Measuring Your Market Share When Nobody Knows the Size of the Market

A share is a ratio, and both halves have to be the same thing

Before any estimating there is a discipline question, and most bad market shares fail here rather than in the arithmetic. Four things have to match between numerator and denominator: the unit, meaning both are money or both are physical units and never one of each; the period, meaning the same twelve months and not your financial year against the statistics office's calendar year; the geography, meaning the territory you sell into and not the country you are registered in; and the product definition, meaning the same set of goods and services on both sides. Fail any one of the four and the ratio is not wrong by a little, it is meaningless.

The unit rule has a consequence people find surprising, and it is an exact identity rather than a rule of thumb: your value share divided by your unit share equals your average selling price divided by the market's average price. Sell 3,200 units into a market of 160,000 and your unit share is 2.0 %; if your average price is $94 against a market average of $47, your value share is 4.0 %. Neither number is wrong and neither is the real one. A premium brand always looks twice as big in money as in volume, and a discounter always looks half as big — so quoting whichever is flattering, without saying which, is the commonest small dishonesty in the whole subject.

Bottom-up: build a denominator you can defend, and give it a band

A bottom-up market is a chain of factors, each of which you can observe, ask about or bound. How many potential buyers are within reach? What proportion of them buy this category at all? How often in a year? At what average price? Multiply. The virtue of the method is not accuracy — it is that every factor is a claim someone can challenge and you can go and check. A top-down figure you copied from a report is a single number nobody in the room can argue with, which feels like strength and is the opposite.

The uncertainty must travel with the estimate, and there is a clean way to carry it. For a product of independent factors, the relative error of the result is approximately the square root of the sum of the squared relative errors of the factors. That is more forgiving than intuition suggests: three factors each known to plus or minus 20 % give a product known to plus or minus 34.6 %, not 60 %, because errors partly cancel. It is also less forgiving in one direction — every extra factor you multiply in adds to the total, so four factors at 25 % each take the whole estimate to 50 %. A longer chain looks more rigorous and is less certain.

Here is the whole method on one case. A catchment of 52,000 households; 18 % of them buy the category at all; those who do buy 2.4 times a year; the average ticket is $38. The market is 52,000 × 0.18 × 2.4 × 38 = $853,632. Put honest error bars on the four factors — 10 % on the household count from the census, 25 % on the penetration rate because it came from a small survey, 20 % on frequency, 5 % on price because you can see competitors' price lists — and the combined relative error is 33.9 %, so the market lies somewhere between $564,151 and $1,143,113. On your own sales of $142,000, the share is 16.6 % centrally, and between 12.4 % and 25.2 % across the band. That range is the finding. It is enough to tell you that you are a significant player and nowhere near enough to tell you whether you are the leader.

Top-down: what the statistics office will and will not give you

There is a free denominator available and most small businesses never look at it. European structural business statistics publish, per country and per activity, the number of active enterprises, turnover and related indicators, value added and related indicators, and the number of employees and self-employed persons. National statistical institutes go further than the European aggregate: many publish at four-digit activity level or at an even finer national breakdown. In France the source is Ésane, which combines administrative tax and social declarations with surveys rather than asking firms twice; other countries build theirs the same way, on registers rather than on questionnaires.

Three limits are worth knowing before you build anything on it. The coverage is not universal: European structural business statistics run over activity sections B to N and P to R plus two divisions of section S, which leaves agriculture and public administration outside, so some activities have no official denominator at all. The classification changed: the current version of the European activity classification was set by a delegated regulation of 2023 and applies to European statistics from reference year 2025, which means a series break sitting in the middle of any multi-year comparison you make. And most importantly, a class of the classification is an activity, not a product. Firms are classified by their main activity, so a class contains the whole turnover of firms that mostly do this and some of whom also do other things, and excludes the turnover of firms that do a little of it as a sideline.

Served or total: the distinction that changes the answer by a factor of a hundred

The total market is everything the category sells anywhere you could conceivably compete. The served market is what you can actually reach, quote for and deliver to at your current prices with your current capability. Take a firm doing $1.2 million a year. Against a national activity class worth $3.0 billion its share is 0.04 %. Against the $24 million it can genuinely reach — its region, its segment, the customer sizes it can serve — its share is 5.0 %. Same firm, same year, same invoices, and a factor of 125 between the two answers.

Neither is the true one; they answer different questions. The total figure is the right denominator for asking whether the category itself is growing or shrinking, because it is the only one whose definition does not move when your delivery radius or your product line moves. The served figure is the right denominator for every operating decision — whether to hire another salesperson, whether the pipeline is realistic, what a further point of share would cost. Using the total where the served belongs produces the most expensive error in the subject: a 0.04 % share tells a management team that the market is effectively infinite and any target is reachable, when the truth is that they already hold a twentieth of what they can reach and the next customer will have to be taken from a named competitor.

