Benchmarking

Benchmarking is the process of comparing your results, processes or metrics with a reference point, such as your own past performance, other teams, competitors or recognized best practice, to see where you stand and what to improve. In surveys, it turns a number like "CSAT 78%" into a judgment: better or worse than last quarter, than other branches, or than comparable companies measured the same way.

Benchmarking as a management practice was popularized by Xerox, which started comparing its costs and processes with competitors around 1979. Robert Camp, a Xerox manager involved in that work, described the method in his 1989 book Benchmarking: The Search for Industry Best Practices That Lead to Superior Performance.

Types of benchmarking

How to benchmark, step by step

1. Choose what to compare. Pick a few metrics that drive decisions, for example NPS, CSAT, response rate or time to resolution.
2. Choose the reference. Start with internal benchmarks (previous waves, other locations), then add external ones if a trustworthy, comparable source exists.
3. Make the data comparable. Same question wording, scale, answer labels, channel, timing, type of respondent and calculation method.
4. Measure the gap and check whether it is larger than the margin of error. See gap analysis.
5. Act and re-measure with the same method, so the next comparison shows whether the change worked.

Worked example: internal benchmarking

A chain of clinics runs the same patient survey in every branch each quarter. The share of satisfied patients (top-2-box on a 5-point scale) across the network is 78%. Branch A scores 82% from 150 responses.

Is Branch A better? Its 95% margin of error is about 1.96 × √(0.82 × 0.18 ÷ 150) ≈ ±6 points, so its true score could be anywhere from roughly 76% to 88%. The 4-point lead over the network is within that range and does not prove a difference. If Branch A stays above the network for three or four quarters in a row, the pattern becomes convincing, and its practices are worth studying and copying. The margin of error calculator runs this check for any share.

How to interpret a benchmark

• Look at the distribution, not just the average. Knowing that you are in the bottom quarter of comparable companies says more than "3 points below average".
• Trend beats snapshot. A metric that improves steadily against your own baseline is good news even if a market leader is far ahead.
• Check the source of external numbers. A credible benchmark states the question, scale, sample, country, channel and year. If it does not, treat it as a rough indication.
• Context changes expectations. Customer type, price segment, country and the moment of the survey all move scores.

Common mistakes

• Comparing numbers from different methods, for example your 5-point CSAT with a study that used a 7-point scale, or an email survey with an in-app one.
• Changing your own questionnaire and then comparing with earlier waves.
• Using "industry norms" with no named source. If you cannot say where a number comes from, do not set targets on it.
• Reading small differences as real without checking sample sizes and margins of error.
• Copying practices without understanding why they work. Benchmarking shows the gap. The reasons usually need qualitative research.

Benchmarking with surveys

• Freeze the core questions and answer scales of recurring surveys, and add new questions around them rather than editing them.
• Record the segment of each response (branch, product, channel) so you can compare units inside one dataset.
• Keep a simple table of key metrics by wave with n and the margin of error. After three or four waves you have a reliable internal benchmark.
• Use a baseline study before a major change, so there is something to compare the result with.