Sizing a franchise territory with ONS data across England and Wales

What to take away

What ONS data can and cannot tell you about a franchise territory

ONS data tells you how many people live in a proposed territory, how many businesses trade there, and how much economic output the area generates. That is the demand side of the unit economics test.

It does not tell you whether a specific site will convert, what rent you will pay, or whether a competitor already owns the best pitches.

Use ONS figures to set the size of the prize. Use your own pipeline, rent quotes and staffing costs to test whether the prize is worth taking. A territory that passes the population test can still fail on rent, parking or access.

Population estimates are the starting point. They are produced annually and rolled forward between censuses, so they are timely but modelled. Census data is richer and more granular, but it ages. Treat the two as a pair: census for structure, estimates for currency.

ONS methodology pages explain how each dataset is built, including sampling error and disclosure control. Read them before you put a number in a board paper. A figure you cannot explain is a figure you cannot defend.

What ONS data cannot do is prove sales. It can tell you that 40,000 people of the right age and income live in a territory. It cannot tell you that they will buy from you. That gap is where demand signals for a franchise territory matter, because they test behaviour rather than headcount.

Census and population estimates for England and Wales

Start with the latest census for England and Wales. It gives you population by age, household composition, tenure, car ownership and economic activity at output area level. Those variables matter more than raw headcount for most franchise models.

A care franchise needs over-75s and households with limited support. A children's activity franchise needs under-10s and car ownership. A trades franchise needs owner-occupiers and older housing stock. The census lets you build a territory profile rather than a population total.

Population estimates then bring the census forward. They are published for local authorities, regions and England and Wales as a whole, with components of change: births, deaths and migration. Migration is the swing factor in many urban territories.

Use mid-year estimates to check whether a territory is growing or shrinking. A boundary that looks strong on census day may be losing households by the time you open. The ONS population estimates for territory sizing collection is the place to start.

Build a simple table for each candidate territory: total population, target age band, households, and change over the last two estimate years. If the target band is falling while total population rises, the territory is ageing away from your model.

Welsh local authority figures need the same treatment. Wales has 22 unitary authorities, and their population profiles differ sharply between the south Wales valleys, Cardiff and rural mid and north Wales. Do not treat Wales as a single market.

English local authority figures cover 317 councils of different types: county, district, unitary, London borough and metropolitan borough. A territory built on districts may not match a territory built on unitaries. Check which geography your data uses before you compare.

For both nations, check whether the data is for the built-up area or the administrative area. A local authority can include large rural tracts that will never generate customers. Population density and settlement patterns matter as much as totals.

Local authority figures and how boundaries change the numbers

Boundaries change. Councils merge, wards are redrawn, and output areas are re-cut at each census. A territory that looked viable on 2011 geography may look different on 2021 geography, even if nothing changed on the ground.

This matters because franchise agreements often define territories by postcode, not by council area. Postcodes do not align neatly with local authority boundaries. A single postcode district can straddle two councils with different demographics.

Build your territory on the smallest stable geography you can. Output areas and lower-layer super output areas are more stable than wards and much more stable than council boundaries. Aggregate them up to your proposed postcode territory.

When you compare territories, hold geography constant. If territory A is measured on a district and territory B on a unitary, you are comparing apples with pears. Re-cut both to the same building blocks.

Check the boundary changes that accompanied the most recent census. Some councils changed name and composition, which breaks time series. A trend line across a boundary change is not a trend.

For Wales, check the Welsh local authority figures separately. Some datasets are published for Wales only, and some England and Wales datasets do not break out Welsh authorities in the same way. Read the notes.

If your territory crosses the England and Wales border, you will need both sets of figures. The border area around Chester, Shrewsbury and Monmouthshire is a common example. Do not assume the data is harmonised.

Business activity, size and location counts as a demand proxy

Population tells you who lives in a territory. Business counts tell you what economic activity happens there. For B2B franchises, business counts are the primary demand measure, not a secondary one.

The ONS business activity, size and location dataset gives counts of businesses by industry, employment size band and location. Use it to count your target customers, not the whole business base.

A commercial cleaning franchise should count offices, clinics and light industrial units above a certain size. A payroll franchise should count employers with 5 to 50 staff. A signage franchise should count retail and hospitality premises. The dataset supports that filtering.

Business counts also reveal competition. If a territory has 200 businesses in your target sector but 12 established competitors, the ratio is worse than a territory with 120 businesses and 2 competitors. Pair the counts with competitor research for a franchise territory using official records.

Size bands matter. A territory with many micro-businesses has different needs from one with a few large employers. Micro-businesses buy differently: lower value, higher churn, more self-service. Large employers buy through procurement, with longer cycles.

