Business licences benchmark research: a method

What to take away

Business licences benchmark research is valid only when the compared cases, measures and periods are alike. A national league table built from unlike local processes can mislead more than no benchmark.
Fix one unit, one permission family, one geography label and one application journey stage before you collect anything.
GOV.UK Find a licence and the ONS business activity tables give context, not licence case counts.
Every figure should carry its coveragecase count, exclusions and period.
Set the reference group before you see results, and keep an internal historical fallback ready.

Start with the decision the number will support

A benchmark answers a decision, not a curiosity. Write down what changes if the comparison shows a gap: a staffing shift, a form rewrite or a channel move.

Government Digital Service guidance on using data to improve your service ties measures to user needs, baselines and decisions. Make a one-page metric card naming the decision, the measure, the unit, the period and the owner.

Define the comparison question

State the decision the benchmark will support
Link measures to needs and baselines
Create a metric card before seeking peers
Do not invent the question from data

Fix the unit before you fix the number

Count people, organisations, premises, applications, permission rows or support interactions, then hold that choice constant. A premises licence row and a company registration row are not interchangeable. For premises licences in England, the Licensing Act 2003 applies; the local authority is the licensing authority and maintains a public register of premises licences (the licensing register).

Read a side-by-side comparison of licensing products before deciding what the measure covers, because documented scope decides which cases enter the count.

Each unit carries its own denominator: applications include withdrawn cases, while premises rows count standing permissions. State it and keep it.

Keep permission families separate: examples include premises licences, personal licences, taxi and private hire vehicle licences, street trading permissions, food business registrations and HMO licensing. Their different scopes mean they should not be pooled as one case type.

Where English licensing data stops being useful

What it supports

GOV.UK Find a licence: Permission families for one activity and location
ONS business activity, size and location: Denominators by industry, size band and area
Council registers and committee reports: Named premises licences and local policy
Freedom of Information requests: Case volumes where a register is thin
Your own case extract: Processing time, withdrawals and open cases

What it cannot do

GOV.UK Find a licence: Counting applications or decisions
ONS business activity, size and location: Licence case volumes or outcomes
Council registers and committee reports: Comparable definitions across councils
Freedom of Information requests: Consistent definitions or timely returns
Your own case extract: Peer comparison without a definition sheet

GOV.UK Find a licence filters by activity and location and warns that its catalogue may not be exhaustive. ONS business activity, size and location tables publish under stated definitions, so a business count is a denominator, not a licence count.

Example: one council, one quarter, one permission family

The figures below are invented to show the arithmetic, not published statistics.

One council, one quarter, one permission family

Define the casepremises licence applications one English council received in a quarter.
Extract it120 received, 96 decided, 14 withdrawn, 10 open at quarter end.
Set the denominator110 closed cases, leaving open cases out of any processing-time measure.
Report the shapemedian 28 days, with 12 of the 96 decided cases over 60 days.

Another council's headline figure skips those checks and says nothing about your own service.

Choose a reference group before you see results

For an English premises-licence comparison, use GOV.UK Find a licence to identify the relevant permission and location, then consult the relevant local authority's public register of premises licences (the licensing register) for published records. Check the register's coverage and record definitions before treating those records as comparable cases.

Fix the permission family, unit, period and geography in writing.
Set a minimum group size in advance, for example five comparable authorities.
Shortlist on activity mix, channel and journey stage or case state (received, decided, withdrawn or open) rather than convenience.
Freeze the calculation with each data owner before results arrive.
If fewer than five match, fall back to your own earlier periods or one descriptive case study.

Do not widen the group after analysis to manufacture a ranking. Note the date you last searched for peers.

Report uncertainty as part of the number

Publish the sample size, the missing records, the exclusions and the period with every benchmark figure. Treat that set as part of the number, not a footnote.

Use a range where one average hides variation, and suppress cells built on fewer than your agreed minimum group size. Record the evidence date and retire the benchmark once the method cannot be maintained.

Ask every data owner to complete one definition sheet: activity, authority type, geography, channel, journey stage or case state, period, exclusions and treatment of unknowns. Compare sample records too, since researching the people who need licensing help shows how differently applicants and staff describe one case.

Keep shared extracts under access control and release the minimum needed. Check UK GDPR duties before publishing anything that could identify an applicant or premises.

Before you act

Write the decision the benchmark supports (see Start with the decision the number will support).
Fix the unit and its denominator (see Fix the unit before you fix the number).
Name each source and its limits (see Where English licensing data stops being useful).
Set the reference group before results arrive (see Choose a reference group before you see results).
Attach coverage notes to every figure (see Report uncertainty as part of the number).
Assign a reviewer who did not run the service being compared.

Common questions

Where can I find English licensing data to benchmark against?

GOV.UK Find a licence lists permission families by activity and location, and ONS tables supply denominators by industry and size band. Neither counts cases, so case-level work needs council registers or your own extracts.

How do I pick a reference group when no peer matches?

Set a minimum group size first, then shortlist on activity mix, channel and journey stage or case state. If too few authorities qualify, use your own earlier periods or a single descriptive case study.

What belongs on the definition sheet?

Activity, authority type, geography, channel, journey stage or case state, period, exclusions and treatment of unknowns. Every data owner completes it before a number moves.

How much uncertainty detail should I publish?

Report the case count, exclusions, period and a range where variation is wide, then suppress thin cells and date the evidence.