Technique

Market sizing: build the tree top-down, fill the numbers bottom-up

One method covers every market sizing question. Build the logic tree before touching a number, then populate it from anchors you already know. Worked example included, plus a daily drill.

On this page

• Step 1: build the tree before any number
• Step 2: fill the numbers bottom-up
• The worked example, out loud
• The anchors worth memorizing
• The classic traps
• Drill this
• Frequently asked questions

Every market sizing question, from "how many coffees does Berlin drink" to "size the European e-bike market," yields to the same method. Build the logic tree top-down before touching a single number. Then fill the numbers in bottom-up, from small anchors you actually know. Candidates who fail sizing questions almost always break this order. They grab at a big number first and try to reason backward from it.

The tree is the deliverable. Interviewers care far more about a clean, complete tree with rough numbers than a precise total hanging from shaky logic.

Step 1: build the tree before any number

Start from the demand side by default. Who wants this thing, how many of them are there, how often do they buy? A demand tree for "running shoes sold in Germany per year" looks like this in words. Population, filtered to people who run, times pairs bought per runner per year, plus the people who buy running shoes without running.

Say the whole tree out loud before calculating anything. This is the sizing version of explain before you calculate. It lets the interviewer bless the logic or redirect you while redirecting is still cheap.

Two structural rules keep the tree honest. Stay MECE. Segments must not overlap and must cover everyone. A common trap is a quiet mismatch between two readings of the question, say "e-bikes sold in Germany" versus "e-bikes bought by Germans." Cross-border orders and tourists sit in one but not the other. Pick one definition, say it, and stick to it. Only segment where it changes the answer. Splitting runners into ages is worth it if age changes purchase frequency. Splitting by shoe color is noise.

One variation to keep in your pocket: when the buyers are hard to count but the sellers are visible, build the tree from the supply side instead. Gyms, hospital beds, charging stations, delivery vans. Number of outlets times throughput per outlet is often a faster and safer tree than guessing at customer behavior. The rule stays the same, tree first. And when both sides are buildable, sizing one and cross-checking with the other is the strongest move available.

Step 2: fill the numbers bottom-up

Now, and only now, attach numbers, starting from the leaves of the tree. Use round anchors you can defend. Germany has roughly 80 million people. Maybe one in five does some running, call it 15 million. A regular runner goes through a bit more than a pair a year, a casual one much less.

Announce every assumption as an assumption and invite the correction: "I'll assume one in five, happy to adjust." Interviewers often carry a real figure and will trade it for a reasonable guess. Every assumption you state is a small check-in, which keeps the sizing a conversation instead of a lecture.

The worked example, out loud

Notice the two turns that make this strong. When challenged, the candidate held the assumption loosely and quantified what would change if it moved. And the closing cross-check did not perfectly agree with the answer, so the candidate said so instead of declaring victory. That honesty scores higher than a tidy number.

The anchors worth memorizing

You cannot derive everything live, so carry a small stock of round numbers. Your home market's population and households (Germany, about 80 million people and 40 million households). US population about 330 million at roughly 2.5 people per household, so about 130 million households. Europe around 450 million in the EU. Life expectancy near 80, which makes each one-year age cohort roughly population divided by 80. World population about 8 billion. Ten anchors cover the vast majority of sizing questions, and for a worldwide sizing it is usually cleaner to size one representative country and scale up than to whittle down from 8 billion.

The classic traps

False precision. An answer of "15.7 million" from assumptions that were themselves rounded says you do not understand your own error bars. Round the total to match the roughness of the inputs.

Double counting at segment borders. The regular runner who also buys a fashion pair sits in two branches. Decide where each person counts, say the decision out loud.

Forgetting the replacement cycle. Sizing durable goods by owners instead of by purchases per year inflates everything. The question is almost always the annual flow, not the installed base.

Pressing on when the number smells wrong. There is no single right answer in sizing, but there are wrong ones, and they are wrong by orders of magnitude. If your tree says Germans buy 200 million pairs of running shoes a year, do not present it and hope. Challenge your own assumption, find the broken branch, and fix it. Catching it yourself is worth more than never slipping.

Drill this

One sizing a day, eight minutes, out loud.

1. Pick any object in view. Coffee cups, e-scooters, umbrellas, dog leashes.
2. Minutes 1 to 3: build and speak the full tree. No numbers allowed yet.
3. Minutes 4 to 7: fill in numbers from your anchors, narrating every assumption.
4. Minute 8: sanity-check the total against a reference you know and say the so-what.

The no-numbers rule in the first three minutes is the whole drill. It forces the discipline the method is named for, and after two weeks of daily reps the tree-first reflex holds up under interview pressure.

Frequently asked questions

Interviewing somewhere specific?

Read next