Seven Common Mistakes in Market Sizing (and How to Fix Them)

Market sizing rarely fails loudly. It fails quietly: a number that looks plausible slips into a slide, then into a hiring plan, and two years later a team is chasing a market that was never there. The strange part is how seldom the arithmetic is at fault. Most bad market sizes are built by capable people from a spreadsheet they trusted, and the damage comes from a few early decisions about definitions, boundaries and assumptions that nobody went back to check. The seven mistakes below cover most of what goes wrong, and each one comes with the correction that usually fixes it.

1. The market means five different things

Ask five people in the same company to define "the European logistics software market" and you will get five markets. One is counting warehouse management. Another quietly included customs filing because a client once asked about it. Someone is thinking about vendor revenue; someone else is thinking about what shippers spend. Both numbers can be correct and still be useless together, because they answer different questions.

Six questions settle most of this, and they are worth writing down before anyone opens a spreadsheet:

• Who is the buyer — the end customer, an intermediary, or both counted separately?
• What exactly is purchased — software licences, implementation, support, hardware, financing?
• Which budget line it sits on — IT, operations, marketing, compliance?
• Which geography — where the buyer is based, or where the spend is booked?
• Which year and currency — and nominal or real figures?
• Revenue or units — value counted at the vendor, at the channel, or volume of transactions?

A useful test: hand the definition to a colleague who has not seen the model and ask them to rebuild the top line from it alone. If they arrive somewhere materially different, the definition is doing less work than you think.

2. Double counting, the invisible error

This one survives review because the total still looks believable. It usually creeps in through three doors.

• Nested segments. "Cloud software" and "software for small businesses" overlap heavily, but both get added to a total as if they were separate rooms in the same house.
• Layers of the value chain. The distributor's revenue and the manufacturer's revenue describe the same product reaching the same customer. Adding them inflates the market without anyone noticing, because each line is individually sourced.
• Channels and end customers. A sale through a reseller gets counted once as a reseller transaction and again as an end-customer purchase.

The correction is structural, not analytical: build a hierarchy in which every euro of spend lands in exactly one bucket. Sum only the leaf nodes, never a mix of parents and children. Add a visible "sum check" row to the model that flags any total that does not reconcile with the last version, and note in the definition whether you are counting spend once at the point of purchase or tracking value along the chain.

3. Assumptions with no paper trail

"Twelve percent penetration by year three" sounds like a forecast. Usually it is a memory of a competitor's slide, a round number that felt ambitious but safe, or worse, the number that made the plan work. Six months later nobody can say where it came from, so nobody can challenge it.

Every assumption in a market model needs four things attached to it: a value, a source or a line of reasoning, a confidence level, and an owner. That does not mean every input needs a research report behind it; it means the reader can tell the difference between "measured", "inferred from two data points" and "placeholder, please replace".

Growth deserves its own warning. A category that grew quickly while it was small will not keep that rate once it is large, and a rate lifted from a global figure rarely survives contact with one country's regulation, language and purchasing habits.

If you cannot name the assumption that would halve your answer, the model is not finished — you have just stopped working on it.

4. Confusing a large market with a reachable one

The classic line is "the global category is enormous; we only need half a percent." It is a comfortable sentence that hides the hard part, which is that the half a percent has to be reachable by a specific sales motion, a specific price point and a specific product.

Separate three numbers explicitly. The total market is everything that could conceivably be bought by anyone, anywhere. The serviceable market is the slice your product, geography, language, pricing and compliance actually fit. The obtainable market is what your go-to-market can touch in the planning period — accounts you can identify, reach and close with the team you have. Only the third one belongs in a revenue plan, and it should be expressed as a share of the second, not of the first.

A quick sanity check: count the named accounts you could genuinely sell to next year. If that list is short, the grand total is decoration.

5. Unit and arithmetic slips that reach the boardroom

These are unglamorous and surprisingly common. A per-site spend that is monthly in one tab and annual in another. Users mistaken for accounts, or sites for companies. Nominal growth compared with real growth. Calendar years against fiscal years. A global figure converted to a country by population share when the category is clearly concentrated in a handful of metro areas.

The fix is mechanical: label every cell with its unit, build a dimensional check into the summary tab, and recompute one line independently by hand. If your market model is large, ask someone who did not build it to reproduce a single segment from scratch. Two people arriving at roughly the same number by different routes is worth more than any amount of formatting.

6. The single-number answer

A figure like "€1.42bn by 2029" signals precision the underlying work almost never supports. Worse, it ends the conversation: there is nothing to argue with, so nothing gets tested. A range forces the team to discuss which driver matters, and that discussion is usually more valuable than the number itself.

1. Pick the three assumptions with the greatest influence on the answer.
2. Give each a low, base and high value, with a short reason for the bounds.
3. Run the combinations and present three coherent scenarios, not nine disconnected cells.
4. State which scenario you are planning against and why.
5. Write one sentence on what evidence would move you to a different scenario.

All three scenarios should be defensible. A "low" case that assumes the category shrinks by a fifth is not a scenario, it is a straw man designed to make the base case look prudent.

Before the number leaves the room

Five habits catch most of these errors. Put the market definition at the top of the document, not in an appendix. Keep a visible sum check so double counting cannot hide. Maintain an assumptions register with sources, confidence and owners. Publish a range with a named scenario instead of a single point. And write down the decision the number is meant to support, because a market size that changes nothing is a hobby, not analysis.

Being roughly right in a way everyone can inspect beats being precise in a way nobody can follow. The models that survive scrutiny are rarely the most elaborate ones; they are the ones where the boundaries are stated, the assumptions are traceable, and the answer arrives with its own uncertainty attached.

Photo: Bia Limova / Pexels