form a partition of the population; "city dwellers" and "working people" overlap, and the calculation double counts.
- Segment by what really changes consumption, not by what is easy to state. For a takeaway coffee, age and activity matter; hair colour does not.
- Offer to handle only the dominant segments. Saying "I will set tourists aside, since they weigh little in a city of this size, and I will note it as an assumption" is a well-regarded move: it shows you know how to prioritize.
Draw your tree on the page as you go. A readable diagram lets the interviewer follow you, interrupt you at the right place and help you if you get lost.
Setting round assumptions and announcing them
Each branch of the tree ends with a numerical assumption. Three rules make them defensible.
Round to calculate fast. Take 500,000 inhabitants rather than 487,300, three euros rather than 2.85 euros, fifty weeks rather than fifty-two. The aim is to be able to do the calculation in your head and out loud. Round numbers are not laziness, they are an expected skill.
Anchor on what you know. To estimate the share of the population that buys a takeaway coffee, think of the people around you, your daily commute, an observable habit. Say where the assumption comes from: "From what I see around me, I would say roughly one adult in three." You do not have to cite a study, and inventing one would be worse than doing without.
Announce each assumption before using it. State it, justify it in one sentence, then offer to correct it if the interviewer has a better figure. Many interviewers supply a number when they judge yours too far from reality: take it without argument and carry on. Bain says as much in its teaching case: there is no "right" or "wrong" answer for assumptions, what counts is the quality of the reasoning and how reasonable what you set down is.
Finally, remember to spell out the scope: which geographic area, which period, exactly which product, in volume or in value. A framing question at the start avoids solving the wrong problem. Among its interview tips, Bain also recommends clarifying the objective and thinking out loud.
Checking the order of magnitude
This is the step candidates forget most often, and the one that most clearly sets strong profiles apart. A result with no check is just a number; a checked result becomes an estimate.
Three checks work almost every time.
- Triangulation through the other approach. Did you estimate from demand? Cross-check from supply, by asking how many outlets exist and what each one must sell to survive. If both estimates are in the same order of magnitude, you can conclude.
- The per-inhabitant or per-unit ratio. Divide the total by a familiar base: amount per inhabitant, volume per site, revenue per employee. A result implying that each inhabitant buys thirty times more than normal is an immediate red flag.
- The ceiling and the floor. Ask yourself whether the result exceeds what is physically or economically possible, or falls below what a single player already sells.
Bain illustrates this reflex in its teaching case: the volume needed to break even is compared with the estimated size of the market, which makes it possible to judge whether the goal is achievable. In other words, the estimate serves to answer a decision question, not to produce a number for its own sake. Adopt the same reflex: after your result, say what it implies for the client's question.
If the check reveals a large gap, do not hide it. Trace back through the calculation, identify the most fragile assumption, correct it and say why. Correcting out loud after a check is seen as a consultant's quality.
A worked example from start to finish
Here is a complete example, with fictitious assumptions chosen for how easy they are to calculate with. They are teaching parameters, not statistics: they describe no real market.
Question: what is the annual size of the takeaway coffee market in a French city of 500,000 inhabitants?
Framing: "I mean the annual revenue, in euros, from coffees sold to take away by businesses in the city, all locations combined. I start from demand and will cross-check from supply."
Step 1: demand
- Population: 500,000 inhabitants (fictitious assumption).
- Potential consumers, meaning teenagers, students, working people and younger retirees, excluding children and the very old: about 60%, or 300,000 people (fictitious assumption).
- Share who regularly buy a takeaway coffee: one third, or 100,000 people (fictitious assumption).
- Frequency: three coffees per week per person (fictitious assumption).
Calculation: 100,000 people times 3 coffees gives 300,000 coffees per week. Over 50 weeks, that makes 15 million coffees per year.
Step 2: value
- Average price of a takeaway coffee: 3 euros (fictitious assumption).
Calculation: 15 million coffees times 3 euros gives about 45 million euros per year.
