Measurement
Part of How marketing agencies measure campaign performance
Why benchmarking at marketing agencies needs a written method
Why benchmarking at marketing agencies holds up when the method is written down: dated sources, stated samples and a clear line between evidence and guesswork.
What to take away
- The Advertising Standards Authority (ASA) published 1,155 rulings in 2024, according to its searchable rulings database, so any benchmark claim about ad standards has public precedent to check against.
- Benchmarking is only as good as its stated sample, date and source. If any of the three is missing, treat the number as an illustration.
- Public sector buying weighs national interest tests as well as price, so benchmarks built only on day rates miss part of the picture.
- Write the method before you collect the numbers, then keep the raw inputs so a client can audit the result.
Why benchmarks go wrong
Most agency benchmark tables fail for a boring reason: the sample is never described. A figure such as "average cost per lead of £85" means nothing without the sector, period and number of accounts behind it.
Same cost, different scope
Team A
- Monthly reporting tooling
- £400
- Brands covered
- One
- Cost per brand
- £400
Team B
- Monthly reporting tooling
- £900
- Brands covered
- Four
- Cost per brand
- £225
For example, a team paying £400 a month for reporting tooling might benchmark that against a rival spending £900. The comparison looks useful until you learn the second team covers four brands. Same cost line, different scope.
Good benchmark research fixes scope first: which services, which client sizes, which months. Then it records what was excluded.
Name the source, not just the type. The IPA Agency Census, WARC's marketing data and Companies House filings for agency accounts each cover a different population, and the Office for National Statistics publishes the pay and inflation series you need if you rebase an old rate. State which one you used, for which year, and what it covers.
Four questions before you trust a figure
Ask who collected it, when, from how many cases, and with what definition. If any answer is vague, downgrade the figure.
Definitions matter more than sample size in small markets. "Retainer" can mean a fixed monthly fee or a minimum commitment. Those are different numbers.
Can you reproduce the figure?
Could a new analyst reproduce it from your notes alone?
benchmark is a measurement
benchmark is a story
Use public rulings as a check
The Advertising Standards Authority's searchable rulings database lets you test whether a claim you plan to repeat has already been challenged. Search the sector, read the reasoning, then adjust the wording.
This is precedent checking, not legal advice. It shows how the regulator has treated similar claims, which is the external anchor a benchmark needs.
Standards and compliance benchmarks
Ad fraud and brand safety numbers travel badly between sources. The IAB UK Gold Standard letter sets out what certification covers, which helps you separate a certified process from a marketing claim.
If your benchmark compares agencies on media quality, state whether each holds that certification and the date you checked. A certification status without a date ages quickly.
These checks sit inside a wider measurement and reporting cycle, so decide where the benchmark file sits in that cycle before you collect the first number.
Public sector buying as a benchmark
Commercial benchmarks often assume price is the main lever. Public sector buying does not work that way. UK public procurement runs under the Procurement Act 2023, and the government's Procurement in the national interest policy sets out how national interest tests sit alongside cost in public contracts.
If you benchmark agency performance for public bodies, include those tests in your criteria. Otherwise you will rank suppliers on a measure the buyer ignores.
Build the benchmark file
Keep one file per benchmark, holding the question, the population, the collection dates, the raw inputs and the known gaps. Version it when you update. Where the benchmark sits beside a directory listing or a review, link the file to that entry so readers can see the sample behind the score.
When you present the result, show the range, not just the midpoint. A range of £300 to £1,400 a month tells a buyer more than an average of £780.
Pair the file with a live view so the numbers stay current, and keep the presentation layer honest when the underlying data changes.
Benchmark checklist
- Question written before collection started.
- Population and exclusions recorded.
- Collection dates stated for every figure.
- Source named for each external number, with the publication, report or register it came from.
- Raw inputs retained and versioned.
- Range shown alongside any average.
- Comparative claims checked against Advertising Standards Authority rulings.
- Certification claims dated, including IAB UK Gold Standard status.
- Public sector criteria included where relevant, including national interest tests.
- Owner named for the next review.
Common questions
How many sources does a benchmark need?
One named, dated source beats five unnamed ones. Add sources only when they measure the same thing in the same way.
Can I use client data as a benchmark?
Yes, if consent and anonymisation are in place. State the sample size and period, and never let one large client dominate the average without saying so.
How often should benchmarks be refreshed?
Set a review date when you publish. Quarterly suits fast-moving media costs; annually is usually enough for pricing structures.
What if the numbers contradict each other?
Report both and explain the likely cause, such as different definitions or periods. A visible conflict is more useful to a buyer than a tidy average.