Why a wrong denominator is worse than no number

A missing number makes people cautious. A wrong number makes them confident, and confidence is what turns an estimate into a decision. The cleanest demonstration is the growth decomposition, because your share change is not your sales change — it is your sales growth divided by the market's growth, minus one. Sales up 12 % in a market you believe grew 8 %: share up 3.70 %, and the sales team gets a bonus for taking ground. The same 12 % in a market that actually grew 15 %: share down 2.61 %, and the right conversation is why competitors grew faster. Identical performance, opposite conclusion, and the only thing that changed was a market growth rate nobody measured.

The same arithmetic rescues a case that looks like failure. Sales down 3 % in a market down 9 % is a share gain of 6.59 % — a good year that reads as a bad one on the sales line alone. This is why the denominator has to be measured in falling markets as well as rising ones, and why the single most useful discipline is never to change the market definition and the reporting period in the same revision. Change one at a time, or you will never know which of them moved the share.

Three checks that cost nothing

First, the sum. Every share in a market must add to 100 % or less. Ask the five firms in your category what share each believes it has and you will routinely be told 20 % five times; add the forty smaller ones at even half a per cent each and the total is 120 %. That check finds a bad denominator in thirty seconds and it works even when you have no idea what the right one is, because it does not need the right one — it only needs the claims to be mutually possible.

Second, triangulation. Build the market bottom-up and separately from the top down, and require the two to overlap within the band you computed. If the bottom-up says between 564,000 and 1,143,000 and the top-down says 4 million, one of them is measuring something else — usually the top-down, because the activity class contains adjacent products or a wider geography. That failure is informative: it tells you exactly how much of the official class is not your market.

Third, concentration. Square every share and add them up. On a served market where the five players hold 38, 24, 17, 12 and 9 per cent, that gives 2,534, and dividing 10,000 by it gives 3.95 — the number of equal-sized competitors that would produce the same concentration. Check that against the number of firms you actually lose deals to. If the index says four and you lose to eleven, your denominator is too small; if it says four and you only ever meet two, it is too big. This check is also the sharpest way to see what a bad denominator does: put the same five firms on a denominator 125 times larger and the index falls from 2,534 to 0.16, which would describe an atomised market with no one in charge — the exact opposite of the truth.

Frequently asked questions

What if I simply cannot estimate the market at all?

Then say so, and measure a partial share you can actually observe instead. Share of a distribution channel whose total throughput you can see. Share of the tenders you were invited to. Share of a named account list you keep yourself. Share of the shelf in the stores you visit. None of these is the market share, and each of them can be computed exactly, tracked over time and defended in a meeting. A precise partial share that you update every quarter is worth far more than a total share you invented once and now quote from memory.

Should I measure share in money or in units?

Whichever matches the decision, and say which one you used. Money is right for anything about revenue, capacity and profit. Units are right for anything about production, logistics and the number of customer relationships you have to service. The two are linked by an exact identity: value share divided by unit share equals your average price divided by the market average. If you sell at twice the market's average price, your value share is exactly twice your unit share — 2.0 % in units becomes 4.0 % in money — and neither figure is more honest than the other. What is dishonest is quoting the flattering one without naming it.

How wrong is a bottom-up estimate likely to be?

Compute it rather than guessing. For a product of independent factors the relative error is about the square root of the sum of the squared relative errors. Three factors each good to plus or minus 20 % give plus or minus 34.6 %; three at 10 % give 17.3 %; two at 15 % give 21.2 %; four at 25 % give a full 50 %. Two lessons fall out. Errors do not simply add, so a chain of rough factors is more useful than it feels. And every additional factor makes it worse, so resist the urge to refine the model by multiplying in a fifth term — go and measure the worst existing factor instead. In the worked example the penetration rate carried 25 % of the 33.9 % total; one small survey would do more for the estimate than any amount of extra structure.

Can I just use my trade association's market figure?

You can, provided you write down two things beside it: which firms it covers and how it was collected. Association figures are usually built from members' returns, so the denominator is a members' market — it excludes non-members, and in a fragmented trade that can be most of the volume. That is not a reason to reject the number; it is a reason to label it. A share computed on a members' market is a perfectly good measure of your position among members, and it becomes misleading only when it is presented as your position in the market. The same caution applies to any figure whose method you cannot describe in one sentence.

How often should I recompute my market share?

Once a year for the number itself, and never in the same revision as a change of definition. The reason is that share movements and definition movements are indistinguishable after the fact: if you widen the territory and recompute in the same quarter, you will never know whether the share fell because you lost ground or because the denominator grew. Change the definition in one year, hold it fixed the next, and note in the file which year the break sits in. The official statistics have the same problem and handle it the same way — the European activity classification changed version for reference year 2025, which puts a break in the middle of any series that spans it.