Customer Satisfaction Score Calculator

The maths behind it

Customer satisfaction score is the share of respondents selecting a positive rating: satisfied responses ÷ total responses × 100.

The scale varies: 1 to 5, 1 to 7, or a binary, and what counts as satisfied is a choice. Counting only the top box gives a much harsher figure than counting the top two.

It is transactional by nature, asked after a specific interaction rather than about the relationship generally.

Putting numbers to it

600 responses to a post-purchase survey: 402 rate 4 or 5 out of 5. Top-two-box CSAT is 67%.

Counting only 5s gives 41%. Both describe the same responses and the second is a considerably more demanding standard.

Tracking the top-box figure over time is usually more informative, since it moves when experience genuinely improves rather than when people shift between adjacent positive ratings.

Segmenting by order type frequently reveals the problem: 78% on standard orders against 44% on anything that shipped late is a delivery problem rather than a product one.

Where it is unreliable

Response rates are low and skewed toward strong opinions, so the sample is not the customer base. A score built on 3% of customers describes the ones who felt strongly enough to answer.

Survey timing also determines the answer: asked at delivery it measures logistics, asked a month later it measures the product.

How to act on this

Ask a single question immediately after a specific interaction, and always include a free-text follow-up. The score is a tracking number; the comments are the useful output.

Then segment by what happened rather than reading the aggregate, since the score is nearly always dragged by one identifiable operational failure.

Why CSAT and NPS answer different questions

CSAT measures satisfaction with a specific interaction and responds quickly to operational changes. NPS attempts to measure the relationship and moves slowly.

Using CSAT for transactional feedback and NPS periodically for the relationship is the conventional split, and running both on the same cadence produces two numbers measuring the same thing badly.

For a small business, transactional CSAT with a free-text field is generally the more useful of the two, because it produces specific, actionable complaints attached to identifiable orders rather than a score that requires interpretation.

Closing the loop with dissatisfied respondents converts a measurement exercise into a recovery one, and the recovery rate on a personally handled complaint is high enough to justify the effort at most volumes.

It also produces the specific detail that a rating alone never contains, which is where the operational fixes come from.

Asking at a consistent point in the journey is what makes the trend meaningful, since moving the survey earlier or later changes what is being measured.

Keeping the survey to one question and one optional comment maximises response rate, and longer instruments reliably reduce it without adding proportionate insight.

Comparing scores across fulfilment methods or carriers frequently isolates a specific operational partner as the cause of a dip.

Publishing the score internally, alongside the verbatim comments, does more to change behaviour than any target attached to it, because the comments are specific and the number is not.