PartnerRound- 01ClarifyProduct, place, period, metric
- 02Choose approachTop-down or bottom-up
- 03SegmentSplit where behaviour differs
- 04EstimateRound numbers, one at a time
- 05Sanity-checkTest against a second view
By the end you can
- Turn any 'how big is this market' question into a clear structure within 60 seconds
- Choose top-down or bottom-up, and demand-side or supply-side, for a given question
- Segment where behaviour really differs and state assumptions the interviewer can follow
- Sanity-check and triangulate an estimate before you present it
On this page
01What is tested, and the five-step method
A market sizing question, or guesstimate, asks you to estimate a number nobody expects you to know: how much the UK spends on electric toothbrushes, how many coffee shops a city can support. It comes as a standalone question or inside a case, as in 'is this market big enough to enter?'
Nobody is grading the final number. The interviewer is grading whether your structure is logical, your assumptions sensible and stated, your arithmetic clean, and whether you notice when an answer looks wrong. A sound approach that lands 30% off beats a lucky guess.
Use the same five steps every time. The diagram at the top of this lesson is the whole method. Here is what each step involves.
1. Clarify scope
Pin down the product (electric only, or all toothbrushes? handles and replacement heads?), the geography, the period (annual is the default) and the metric (units or revenue, retail or manufacturer prices). Two or three questions, not ten.
2. Choose the approach
Top-down starts from a big number you know, usually population or households, and narrows it. Bottom-up starts from one unit you can picture, a shop, a machine, a customer, and scales it up. Pick the one whose inputs you can estimate with most confidence.
3. Segment
Split the population only where behaviour differs enough to change the answer. Two to four segments is the sweet spot.
4. Estimate with round numbers
Write the formula, then fill each box with a round, defensible number: 70m, not 68.3m; 50%, not 47%. Do the maths on paper and label the units.
5. Sanity-check
Check the answer per person, per household or per outlet, or against a second method. If it fails, find the assumption that is off. Do not just shave the total.
02Top-down vs bottom-up, demand vs supply
Start big and narrow down. Best for products bought by a share of all people.
Start from one unit you can picture and scale it up.
Demand side vs supply side
Demand-side estimates count buyers and what they buy: people × usage × price. Supply-side estimates count sellers and what they can deliver: outlets × capacity × utilisation. Top-down sizing is usually demand-side; bottom-up can be either.
Size whichever side is the bottleneck. Daily flights from a busy airport are capped by runway slots, not by how many people want to fly. Most consumer goods are capped by demand, so size the demand.
Check yourself
Which of these is best sized from the supply side?
03Segment where behaviour differs
Segmentation is where judgement shows. Only split when segments differ on a number in your formula: penetration, frequency or price. Splitting UK adults by star sign adds work and no accuracy.
For B2B markets, swap people for firms: businesses by size band × share that buy × units or spend per business. Always split by size, because a handful of large firms can outbuy thousands of small ones.
Large firms are 1% of businesses but over a third of the machines (10,000 of 28,000). If a machine lasts about seven years, the region buys 28,000 ÷ 7 = 4,000 a year. An unsegmented average would have hidden where the market really sits.
04Triangulate, sanity-check, conclude
One method gives you an estimate. Two methods that agree give you confidence. Triangulating means sizing the same thing a second way, ideally from the other side (supply if you started with demand), and seeing whether the two land within about 25% of each other.
If they disagree, that is useful: it tells you which assumption to revisit. Say so out loud rather than quietly averaging the two.
- Per person or per household: does the spend per user a year feel right?
- Per outlet: does each shop serve a believable number of customers an hour?
- Against the economy: a single consumer product cannot be a big slice of household spending.
- Against the client: if the client's revenue is bigger than your market, one of them is wrong.
Land the answer
Close in three moves: the number as a range, the assumption that drives it, and what it means for the client. For example: 'About £800m to £850m a year. Penetration drives it most. For the client, 5% share would be roughly £40m of revenue a year, enough to justify a launch.'
A range signals judgement, not doubt. Naming the driver shows you know where the estimate is fragile, and the so-what turns a maths exercise back into a business answer.
Check yourself
You estimate annual UK toothpaste sales at £5bn. Which check exposes the problem fastest?
05Stating assumptions out loud
Every number you use is given, known or assumed. Say which. An assumption stated with a reason is a judgement; one slipped in silently is a guess. The reason can be short: an observation, a logical bound or an anchor number.
- Give the number and the reason in one breath: 'Say half of adults use an electric brush. They're common, but price keeps some on manual.'
- Round on purpose and say so: 'I'll use 70m to keep the maths clean.'
- Invite challenge on the big drivers only: 'Penetration matters most here. Does 50% sound reasonable?'
- Keep a list of assumptions on your page so you can revisit them in the sanity check.
Weak
Let's say 50% of people have an electric toothbrush.
Strong
Of the 56m adults, I'll assume half use an electric brush. They're common now, but a good one costs £50 or more, which keeps some people on manual. Children will be lower, say a quarter, because parents often buy cheaper manual brushes for small children. Penetration moves the answer most, so I'm happy to adjust it.
Why: The strong version segments before assuming, gives a one-line reason for each number and flags the assumption that drives the answer. The interviewer can push on the right number instead of doubting the whole estimate.
