Branding agencies benchmark research: set the specification

Use branding agencies benchmark research to assess performance claims in agency profiles, reviews and pitches, not to treat unlike figures as comparable. Define the comparison before looking for a figure. A usable benchmark needs the same population, geography, service, period, event, numerator, denominator and unit as the decision it will inform. Choose the geography to fit the decision: UK-wide evidence is not interchangeable with England-only evidence. If suitable evidence for the chosen geography is absent, report the gap instead of localising a UK or global result.

This article teaches a method; it publishes no market benchmark. Its evidence search was completed on 5 September 2026. Government statistical guidance is used as a quality reference and does not make private agency analysis an official statistic.

What to take away

Write the benchmark specification

Start with a one-line buyer decision, such as whether an agency’s performance claim in a profile, review or pitch is relevant to the brief. Then fix the population, geography, service, period, event, numerator, denominator and unit.

Template: population = [eligible projects or clients]; geography = [target market]; service = [defined branding service]; period = [defined period]; event = [precise outcome]; numerator = [cases meeting that outcome]; denominator = [all eligible cases]; unit = [measure used]. Hypothetical completed specification: population = all branding-agency brand-identity projects for UK clients completed in a single calendar year; geography = UK; service = brand identity; period = that calendar year; event = client approval at first presentation; numerator = eligible projects approved at first presentation; denominator = all eligible projects in that cohort; unit = percentage of projects. This specification describes a claim a buyer could test in a profile, review or pitch; it is not a reported agency result.

A label such as "agency approval rate" is insufficient. It could count files, projects or clients, and approval may mean first submission, final acceptance or absence of complaint. Those definitions cannot be pooled silently.

Publish the source universe

Record where candidates were sought for branding-agency performance claims, and name the sources rather than describing categories. For a UK decision the practical universe is: the agency profiles and reviews in this directory; client-review platforms such as Google Business Profile reviews, Trustpilot and Clutch; the company records filed at Companies House; the member, award and standards material of the Institute of Practitioners in Advertising (IPA) and the Design Business Association (DBA); the Design Council’s published work; industry press such as The Drum and Design Week; and official releases, including ONS statistics and the DCMS sector economic estimates for the creative industries. Log every candidate’s publisher, date, geography, population, method and exclusion decision. A supplier case study remains that supplier’s claim even if its number is precise.

The UK Statistics Authority's Standards for Official Statistics require official producers to consider suitable sources, rigorous methods, comparability and openly communicated quality. Applied as a voluntary reference here, those fields help readers understand why two attractive figures may not be comparable.

Logging a candidate source

Every candidate gets a row before any figure is compared. Worked example for a UK decision: publisher = the agency’s profile and review entries in this directory; date = the date the entry was last updated, as printed on the page; geography = UK; population = the clients and projects the agency names; method = self-reported, with no stated cohort or definition; exclusion decision = retained as a supplier claim and excluded from benchmarking.

The same six fields work for a Companies House filing, an IPA or DBA award entry, a set of Trustpilot reviews or a DCMS sector estimate. If a field is blank on the source, write “not stated” rather than filling it in.

Test compatibility before calculation

Create a row for every required field and compare candidates with the target specification. Keep incompatible evidence separate. Do not average a project-level measure with a client-level measure, a monthly rate with an annual total or a UK population with an unexplained sample from another geography.

For survey evidence, inspect recruitment, response, weighting and uncertainty. The ONS guide to survey uncertainty explains that sampling produces uncertainty and that non-response, inaccurate answers and processing can create non-sampling error. A commercial study needs its own disclosures and cannot assume the same controls.

Separate a baseline from a market reference

An organisation's own history may be the closest usable comparator. Call it an internal baseline, retain a stable definition and mark changes to systems, services or cohort. It describes that organisation under those conditions; it is not evidence of an England norm.

If external data are suitable only for context, present them alongside rather than using them as a performance threshold. Geography, period and method should appear next to the figure, not in a distant footnote.

Assure and update the result

The 2025 AQuA Book calls for analysis to be verified and validated for its intended use. Have one reviewer reproduce extraction and calculation while another challenges the population, method and decision fit. Preserve the source snapshot, code or formula, exclusions and approval.

Set an expiry or review trigger for a new release, source revision, definition change, population shift or material missing data. Withdraw the comparison if its denominator cannot be reconstructed or the source no longer supports the stated geography.

The defensible conclusion may be that no comparable benchmark for the chosen geography was found. That is a research result, not a failure. It prevents a buyer from turning an unrelated figure into a target and points to the exact evidence a future study would need.

Before you act

Common questions

What should I do if no suitable England benchmark exists?

If no suitable evidence for the target geography exists, report the gap rather than localising a UK or global result. The defensible conclusion may be that no comparable benchmark for that geography was found. That is a research result, not a failure, and it prevents a buyer from turning an unrelated figure into a target.

How do I test whether two figures are compatible?

Compare each candidate against your target specification field by field, using a row for every required field. Keep incompatible evidence separate rather than averaging different measures or populations.

What quality checks apply to survey evidence?

Inspect recruitment, response, weighting and uncertainty. Sampling creates uncertainty, and non-response, inaccurate answers and processing can add non-sampling error, so a commercial study needs its own disclosures.