How to size a niche market when there is no analyst report

Every deal team runs into the same wall. The target sells industrial sealing systems for cryogenic applications, or compliance software for municipal utilities, or refurbished dental imaging equipment. You need a market size for the investment memo, you search for a report, and there is nothing. No Gartner quadrant, no Statista figure, no association study. The niche is real, the revenue is real, the report does not exist.

The reflex is to reach for the nearest bigger number and cut it down with an assumed share. That produces a figure that looks precise and falls apart under the first serious question in the investment committee. The alternative is to build the market from the bottom up, out of the actual companies that operate in it. It takes more discipline, and it gives you a number you can defend line by line.

In this article you'll learn:

• Why top-down desk research systematically breaks down below a certain market size, and how to recognise it before you build a model on it
• How to define a niche by what companies actually do rather than by the industry code they happen to carry
• The bottom-up, company-level sequence that turns a company population into a defensible market size
• How to triangulate two independent estimates so the result holds up against real demand data
• What to write down so the number survives investment committee questions and later audit or valuation review
• Where on-demand niche market reports replace the multi-week research project

Why top-down desk research breaks down in niche markets

Top-down sizing works when someone else has already done the segmentation for you. You take a published market figure, apply a regional share, apply a segment share, and you have a number. For broad categories with well-funded analyst coverage, that is a reasonable starting point.

Niches are different for a structural reason: nobody has monetised research on them. Published market studies exist where there is a paying audience of vendors and investors large enough to justify the study. A market with 60 relevant suppliers and no listed pure play does not clear that bar. So either no report exists, or a report exists whose market definition is so broad that your target's actual competitive arena is 2% of it.

That is where the damage happens. Two assumed shares stacked on a broad figure, say 15% regional share times 8% segment share, turn a rounding error at the top into a factor-of-two error at the bottom. The model still looks quantitative. It just is not evidence.

How to tell you are in bottom-up territory

Three signals, any one of which is enough:

• The published market definition does not match the way buyers actually buy, so the boundary of the market has to be redrawn by hand.
• You can plausibly name the relevant suppliers in the low hundreds or fewer. If you can enumerate the market, you should count it, not estimate it.
• The available figures disagree by more than a factor of two. That is not a range, it is an absence of data.

The bottom-up method in one sentence

Identify every company that serves the niche, attach revenue to each, decide how much of that revenue actually belongs inside your market definition, then add it up and test the result against something independent.

The work is in the details of each step, which is where most sizing exercises quietly go wrong.

Step 1: define the niche by activity, not by industry code

Industry classifications were built for statistical reporting, not for deal work. NACE, SIC and NAICS codes are assigned at company level, they lag business model changes by years, and diversified companies carry a single code for everything they do. A specialist in cryogenic sealing and a generic gasket manufacturer can sit in the same code.

Write the market definition as a set of testable criteria instead:

• What the product or service does, in functional terms
• Which customer problem it is bought for, and by which buyer function
• The delivery model, since equipment, consumables and service revenue behave differently
• The geographic boundary, based on where customers actually buy from
• The explicit exclusions, which matter as much as the inclusions

Write down what is out of scope and why. When someone in the investment committee argues that the market is bigger than you claim, the exclusion list is the answer.

Step 2: build the complete population of companies

This is the step that decides whether the exercise is credible. A bottom-up size built on the 20 companies you already knew about is wrong by construction, and it is wrong in a predictable direction, because the companies you already knew about are the large and visible ones.

Completeness comes from screening a broad company universe against the activity criteria from step 1 rather than from starting with a shortlist. In practice that means searching descriptions of what companies actually do, their products, their own website language, across a wide base rather than filtering a narrow one. Company-level databases that combine financial data with business-activity descriptions across many countries make this tractable, which is what the StrategyBridgeAI company database is built for: around 50 million companies across more than 100 countries, screened by what the business does rather than by its classification code.

Two checks before you move on:

• Take five companies you know belong in the market and confirm the screen finds them. If it misses one, the criteria are too narrow.
• Take five companies the screen returned that you do not recognise and read what they do. If two do not belong, the criteria are too loose.