Location quotients help. Divide the territory's share of a sector by the national share. A quotient above 1 means the sector is over-represented locally. That is a concentration you can target.

Use business counts as a demand proxy, not a sales forecast. A business in the count may already have a supplier, may be too small to buy, or may not buy your category at all. The count sets the ceiling; your conversion rate sets the floor.

GVA and regional accounts for comparing territories

GVA measures the value of goods and services produced in an area. Regional accounts break GVA down by region and, in some releases, by local authority. It is the best available single measure of an area's economic weight.

The ONS GVA data for comparing territories collection covers the UK, including England and Wales. Use it to rank candidate territories by economic output per head, not just by population.

GVA per head separates affluent low-population areas from poor high-population areas. A rural district with high GVA per head may support premium services that a denser but poorer urban area cannot. Population alone would hide that.

Use GVA to compare regions: Greater London, the South East, the North West, the West Midlands, Yorkshire and the Humber, and Wales. These are the geographies where regional accounts are most reliable. Finer breakdowns carry more uncertainty.

GVA is not household income. It measures production, not spending power. A territory can have high GVA from a single large employer whose workers commute in and spend elsewhere. Check commuting patterns before you rely on GVA alone.

For a franchise territory, GVA is a sanity check on price positioning. If your model assumes premium pricing, the territory should have the GVA per head to support it. If it does not, either the model or the territory is wrong.

Do not use GVA to forecast unit sales. It is a comparison tool. The conversion from economic activity to your till is the part you must test yourself, with local evidence.

Building a territory test from population, business counts and unit economics

A territory test is a single page that says whether a proposed boundary can carry your unit economics. It uses ONS data for the demand side and your own cost data for the supply side.

Here is the sequence.

The penetration rate is the decision variable. If the territory needs 8 per cent of target households to break even and your best site achieves 3 per cent, the territory is too small or the model is too expensive.

Use a table to compare candidates. The table below shows the shape of the test, with illustrative figures for two territories.

Territory B looks stronger on population and penetration, but weaker on business counts. If your model is B2B, Territory A may be the better bet despite the lower population. The table forces that argument into the open.

Run the test at three boundary sizes: the proposed territory, a 20 per cent smaller version, and a 20 per cent larger version. If the verdict flips between them, the boundary is doing too much work.

Add a downside case. Cut the target population by 10 per cent and raise fixed costs by 10 per cent. If the territory still breaks even, it has headroom. If it does not, you are buying a marginal site.

For a longer view, the 2027 outlook for one franchise territory shows how to build a forward view from local evidence rather than national averages.

Timing the analysis around the ONS release calendar

The value of a territory analysis decays. Population estimates are updated annually, business counts annually, and GVA annually with regional detail. Census data is refreshed only every ten years.

Check the ONS release calendar for timing data use before you commission the work. If a new population estimate is due in eight weeks, wait for it rather than building on last year's figures.

The calendar also tells you when methodology changes are coming. A change in the way business counts are compiled can break your trend. Better to know before you sign a territory agreement.

Build a data refresh date into the franchise agreement or the board paper. State which release the analysis used and when the next update is due. A territory model without a refresh date is a snapshot pretending to be a forecast.

If a decision cannot wait for the next release, use the most recent data and label it clearly. Do not mix vintages silently. A population estimate from one year and a business count from another can be combined, but say so.

For fast-moving decisions, use the most frequent releases. For boundary-setting decisions, use the most granular. The calendar helps you match the release to the decision.

Documenting assumptions so the territory model can be challenged

A territory model is an argument. Like any argument, it should show its workings. Document the assumptions so a colleague, a franchisee or a lender can challenge them.

Start with the geography. Name the postcode districts, the local authority areas and the output areas used. State the census year and the estimate year. State the release date of each dataset.

List the demand assumptions: target age band, target business sector and size band, and the conversion rate assumed. List the cost assumptions: rent, rates, staffing, vehicles and marketing.

Record the source for every figure. Link to the ONS collection, not to a blog that cited it. The ONS methodology for interpreting statistics explains the limits of each dataset, including sampling error and revisions.

State what would change your mind. If target population falls below a threshold, or if a competitor opens within the territory, the model should be revisited. Write the trigger down.

Keep a version history. When the ONS releases new figures, update the model and record what changed. A model that cannot be traced back is a model that cannot be trusted.

Finally, separate facts from judgements. The population figure is a fact. The assumption that 2 per cent of households will buy is a judgement. Label them differently so the reader knows which is which.

Common questions