NPS Calculation: Formula, Worked Example and B2B Checks

The Net Promoter Score calculation is simple: percentage of promoters minus percentage of detractors. The common mistake is in the population behind those percentages. Passives belong in the denominator. People who did not answer do not become detractors. Several respondents from one company do not become several customer accounts.

This guide gives you the formula, a worked example, a spreadsheet check and a reporting checklist for B2B teams. It measures stated willingness to recommend. It does not turn that answer into a renewal forecast or an automatic sales instruction.

The NPS formula and the three response groups

Bain's scoring definition uses a recommendation question answered on a 0–10 scale. Classify valid answers as promoters (9–10), passives (7–8) and detractors (0–6).

NPS = 100 × (promoter count − detractor count) ÷ total valid responses. The total is promoters + passives + detractors. The result ranges from −100 to +100. Write it as a score, such as +25, rather than as “25% of customers recommend us.”

A zero rating is a valid detractor response. A blank field, an invalid value or a “not applicable” answer is not a zero. Keep those records visible in the survey report, but outside the standard score calculation. When there are no valid answers, report “no score available”; zero NPS would incorrectly imply equal promoter and detractor shares.

A worked calculation, including passives

Imagine a fictional B2B service receives 40 valid ratings: 20 promoters, 10 passives and 10 detractors. Promoters are 20 ÷ 40 = 50%; detractors are 10 ÷ 40 = 25%. The NPS is 50 − 25 = +25. This is an illustration, not customer data or a benchmark.

If you remove the 10 passives, the denominator falls to 30 and the result becomes +33.3. That is a different calculation. A passive answer contributes zero to the numerator while still contributing one to the denominator.

Do not average the original ratings and call that NPS. Ten ratings of 8 produce a mean rating of 8 and an NPS of 0: every response is passive. Both summaries can be calculated, but their labels and meanings must stay separate.

Check a spreadsheet before copying the number

For a clean column of valid numeric ratings in A2:A41, an Excel or Google Sheets formula is:

=(COUNTIF(A2:A41,">=9")-COUNTIF(A2:A41,"<=6"))/COUNT(A2:A41)*100

This formula assumes every numeric cell contains an allowed integer from 0 to 10. It is not a substitute for validation: a stray −1, 11 or duplicate submission can corrupt the result. Confirm the three group counts add up to the valid-response count. Recalculate one small sample manually. Use the same rounding rule in every report, preferably rounding only the final result.

Keep one row per retained response with a respondent identifier, account identifier, role, survey wave, date, rating and optional reason. Decide before analysis how to handle repeated submissions within a wave, then apply the rule consistently rather than keeping whichever rating improves the score.

In B2B, report people and accounts separately

An operational user, an administrator and the budget owner can experience the same supplier differently. Bain's B2B guidance explicitly discusses the need to choose whom to ask within complex customer relationships. A practical consequence is to record the respondent's role and the account they represent.

Suppose one account contributes eight responses and another contributes one. In a respondent-level NPS, the first account supplies eight observations. That may be appropriate for a question about people using the service, but it is not an equal-account view of the customer base. State the unit clearly and report how many accounts are represented.

Do not silently average the contacts' ratings, reclassify that average and label it the standard respondent NPS. If you also build an account-level or revenue-weighted view, publish its aggregation and weighting rules under a separate label. Keep conflicting views within an account available for follow-up; an average can hide the disagreement worth investigating. A sales account plan can help identify the relevant roles, but it cannot supply feedback from someone who has not answered.

Show who answered, not just how many

Put the eligible population, unique invited contacts, valid responses and represented accounts beside the score. Define any participation percentage you report. “40 valid ratings from 100 unique invitees” is more informative than an unexplained “40% response rate.” Track undeliverable invitations and missing ratings separately.

The AAPOR survey research guidance emphasizes transparent sampling and reporting. For an operational customer survey, compare the invited and responding groups using attributes you already know, such as tenure, role or product plan. If almost every answer comes from long-standing daily users, say so.

Nonresponse is not a sentiment category. A silent contact might be busy, unreachable, disengaged or unwilling to answer; the survey alone cannot distinguish these explanations. Investigate coverage gaps without assigning invented ratings. A high participation figure also does not, by itself, establish that the responses represent every customer.

Compare survey waves with a stable method

Keep the question, scale, audience rules, channel and timing as consistent as possible. Record any change beside the result. A survey sent after a successful onboarding milestone and a relationship survey sent across the whole customer base do not measure the same situation. Bain distinguishes relationship, experience and benchmark feedback.

Avoid averaging departmental or monthly NPS figures without their counts. In a hypothetical example, a group of 10 respondents scores +80 and a group of 90 scores 0. Their simple average is +40. If they are disjoint groups from the same compatible survey, the combined respondent-level score is (10 × 80 + 90 × 0) ÷ 100 = +8. Use underlying counts whenever possible, and do not pool incompatible populations merely because both reports say NPS.

Small samples can move sharply when a few answers change. Show the counts and distribution before calling a movement an improvement. There is no universal “good B2B NPS” that removes the need to inspect the method. An external benchmark is useful only when its audience and collection method are sufficiently comparable.

Use the reason for a response to choose the next action

Add a short, neutral request for the reason behind the rating. Review that explanation with the account context before assigning an action. An invoice problem, a missing onboarding step and an unresolved product issue need different owners. Record an owner, a next step and a review date, then tell the respondent what happened when appropriate.

A promoter rating is not permission to publish a testimonial or ask for an introduction automatically. Verify the actual experience and obtain the relevant agreement. A detractor label is likewise not proof of imminent cancellation. Analyze observed departures separately and use customer onboarding for first-value problems.

A checklist for the next NPS report

1. State the recommendation question, survey dates and eligible audience.
2. Count valid ratings, including passives, and distinguish blanks from zeroes.
3. Show the three group counts and calculate the score from the same denominator.
4. Report contacts invited, accounts represented, exclusions and repeat-response rules.
5. Check whether comparisons use compatible audiences and collection methods.
6. Link feedback to a documented action without inventing a renewal probability.

Start by making the current survey auditable before adding another dashboard. If the feedback points to a mismatch between the customers you acquire and the value you can deliver, revisit lead qualification. For your prospecting workflow, explore Lead Scorer's plans; the survey calculation described here works independently in your existing survey tool or spreadsheet.