Step 3: attach revenue, and be explicit about what is estimated

For each company in the population you need a revenue figure and a note on where it came from. Disclosure quality varies by country and legal form, so expect a mix:

Never let an estimated figure enter the model unmarked. The share of estimated revenue in your total is itself a finding: if 60% of the market is estimated from employee ratios, that belongs in the memo, not in a footnote nobody reads.

Two mechanical errors to avoid: double counting a subsidiary and its parent, and mixing financial years. Both are common and both inflate the result.

Step 4: apply a relevance share per company, not one share overall

Almost no company in a niche earns 100% of its revenue there. The generic gasket manufacturer might have 4% of revenue in cryogenic applications, the specialist 95%. Applying a single blended share to the whole population destroys exactly the precision you built in step 2.

Assign each company a relevance share, taken from segment reporting where it exists, from product portfolio and customer references where it does not, and record the basis. Group the population into tiers, pure plays, strong specialists, occasional suppliers, so a reviewer can see where the volume actually sits and challenge one tier without discarding the whole model.

Step 5: triangulate against an independent estimate

A single method is a hypothesis. Build a second estimate that does not share the first one's inputs, then compare:

• Demand side: number of buying units times average annual spend, sized from the installed base or from a proxy such as production capacity or regulated asset counts.
• Adjacent anchors: import and export statistics for the relevant product group, procurement or tender volumes, regulatory filings that scale with usage.
• The target itself: if your bottom-up market implies the target holds 40% of it, and management believes they hold 10%, one of the two is wrong and it is worth knowing which before signing.

A spread within roughly 20% between two independent methods is a usable result. A factor of two means a definition problem, usually in the boundary you set in step 1. Fix the definition rather than averaging the two numbers.

The number you can defend is worth more than the number you like. A defensible market size is one where every input has a stated source and every assumption has a name attached.

Step 6: document so it survives IC, and later review

Bottom-up sizing carries assumptions. That is not a weakness, as long as they are visible. A sizing that cannot be reconstructed six months later is not evidence, it is an opinion with a decimal point, and that becomes a real problem when the same figure is reused in a valuation, a fairness opinion, or an impairment test that an auditor will review.

Keep an assumption log next to the model:

Run the sensitivity before the meeting. If the strategic conclusion holds across the range, say so and the discussion moves on. If it flips, that is the actual finding and the committee needs it.

Top-down and bottom-up compared

The by-product row is worth pausing on. A completed bottom-up sizing is a structured map of every relevant company in the niche, with revenue and specialisation attached. That is the same artefact a buy-side longlist requires. Teams that treat market sizing and target screening as one exercise get two deliverables from one piece of work.

Where on-demand niche market reports fit

The reason niches get sized top-down is not that practitioners prefer bad method. It is that the good method used to cost several analyst weeks per market, and deal timelines do not have several weeks. So the shortcut gets taken, and the number goes into the memo with a caveat nobody reads.

That constraint is what changes when the company population, the financial data and the activity descriptions sit in one place and can be screened on demand. On-demand niche market reports produce the market definition, the supplier population, the size estimate and the competitive structure for a specifically defined niche in hours rather than weeks, built on verified, documented sources so every figure stays traceable to the companies behind it. Outside-In business analysis then goes one level deeper on individual companies in that population, which is what you need once the market view turns into a target discussion.

The method does not change. The sequence in this article is the sequence. What changes is that the sequence becomes affordable inside a live deal timeline, which means the defensible number and the fast number are finally the same number.

Common mistakes in bottom-up niche sizing

• Starting from a known-companies list instead of a screened population, which systematically undercounts the long tail
• Using industry codes as the market definition
• Applying one blended relevance share to a mixed population
• Mixing financial years and currencies inside the total
• Double counting parents and subsidiaries
• Reporting a single point figure with no sensitivity range
• Losing the audit trail, so the number cannot be reused or defended later

Want to see a niche market sized end to end, from company population to competitive structure? Book a demo and bring one of your own niches.