06Worked estimate 1: top-down
Worked example
Annual UK electric toothbrush sales
Question: estimate annual revenue from electric toothbrush sales in the UK. Scope agreed: retail value in pounds, handles and replacement heads, one year.
Structure
Revenue = handle sales + head sales. Users = people in the segment × electric penetration. Segment by age, because children use cheaper brushes and replace them faster.
Assumptions
Maths
- Adult handles: 28m ÷ 4 = 7m a year × £60 = £420m.
- Child handles: 3.5m ÷ 2 = 1.75m a year × £20 = £35m.
- Heads: (28m + 3.5m) = 31.5m users × 3 heads × £4 = £378m.
- Total: £420m + £35m + £378m = £833m. Call it £800m to £850m.
Sanity check
Per user: £833m ÷ 31.5m ≈ £26 a year. An adult spends £15 a year on the handle (£60 over four years) plus £12 on heads, so £27. That feels right for someone with a mid-range brush. Per household: £833m ÷ 28m ≈ £30 a year, also plausible.
So what: a market of under £1bn, with heads worth nearly half of it (£378m of £833m). A brand whose handles only take its own heads locks in the recurring half.
07Worked estimate 2: bottom-up
Worked example
Coffee shops a city of 1m people can support
Question: how many coffee shops can a city of 1m people support? Scope agreed: chains and independents, sit-in or takeaway, ignoring commuters and tourists from outside the city.
Structure
Shops supported = cups bought a week ÷ cups one shop must sell a week to break even. Demand first, then the economics of one shop.
Demand: cups a week
Supply: one shop's breakeven
- Weekly fixed costs: rent $1,500 + staff $3,000 + other $500 = $5,000.
- Contribution per cup: price $3.50 - beans, milk and cup $1.00 = $2.50.
- Breakeven: $5,000 ÷ $2.50 = 2,000 cups a week, about 290 a day over seven days.
Answer
1m cups a week ÷ 2,000 cups per shop = about 500 shops, or one for every 2,000 residents.
Triangulate from the supply side
Picture the city instead. About 100 neighbourhoods of 10,000 people, each with a high street of around four coffee shops, gives 400. Add the centre, stations and campuses, say 150 more: about 550. The two methods are within 10% of each other, so 500 to 550 is a confident range.
Sanity check
290 cups across a 12-hour day is about 24 an hour, one every two and a half minutes: steady but not frantic, which is what breakeven should look like. One cup per resident a week, with half the city never buying, also feels right.
So what: the city supports about 500 shops at breakeven volume. If it already has 600, the market is saturated, and a new entrant must win on location or quality, not on market growth. Food sales would lower each shop's breakeven, so treat 500 as a floor if time allows a refinement.
08Common mistakes and anchor numbers
- Skipping scope: sizing all toothbrushes when the question was electric, or units when the client wanted revenue.
- False precision: 68.3m people and 47% penetration add effort, not accuracy.
- Over-segmenting: six segments with made-up splits multiply the guesswork.
- Mixing time units: daily cups with annual prices, or monthly users with weekly frequency.
- Confusing stock and flow: 31.5m people own an electric brush, but only about 9m handles are bought a year.
- No so-what: ending on a number instead of what it means for the client.
Check yourself
30m people in a country own a bike, and a bike lasts about 6 years. Roughly how many bikes are sold there each year?
Cheat sheet
The lines worth rereading the night before.
- 1Clarify scope first: product, geography, period, and units or revenue.
- 2Top-down for mass consumer products: population × % who buy × units per buyer × price.
- 3Bottom-up for local, B2B or capacity-limited markets: outlets × capacity × utilisation × price.
- 4Segment only where penetration, frequency or price really differs. Two to four segments.
- 5State every assumption with a one-line reason, and flag the one that drives the answer.
- 6Annual purchases = installed base ÷ lifespan. Do not confuse stock with flow.
- 7Sanity-check per person, per household or per outlet, and triangulate with a second method.
- 8End on the so-what, not the number.
Now drill it
Reading builds the model; drills make it automatic. Start with the first topic, easiest drills first.
Read the full guide
The same ground on the blog, with more worked examples and practice on the page.
Questions candidates ask
How accurate does a market sizing answer need to be?
Not very. Interviewers know you cannot know the real figure, so they score the structure, the assumptions and the arithmetic. Aim for the right order of magnitude with logic they can follow, and say which assumption you are least sure of.
Should I use top-down or bottom-up market sizing?
Use top-down for mass consumer products where you can estimate the share of people who buy. Use bottom-up for local markets, B2B products and anything capped by capacity, such as clinics or shops. If time allows, run the other method as a quick cross-check.
Can I ask the interviewer for numbers in a market sizing question?
You can ask, and some interviewers will give you a population or a price. Many will ask you to estimate instead, so have anchor numbers ready. Either way, ask during scoping, not halfway through the maths.
What is the difference between TAM, SAM and SOM?
TAM, the total addressable market, is everything the product could sell if everyone who could buy did. SAM, the serviceable available market, is the part your business model and geography can reach. SOM, the serviceable obtainable market, is the share you can realistically win. Most case questions want the total market unless they say otherwise.
How long should a market sizing question take?
As a rough guide: about a minute to clarify and structure, three to five minutes of estimates and maths, and a minute to sanity-check and conclude. If you are well past ten minutes, you have probably over-segmented or lost the thread.