Benchmarking done properly

Benchmarking means comparing how you perform, and how you work, against someone else, so you can see where you’re behind and learn how others do better. Done well, it gives a team a realistic target and a set of methods to copy or adapt. Done badly, it produces a league table that compares unlike things and an argument about why the numbers don’t apply.

This guide covers the four classic types of benchmarking, a practical process, how to get data without crossing ethical or legal lines, and how to normalise figures so a comparison means something.

Where benchmarking comes from

Comparing yourself with rivals is old, but the modern business practice has a clear starting point. The Global Benchmarking Network’s short history says Xerox coined the term in 1979, when Japanese competitors were selling copiers below Xerox’s own production costs. Xerox studied how they did it, then extended the approach across the company in 1981. Its logistics projects that year showed that useful partners don’t have to be in the same industry.

Robert (Bob) Camp wrote one of the first books on the subject in 1989, and the history above links it to benchmarking’s spread in the United States in the late 1980s. Wikipedia’s benchmarking article sets out the 12-stage process from that book, from selecting a subject through to reviewing and recalibrating. Most later methods are shorter versions of the same idea.

The four types of benchmarking

The Institute for Manufacturing at the University of Cambridge describes four types. They differ in who you compare yourself with.

Many teams start internally because it is quick, then look outside once they’ve fixed the obvious gaps. Competitive benchmarking tells you where you stand. Functional and generic benchmarking more often tell you how to improve.

A practical process

1. Pick a narrow subject. “Customer service” is too broad. “Time to resolve a delivery complaint” is workable.
2. Understand your own process first. Map how the work is done today and measure it. You can’t compare what you haven’t measured.
3. Define the metrics precisely. Write down what counts, what doesn’t, the period and the unit. Ambiguous definitions ruin more benchmarks than bad data does.
4. Choose partners and sources. Decide which type of benchmarking fits, then find the partners or published data.
5. Collect and normalise the data. Adjust for size, mix and definitions so the comparison is fair (see below).
6. Find the gap and its causes. The number shows the gap. Interviews, process walk-throughs and site visits show why it exists.
7. Set targets and act. Adapt the practice that explains the gap, set a target, and review results after an agreed period.

Sourcing data ethically and legally

There are three main sources of comparison data: published information (annual reports, industry studies, trade association surveys), benchmarking groups where a neutral body collects and anonymises members’ data, and direct exchange with a partner. Each has rules.

For direct exchange, the Global Benchmarking Network’s code of conduct (March 2025 version) is a sound guide. Its principles include preparing properly, sharing the same type and level of information you ask for, keeping findings confidential unless the partner agrees, using information only for the agreed purpose, and never attributing data to a partner without permission. It says that if there is any legal doubt, you should take legal advice or not go ahead. It also warns against discussing costs with competitors where those costs feed into pricing.

That warning is real in the UK. In 2017 the Competition and Markets Authority reported that the Competition Appeal Tribunal had upheld a £130,000 fine on a water tank supplier that exchanged information on prices and pricing intentions with three competitors at a single meeting. The company had refused to join their cartel. The CMA’s message was that sharing competitively sensitive information, such as prices or pricing plans, can break competition law even at one meeting.

In practice:

• Never share or ask a competitor for current or future prices, margins, pricing plans or customer-specific terms.
• Use a neutral third party to collect and anonymise competitor data, and only receive aggregated, historical results.
• Don’t misrepresent who you are to get information, for example by posing as a customer or job applicant.
• Take legal advice before any benchmarking exercise that involves direct competitors.

Normalising metrics

Raw numbers rarely compare fairly. A large site will have higher costs than a small one, and a business selling bulky goods will have higher delivery costs than one selling jewellery. Normalising means converting figures into a common basis before comparing.

• Scale by a driver of the work: cost per order, per unit, per employee or per pound of revenue.
• Match definitions: does “customer” mean an account or a person, and does “headcount” include contractors?
• Align periods and seasons: compare the same months, or use 12-month totals.
• Adjust for currency, inflation and accounting treatment where partners are in different countries or years.
• Account for mix: a site handling complex orders will look slow if you only count orders.

An example (illustration only)

A retailer runs two warehouses and wants to know which is more efficient. Both ship 30,000 orders a month.

Measured per order, Site B looks a third more expensive. But Site B serves trade customers whose orders contain twice as many lines, and each line is a separate pick. Per order line, Site B is the cheaper site. The useful question changes from “why is B so costly?” to “what does B do that A could copy?”

Common mistakes and when to use it

• Comparing headline numbers without normalising for size and mix.
• Copying a practice without understanding why it works where it came from.
• Benchmarking only against direct competitors and missing better methods elsewhere.
• Treating benchmarking as a one-off report rather than a repeated check.
• Sharing commercially sensitive data with rivals.

Benchmarking works best when a team suspects it is behind but doesn’t know by how much or why, and when the process can be measured. It’s less useful for brand-new activities with no comparable peers, or where the gap is already obvious and the team needs to act rather than measure again.