Step 3: the supply cross-check
- Number of outlets selling takeaway coffee in the city: about 250, counting cafés, bakeries, chains and vending machines (fictitious assumption).
- Annual coffee revenue per outlet: 45 million divided by 250, or about 180,000 euros.
- Brought down to a day: 180,000 divided by 360 gives about 500 euros, or a little under 170 coffees at 3 euros per day per outlet.
Reading: an outlet serving 170 takeaway coffees a day is a plausible order of magnitude for an active business, neither ridiculous nor wild. The two routes converge, so the result of 45 million euros per year is consistent with the assumptions set. Had the check produced 20 coffees a day, you would have suspected a demand assumption that was too high; had it produced 2,000, an assumption that was too low on the number of outlets.
Step 4: the conclusion
Conclude in one sentence that goes back to the question: "Under these assumptions, the market is worth a few tens of millions of euros a year. The main sensitivity is the share of the population that buys regularly; if that share doubled, the result would double." By naming the most sensitive variable, you show where to refine first.
Where candidates lose points
- Diving into the calculation without a structure. The candidate states numbers in no particular order and can no longer explain where the total comes from.
- Multiplying overlapping segments, and counting the same people twice.
- Hiding assumptions. A number put down without explanation can be neither discussed nor corrected.
- Getting lost in precision. Spending two minutes on 2.85 euros rather than 3 euros is a poor trade-off.
- Forgetting the unit. Mixing volume and value, or weeks and days, is the most frequent mistake under pressure.
- Never checking. A result stated with no cross-check leaves the interviewer wanting more.
- Freezing in the face of silence. Think out loud; if you need thirty seconds, ask for them.
Practising and mentioning it in your application
A market sizing is practised like scales on an instrument. Take one question a day, give yourself four minutes, structure it in writing, announce your assumptions out loud, then check. Vary the topics: a consumer product, a B2B service, an infrastructure, a digital market. The firms publish practice cases themselves: McKinsey posts examples with suggested answers, and Bain offers several guided cases, including the coffee one already cited. Practising out loud with another student is useful, because the exercise is as much oral as numerical.
These skills can also appear in your CV, provided you prove them. A line such as "built the market size estimate for a student association project and validated it through two approaches" says more than a generic "analytical mind". As for layout, the ConseilCV tool takes care of the structure of the document and produces a PDF that applicant tracking software reads correctly, which leaves you time to work on the content. You can create your CV for free and place these quantified achievements in it, with figures you are able to explain in an interview.
Frequently asked questions
Should I always start from demand?
No. Demand is natural for a consumer good, supply for equipment or a B2B market. What matters is to choose, to say why, and then to cross-check through the other route when time allows.
Can I use a calculator or paper?
The rule depends on the firm and the format, including remote interviews. In most cases paper is allowed and a calculator is not: practise calculating in your head with rounded numbers. Ask what is permitted at the start of the interview rather than assuming.
What if I do not know an order of magnitude?
Say so and break it down. For example, estimate a purchase frequency from what you observe, or a capacity from the size of a venue. If the interviewer offers a number, use it without hesitation. Saying that a value is uncertain and explaining how you would refine it is better received than presenting it with unjustified confidence.
How close to reality does the result need to be?
It must be plausible and consistent. The recruiter expects a defensible order of magnitude, not an exact value: the criterion is that your reasoning holds and that the check confirms it.
How long should I spend on it in an interview?
Allow a few minutes for a standalone market sizing, and less if it is one step in a broader case. Framing, structuring and announcing your assumptions takes half the time; calculation and checking take the rest. Watch the clock and summarize your result before the interviewer presses you.
Sources
- Bain & Company, Case interview preparation 🔗: description of the case interview and preparation tips.
- Bain & Company, Coffee Shop Co. 🔗: teaching case with top-down and bottom-up approaches, assumptions and a consistency check.
- McKinsey & Company, Interviewing at McKinsey 🔗: presentation of the problem-solving interview and practice cases.