Search theflght

Find a useful decision.

Join theflght

Get practical guides, straight to your inbox.

Pricing, hiring, positioning — the decisions that come after the idea. No spam, no fluff.

Market Research

How to Size a Small Market From the Bottom Up

Build a small-market estimate from eligible buyers, buying frequency, reach and capacity, with UK public-data sources and an auditable worksheet.

How to Size a Small Market From the Bottom Up

Short answer. Count eligible buying units, estimate purchase frequency and realistic spend, then constrain the result by reachable channels and delivery capacity. Reconcile it with top-down data, but do not treat a national sector total as the market a small business can serve.

An industry report may say a market is worth billions and still tell a local or specialist founder almost nothing. The business does not sell to an abstract industry. It sells a defined unit, in a defined place, through a channel it can reach, at a frequency and price it can deliver.

The cost of a bad market-size estimate cuts both ways. An inflated number can justify years of work in a niche that cannot support the required sales. An unnecessarily broad definition can also hide a small, reachable segment that would support a good business.

Decide what the market must support

Do not start with "What is the total market?" Start with the decision.

Examples include:

  • Is the niche large enough to support £120,000 of annual revenue at the proposed gross margin?
  • Are there enough buyers within a 45-minute service area to fill two technicians' available capacity?
  • Can one founder reach 40 qualified buyers a month without relying on an expensive national campaign?
  • Does the segment contain enough repeat purchasing to recover the acquisition cost?

Theflght's break-even sales guide can turn fixed costs, contribution per sale and owner-pay requirements into a minimum sales target. The market-sizing exercise then asks whether that target is plausible inside the defined segment.

Use a unit that can actually buy

The correct unit might be a household, person, enterprise, local site, school, clinic, property, vehicle or buying team. Choosing the wrong unit can multiply the answer without adding buyers.

The ONS UK business-count dataset distinguishes an enterprise, essentially the business with autonomy or control, from a local unit, an individual site such as a shop or factory. A service sold once through head office may need enterprise counts. A maintenance service bought by each branch may need local-unit counts. Mixing them double-counts demand.

Nomis reported 2,786,985 UK enterprises for the 2026 reference year, based on the Inter-Departmental Business Register. That national total is authoritative for its defined population, but it is not a useful addressable market until the founder filters industry, geography, size, legal status and buying conditions. The Nomis release record says the business-count datasets are based on enterprises and local units live at a reference date in March and can be broken down by detailed industry and employment size.

Write the unit before opening a database:

One potential buyer equals [unit] that meets [eligibility rules], is located in [geography], experiences [buying trigger], and has [authority or route to purchase].

Build the estimate in seven auditable steps

1. Define eligibility as inclusions and exclusions

"Small retailers" is not a segment definition. A stronger definition might be "independent food retailers with one to three customer-facing sites in Greater Manchester, excluding temporary stalls, national chains and sites without on-premise refrigeration".

Each condition should have a reason connected to the offer. If the reason is merely "they seem likely", label it as an assumption to test.

2. Find the closest public count

For UK business markets, useful starting points include:

  • sector registers maintained by the relevant regulator or public body, where the buying unit is a licensed establishment rather than a generic company.

Record the dataset edition, reference date, geography, table, filter and access date. A link to a landing page without the query settings is not a reproducible source record.

3. Audit what the count misses or includes

The current ONS quality and methodology information says its business-location publication covers businesses registered for VAT and/or PAYE. It does not include a sizeable population of low-turnover, non-employing businesses outside those registrations. That matters if the offer targets sole operators; it may matter little if the buyer must employ 20 people.

Companies House is useful for inspecting named companies, but its own search guidance warns that the service is not comprehensive and may contain errors or omissions. A company can also use a broad or stale SIC code. Treat the register as evidence to check, not a verified customer list.

4. Reconcile the classification with the commercial segment

SIC codes classify a unit by primary economic activity. Commercial segments often cut across codes: "businesses with refrigerated displays", "firms exporting fragile goods" or "practices with a particular workflow".

At the fact-check date, the detailed 2026 business-count release still describes its industry breakdown using SIC 2007, while ONS guidance on SIC 2026 says the new classification was finalised in April 2026 but will be introduced in stages. Record which classification a dataset actually uses. Do not silently combine time series or codes across a transition.

5. Apply an evidence-based eligibility rate

If the public count is broader than the target, estimate the eligible share from a transparent audit. For example, randomly inspect a stated number of register entries using pre-written rules, or obtain a sector dataset with the missing characteristic.

