How to size a market without buying a market-size report

Someone asks how big your market is. You search, and you find a headline: the global category is worth some enormous figure and growing at some confident percentage a year. It is behind a paywall, but the abstract gives you enough to put a number on a slide.

The number has two problems. The first is that you cannot check it — you are not shown how it was built, so you cannot tell whether it is a survey, a model, or a previous report restated. The second is worse: it is almost certainly not about your market. "Global category X" bundles enterprise and consumer, every geography, and adjacent products you do not sell, into a total that would not change if your specific opportunity did not exist at all.

A number you built yourself from public counts will be rougher. It will also be defensible, arguable in its parts, and actually about the business you are deciding on.

Start from the decision, not the market

Before sizing anything, write down the decision the number is for. Almost always it is one of three:

• Is this worth doing at all? You need to know whether the reachable market can support the revenue you need — not what the global category is worth.
• Which of these two markets is bigger? You need a consistent method applied to both. Absolute accuracy matters much less than treating them the same way.
• Someone external requires a number. An investor or a board wants a figure. Here the method matters more than the total, because the first question you will be asked is how you got it.

The first two rarely need a market size at all. They need a buyer count and a plausible price, which is a much easier thing to get right.

Build it bottom-up: count, price, repeat

A bottom-up estimate multiplies three things you can source separately:

How many potential buyers exist in the segment and geography you can actually serve, what they pay per purchase, and how often they buy. Each factor gets its own evidence and its own uncertainty. When you present the result, you present the three inputs, not just the product — because that is what lets someone disagree usefully with you rather than just disbelieving the total.

Counting the buyers

This is the part people assume is impossible and is usually the easiest. Depending on the market:

• Company registers let you filter by activity classification and location. Coverage and searchability vary by jurisdiction, so check what that register actually exposes rather than assuming.
• Licence and regulator registers are close to a census in licensed trades — if your buyer must hold a licence to operate, the register is the buyer list.
• Trade association and professional body directories capture a real but partial share of a sector. Useful as a floor, not a total.
• Map and local-listing data works well for physical, local businesses. Counting them across a defined area gives a hard, repeatable number.
• Procurement and tender portals enumerate public-sector buyers, and often publish what they actually spent.
• Official statistics agencies publish business counts by sector and size band. This is frequently the single best source and it is free.

Whatever you use, record the date, the filter you applied, and the source. A count without those three is not reproducible, including by you in six months.

Pricing the transaction

You need what a buyer pays, not what you hope to charge. Published pricing pages are the obvious source. Where pricing is quote-based, look for published contract awards and tender results, rate cards, price lists in trade publications, and archived versions of pricing pages that show the direction of travel. In many sectors, a handful of real published contract values tells you more about the going rate than any survey would.

If prices vary widely, do not average them into a single fake number. Split the population into tiers and size each one. A market with a large low-paying tail and a small high-paying head behaves nothing like its average suggests, and the difference usually decides your strategy.

Frequency and replacement cycle

The factor most often skipped. A one-off purchase every seven years and a monthly subscription produce wildly different annual markets from the same buyer count. If you cannot source a frequency, say so and present the number as "per purchase" rather than inventing an annual figure.

Triangulate: build it a second way

One bottom-up estimate is a guess with sources attached. Two independent estimates that land near each other is evidence.

The second path should use different inputs, not the same ones rearranged. If your first estimate was buyers × price, try sizing the supply side instead: count the businesses currently serving this market, estimate what a typical one turns over from staff counts or published accounts where they exist, and total it. Or work from an adjacent measurable: volume of a complementary product, publicly reported spend by a segment of buyers, or the size of a comparable market in another region scaled by an official population or business count.

If the two estimates land within the same order of magnitude, you have something usable, and the gap between them is your honest uncertainty range. If they are wildly apart, do not average them — one of your assumptions is wrong, and finding out which one is the most valuable hour of the whole exercise.

Size the market you can reach, then the one you can win

The total is the least useful of the three numbers you should produce.

Narrow from all potential buyers to the ones you can actually serve — right geography, right segment, right size band, no disqualifying requirement you cannot meet. Then narrow again to the ones you could realistically reach and convert in the period you are planning for, given the channels you actually have. That last number is the one your plan lives or dies on, and it is usually a small fraction of the first.

This is where the "we only need one percent of the market" line comes from, and why it is treated as a red flag. One percent is not a plan; it is an admission that no reachable path was worked out. The bottom-up version is the opposite: here are the buyers, here is how I reach them, here is what that adds up to.

Work backwards from your own number

For a go/no-go decision, invert the whole exercise. Take the revenue you need, divide by realistic revenue per customer, and you have the customer count you must win. Then ask a single question against your counted population: is that count a plausible share of the buyers I can reach?

The answer is often immediately obvious, and it is a far better decision input than a category total. Needing a modest number of customers from a large, well-defined, reachable population is a workable plan. Needing an implausible share of everyone who exists is a no — and you got there without buying anything.

Where these estimates go wrong

• Double counting. The same business appears in a register, a directory and a listing set. Deduplicate before totalling, or your count inflates quietly.
• Borrowing a global figure for a local decision. Scaling a worldwide total down by population share assumes a uniformity that rarely holds.
• Growth rates with no parent. A projected growth percentage that traces back only to another report is not a fact you can build on. If you cannot find how it was derived, leave it out.
• Averaging a bimodal market. As above: tier it instead.
• Counting buyers who cannot buy. If a large share of the population is locked into long contracts, served by an incumbent they own, or below the size where your product makes sense, they are not in your reachable market.
• Precision theatre. Reporting an estimate to the nearest thousand implies an accuracy the method does not have. Round to the confidence you genuinely have and give a range.

Present it as a model, not a number

Write the estimate as its inputs: the count and where it came from, the price and where it came from, the frequency, the narrowing steps, and the second estimate you checked it against. Anyone reading it can then attack a specific assumption instead of the conclusion — which is the point. When a real figure arrives later, you update one input and the total moves, rather than starting over.

That is also the difference between an estimate and a guess. A guess is a number. An estimate is a number plus the reasoning that would let you notice it was wrong.

The short version

• Decide what the number is for first — most decisions need reachable buyers, not category totals.
• Build bottom-up: buyer count × price per purchase × frequency, each separately sourced and dated.
• Count buyers from registers, licence lists, official business statistics, directories, map data and tender portals.
• Price from published pricing, contract awards and rate cards; tier a market rather than averaging it.
• Build a second estimate a different way; the gap between them is your uncertainty.
• Narrow to reachable, then to winnable. "One percent of the market" is not a plan.
• Invert it: customers needed to hit your number, tested against the counted population.
• Publish the inputs, not just the total, and round to the confidence you actually have.