Pro free for 30 days.No card.Start free →

Measure customer satisfaction honestly: CSAT and NPS for small business

A plain-English guide to CSAT and NPS: what each measures, how to ask the questions, the traps that inflate or deflate scores (response bias, small samples, method changes), and a monthly routine that turns feedback into decisions.

By WorkBOS Content Desk7 min read
Row of smiley face icons rating customer satisfaction (Photo by Tumisu on Pixabay)
Photo: Tumisu / Pixabay

Most small business owners want to know one thing: are my customers happy? Satisfaction scores promise an answer. A number like "92% satisfied" or "NPS of 45" looks precise and is easy to put on a slide. But a number can also mislead you. It can reflect who chose to answer, how the question was worded, or when you asked, and not how customers actually feel.

This guide explains the two most common measures, CSAT and NPS, in plain English, and shows how to collect and read them honestly. You will learn what each one tells you and what it cannot, how to ask the questions, how to avoid the traps that inflate or deflate results, and how to turn the numbers into decisions. It is general business guidance, not statistical or professional advice.

What CSAT and NPS actually measure

Both are survey scores, but they answer different questions.

CSAT: how did this go?

Customer satisfaction score (CSAT) asks about a specific moment, such as a support chat, a delivery or a project handover: "How satisfied were you with the service you received?" Customers usually answer on a short scale, from very dissatisfied to very satisfied. A common way to report it is the share of customers who picked the top two options.

CSAT is good for judging a particular interaction soon after it happens.

NPS: how do you feel about us overall?

Net Promoter Score (NPS) asks a broader question about loyalty: how likely are you to recommend us to a friend or colleague? IBM's overview of Net Promoter Score explains the calculation. Customers who score 9 or 10 are promoters, 7 or 8 are passives, and 6 or below are detractors. You subtract the percentage of detractors from the percentage of promoters, giving a result between -100 and +100.

NPS is good for tracking relationship health over time, for example once or twice a year.

A quick comparison

| | CSAT | NPS | |---|---|---| | Question | How satisfied were you with this? | How likely are you to recommend us? | | Best used | Right after a specific interaction | Periodically, for the overall relationship | | Result | Percentage satisfied | Promoters minus detractors, -100 to +100 | | Strength | Pinpoints which touchpoint needs work | Simple trend line for loyalty | | Weakness | Says little about loyalty | Says little about why |

Where NPS came from, and why to be modest about it

NPS became popular after a 2003 Harvard Business Review article called The One Number You Need to Grow. Its argument was that the best predictor of growth can usually be captured in a single survey question about recommending the company.

Read the claim carefully

The article's wording is hedged. It says "usually," and it describes a correlation with growth in most of the industries studied, not a guarantee. It is also more than twenty years old. So treat NPS as one useful signal, not proof that your business will grow or shrink.

IBM's overview adds its own limits. Satisfied customers may still not recommend a company, which can make NPS understate their happiness. Without an open follow-up question, the score gives you little to act on. And repeated survey requests can frustrate customers and push scores lower.

How to ask the questions well

Good measurement starts with good questions, asked at the right time.

Keep it short

One rating question plus one open question is enough for most small businesses:

  1. The rating: "How satisfied were you with how we handled your request?" or "How likely are you to recommend us?"
  2. The open question: "What is the main reason for your score?"

The open question is where you learn something. IBM's guidance on NPS recommends asking customers why they gave their score and using the answers for root-cause analysis, especially when scores are low.

Time it to the event

Ask for CSAT within a day of the interaction, while it is fresh, and ask once. Send NPS on a schedule, such as every six months, and not after every contact.

Write neutral questions

Avoid wording that nudges: "How great was our service?" invites a high rating. Keep the scale the same every time so results can be compared.

Do not coach the answers

Asking staff to hint at "please give us a 5" or asking customers to rate you in front of the person who served them will inflate results. If a score is going to affect a bonus, expect it to drift upward for the wrong reasons.

The traps that make scores dishonest

The iSixSigma article How to avoid the evils within customer satisfaction surveys lists several well-known problems. Here they are in plain terms.

Non-response bias

People who answer surveys are often those with very good or very bad experiences. The article says response rates below about 80 percent can start to skew results. Most small businesses will never reach that, so treat your score as a signal from the people who responded, not a verdict from all customers. Always report the number of responses beside the score.

Method bias

The same article notes that phone or in-person surveys tend to produce higher scores than email or web surveys. If you switch method, a change in the score may reflect the method and not a change in service. Keep the method steady when comparing over time.

Small samples

With few responses, the margin of error is wide, so an apparent change can be random noise. If your NPS moves from 40 to 30 on ten responses, one or two customers could explain it. Avoid reacting to small swings; look at trends over several periods.

Different people read scales differently

Each respondent uses a rating scale in their own way. A "7" from one customer may be a "9" from another. Another reason to read the written comments, not just the number.

Turn the numbers into decisions

A score is only useful if it changes something.

A simple monthly routine

  • Collect: gather responses from the month.
  • Report: show the score, the number of responses and the response rate.
  • Read: go through every low score and every comment.
  • Group: tag the reasons, such as speed, accuracy, price, communication.
  • Decide: pick one issue to fix and one thing to keep doing.
  • Close the loop: contact unhappy customers personally, within a few days.

Focus on the middle

IBM suggests that passives, those who score 7 or 8, can be converted into promoters more easily than detractors. A short conversation with them, asking what would have made it a 10, often reveals cheap improvements.

Pair surveys with behavior

Scores tell you what customers say. Also track what they do: repeat orders, renewals, refund requests and complaints. If scores look healthy but repeat business is falling, trust the behavior. Our guide to handling angry customers shows how to record complaints in a way that you can analyze alongside survey data.

Common mistakes

  • Reporting a score without the sample size. Ten responses is not a trend.
  • Chasing the number. Pressuring customers or staff makes the score useless.
  • Asking too often. Survey fatigue lowers scores and response rates.
  • Ignoring the comments. The reason behind the rating is the valuable part.
  • Switching methods mid-stream. Comparisons stop being fair.
  • Never telling customers what changed. People answer again when they see their feedback matters.

Where WorkBOS fits

WorkBOS includes surveys and forms, so you can send a short CSAT or NPS survey after a project closes or a request is resolved, and store the responses alongside the client's record in the CRM. Automations can trigger the survey at the right moment, and the team can discuss low scores in chat and assign a follow-up task to one owner. An AI agent can summarize comments, but any action it proposes needs a person's approval. See features, pricing and the docs for details. For related reading, see setting response-time standards and our guide on client onboarding, where early expectations shape later satisfaction.

Frequently asked questions

Q: Which is better, CSAT or NPS? A: Neither is better. CSAT judges one interaction, and NPS tracks loyalty over time. Many small businesses use CSAT after support or delivery and NPS once or twice a year.

Q: What is a good NPS? A: It depends heavily on your industry, so compare against your own past results rather than a headline number. Track the trend, the sample size and the reasons customers give.

Q: How many responses do I need? A: More is better, and small samples have wide margins of error. If you have only a few responses, rely on the comments and avoid reading too much into small changes.

Q: Should I offer a discount for completing a survey? A: Be careful. Incentives can attract people who want the reward and may bias answers. If you use one, offer it regardless of the score given.

Q: What should I do about a low score? A: Contact that customer personally, ask what went wrong, fix what you can, and add the reason to your monthly review.

Sources are linked in the text above: IBM, Harvard Business Review, iSixSigma.

Run it all from one place

Projects, clients, finance, people and AI employees in one system. Start free, no card needed.

Start free →