How to Run an NPS Survey Step by Step

Knowing the theory behind Net Promoter Score is one thing; actually running a survey that produces clean, trustworthy data is another. Many teams launch an NPS survey, collect a pile of scores, and then realize the data is biased, the timing was wrong, or there is no plan for what to do with the results. This step-by-step guide walks you through the entire process — from defining your goal to closing the loop — so your first NPS survey delivers insight you can act on, not just a number on a slide.

• Step 1: Define your goal and survey type
• Step 2: Write the question and follow-up
• Step 3: Choose your channel
• Step 4: Get the timing right
• Step 5: Decide who to survey
• Step 6: Collect responses and calculate the score
• Step 7: Act on the results
• Frequently Asked Questions

Step 1: Define your goal and survey type

Before you write a single word, decide what decision the survey will inform. Are you tracking overall brand loyalty over time, or measuring satisfaction with a specific interaction like onboarding or a support ticket? This choice determines whether you run a relational NPS (a periodic pulse on the whole relationship) or a transactional NPS (triggered after a defined event). A vague goal produces a vague survey; a sharp goal tells you exactly who to ask and when.

Write your goal down in one sentence, for example: "Measure relational NPS quarterly to track whether product changes are improving loyalty." That sentence will anchor every decision that follows.

It also helps to name the audience for the results at this stage. If leadership will use the score for strategic planning, you lean relational. If a product or support team will use it to fix a specific journey, you lean transactional. Knowing who consumes the data keeps the survey design focused and prevents the all-too-common outcome of a report that interests no one and changes nothing.

Step 2: Write the question and follow-up

Use the standard, proven NPS question rather than inventing your own: "On a scale of 0 to 10, how likely are you to recommend us to a friend or colleague?" Resist the temptation to tweak the scale or wording, because doing so breaks comparability with benchmarks and your own historical data.

Always pair the rating with one open-ended follow-up such as "What's the main reason for your score?" The verbatim responses are where you learn what to fix and what to celebrate. Keep the survey to these two questions; every extra field you add lowers your completion rate. If you want a ready-made structure, start from an NPS survey template and customize the branding.

Step 3: Choose your channel

Deliver the survey where your customers are most engaged. The main options are email, in-app or in-product prompts, SMS, and post-interaction widgets. In-app surveys typically get higher response rates because they reach customers in context, while email has broader reach and works well for relational pulses. SMS can be effective for service businesses with mobile-first customers.

Match the channel to your audience. A SaaS product, for instance, can embed the question directly in the app at a natural moment — you can see how this works for a SaaS NPS survey built for product-led teams. Whatever you choose, keep it consistent across waves so channel bias does not distort your trend.

Step 4: Get the timing right

Timing is one of the most underrated factors in NPS quality. For transactional NPS, send the survey soon after the relevant event — a completed purchase, a resolved support ticket, a finished onboarding — while the experience is fresh, but not so immediately that the customer feels ambushed. A delay of a few hours to a day is often a good balance.

For relational NPS, avoid surveying customers who just signed up and have not yet formed an opinion; give new customers enough time to experience the core value first. Also avoid sending around known pain points, like right after a price increase or an outage, unless that is specifically what you want to measure. And never survey the same person too frequently — set a minimum gap (such as 90 days) to prevent fatigue.

Step 5: Decide who to survey

You do not always need to survey every customer. For relational NPS, a representative sample is fine, and rolling samples (surveying a fraction each month) smooth out seasonality and reduce fatigue. The important thing is that your sample is unbiased: do not only survey your happiest power users or your most vocal complainers, because either skew will distort the score.

Aim for enough responses that the result is stable. While there is no universal magic number, larger samples give you tighter confidence and let you segment by plan, region, or tenure. If your customer base is small, survey everyone; if it is large, sample systematically rather than letting self-selection decide who answers.

Beware of survey fatigue, which is a silent killer of data quality. If you contact the same people too often across all your research, response rates fall and the customers who do reply skew toward the most engaged or the most annoyed. A shared contact policy across your organization — capping how often any individual is surveyed — protects both your response rate and the representativeness of the people who answer.

Step 6: Collect responses and calculate the score

Once responses arrive, bucket each one: promoters scored 9–10, passives scored 7–8, and detractors scored 0–6. Then apply the formula:

NPS = % Promoters − % Detractors

For example, with 300 responses where 165 are promoters (55%), 75 are passives (25%), and 60 are detractors (20%), your NPS is 55 − 20 = 35. A good survey tool calculates this automatically and breaks it down by segment, so you can see, for instance, that your enterprise customers are promoters while your self-serve tier skews toward detractors. That segmentation is often more actionable than the headline number.

Step 7: Act on the results

The survey is only half the job; the response is what creates value. Build a simple workflow: route detractor responses to a team member who can reach out personally, tag recurring themes in the verbatim comments, and feed product and process fixes back into your roadmap. Thank promoters and, where appropriate, invite them to leave a review or refer a colleague.

This follow-up discipline is called "closing the loop," and it is the single biggest differentiator between teams that improve their NPS and teams that just watch it. Schedule a recurring review of NPS verbatims with the relevant stakeholders so the feedback consistently turns into action. To launch quickly, you can adapt a customer satisfaction survey template and add your NPS question.

Finally, treat your first survey wave as a baseline rather than a verdict. A single score in isolation has limited meaning; its value emerges when you compare it to the next wave and the one after that. Set the expectation with stakeholders early that NPS is a trend metric, so nobody overreacts to a single number or, worse, abandons the program because the first result was lower than hoped. The discipline of measuring consistently and acting on what you learn is what eventually moves the line.

Frequently Asked Questions

How many responses do I need for a reliable NPS?

There is no single required number, but more responses produce a more stable score and tighter confidence intervals. If your customer base is small, survey everyone. For larger bases, aim for a representative sample large enough to segment meaningfully — often a few hundred responses per wave is a practical starting point.

Should I send NPS by email or in-app?

Both work; the right choice depends on your audience. In-app surveys reach customers in context and usually get higher response rates, while email offers broader reach for relational pulses. Pick the channel your customers engage with most and keep it consistent across survey waves.

How soon after an interaction should I send a transactional NPS?

Send it while the experience is still fresh — typically within a few hours to a day of the event — but not so instantly that it feels intrusive. The goal is to capture an informed reaction, not a reflex.