Do not inspect the most promising names and call the resulting percentage representative. Record duplicates, missing data and uncertain cases. Use a low, base and high rate when the audit is small.

6. Add purchase frequency and realistic value

Market revenue is not simply buyer count multiplied by the founder's preferred price. Estimate:

  • how often an eligible buyer experiences the trigger;
  • what proportion of triggers lead to an external purchase;
  • units bought per purchase;
  • realistic net price after discounts; and
  • whether the purchase displaces an existing option or creates new spend.

Use invoices, tender records, buyer interviews and observed purchase histories where available. Keep stated intention separate from observed spend.

7. Constrain by reach and capacity

The market a founder can serve now is the smaller of:

1 eligible demand reachable through known channels;

2 demand that can be converted at a defensible rate; and

3 orders the business can fulfil while maintaining quality and cash flow.

This is where a large market can still produce a small first-year opportunity. Search interest may help locate people, but search volume is not the same as buying intent.

The bottom-up calculation

Use three nested figures rather than one headline:

Eligible annual demand

countable units x eligible share x purchase frequency x units per purchase

Reachable annual demand

eligible annual demand x reachable share x realistic conversion rate

Serviceable first-year demand

the lower of reachable demand and safe delivery capacity

Then calculate revenue and contribution separately:

serviceable orders x net price = serviceable revenue

serviceable orders x contribution per order = contribution before fixed costs

Illustrative example - not a real market estimate

Assume a founder queries an official business-count dataset and, after documenting geography and SIC filters, obtains 1,200 local units. The numbers below are hypothetical and exist only to show the method.

InputLowBaseHighEvidence status
Public local-unit count1,2001,2001,200Dataset result to be cited
Share meeting the operational condition50%70%80%Sample audit assumption
Purchases per eligible site per year1.01.52.0Buyer records/interviews needed
Reachable share in year one10%20%30%Channel assumption
Conversion of reachable qualified sites15%25%35%Test assumption
Safe annual capacity607284 ordersOperating constraint

In the base case, eligible sites are 1,200 x 70% = 840. Annual buying events are 840 x 1.5 = 1,260. The founder expects to reach 20%, or 252 events, and convert 25%, or 63 orders. Capacity is 72, so reach and conversion constrain the first year to 63 orders.

At an illustrative net price of £600, serviceable revenue is £37,800. At £360 contribution per order, contribution before fixed costs is £22,680. If the business needs 100 orders to break even, this base case fails even though the theoretical market contains 1,260 annual buying events.

The right response is not to increase the conversion assumption until the model passes. It is to test whether eligibility, frequency, reach, price, contribution or geography can be improved with evidence.

Practical asset: the source-and-assumption worksheet

Model lineValue or rangeUnitGeogr aphyReference dateSource/ queryAssumption or observed fact?Main bias or gapTest or update owner
Countable buyer units
Eligible share%
Purchase frequencyper year
Net price£ per order
Reachable share%
Qualified conversion%
Model lineValue or rangeUnitGeogr aphyReference dateSource/ queryAssumption or observed fact?Main bias or gapTest or update owner
Safe capacityorders per year
Contribution per order£

Give every assumption an owner and a next test. The estimate becomes more useful as the business replaces assumptions with observed buyer and delivery data.

Reconcile, do not average, the top-down number

A top-down report can expose a missing category, but it should not be averaged with the bottom-up result. Reconcile definitions:

  • Does the report include products, geographies or customer types the business will not serve?
  • Is it measuring supplier revenue, end-customer spend or transaction value?
  • Is the year historical, forecast or current?
  • Is the denominator all organisations, registered enterprises or active buying sites?

If the figures differ, explain the difference. A large gap is often a definition problem, not proof that one calculation is wrong.

Complete one worksheet using the narrowest segment the business can identify and reach. Compare serviceable contribution with the minimum required sales, then run a small acquisition or buyer-interview test on the most sensitive assumption. Use Theflght's guide to identifying a first profitable segment and decide how much demand is enough for this business.

Rebuild the model when an annual dataset is released, the SIC mapping changes, the geography expands, price or contribution changes, or actual reach and conversion materially depart from the range.

Sources

Cover photograph: Shoppers on a city retail street by Julian. Unsplash License. Adapted with an editorial headline and The FLGHT branding.

Related guides

BUSINESS ADVISER — Editor at theflght

Practical guides for founders making the decisions after the idea.

Comments (0)