Candidate NPS for recruiters: the calculation, shown

Candidate NPS is the standard Net Promoter Score calculation applied to one question asked of candidates: on a 0-to-10 scale, how likely are they to recommend interviewing at your company to a friend. You sort responses into three buckets by score, then subtract the percentage of low scorers from the percentage of high scorers. The result is a single number from -100 to 100 that compresses a batch of survey responses into something you can report and track over time.

Below is the exact question to ask, the bucket definitions, the formula, and a full worked example with the arithmetic shown, so you can check your own calculation against it line by line.

The question to ask

Based on your experience interviewing with us, how likely are you to
recommend [Company] to a friend or colleague as a place to interview?

0 — Not at all likely  ................................  10 — Extremely likely

[Optional follow-up, open text:]
What is the main reason for your score?

Ask this as its own question, on its own 0-to-10 scale, not folded into a 1-to-5 satisfaction question elsewhere in the survey. Mixing scales in one form is covered in candidate experience survey questions, and it applies doubly here, since the whole calculation depends on the 0-to-10 scale being used consistently.

The three buckets

Passives are the bucket most people get wrong. A 7 or an 8 feels like a good score on a 10-point scale, but the Net Promoter method treats it as neutral: it counts toward your total number of responses, but it does not add to the promoter count or subtract as a detractor. This is the part of the method most worth double-checking if your number looks different from someone else's calculation on the same data.

The formula

NPS = (Number of Promoters ÷ Total Responses × 100)
    − (Number of Detractors ÷ Total Responses × 100)

Or, equivalently:
NPS = % Promoters − % Detractors

Passives are excluded from both terms, but included in Total Responses.
The result is a number, not a percentage, typically written without a
"%" sign, ranging from -100 (everyone is a detractor) to 100 (everyone
is a promoter).

A worked example, arithmetic shown

An invented quarter, 40 candidates who completed the survey after their process ended, hired and rejected combined:

Scores received (40 total):
Score 10: 4 responses
Score 9:  6 responses
Score 8:  8 responses
Score 7:  5 responses
Score 6:  3 responses
Score 5:  4 responses
Score 4:  3 responses
Score 3:  2 responses
Score 2:  2 responses
Score 1:  2 responses
Score 0:  1 response
                                          Total: 40 responses

Step 1 — Bucket the scores:
  Promoters (9-10):   4 + 6 = 10
  Passives  (7-8):    8 + 5 = 13
  Detractors (0-6):   3 + 4 + 3 + 2 + 2 + 2 + 1 = 17

  Check: 10 + 13 + 17 = 40 ✓ (matches total responses)

Step 2 — Convert to percentages:
  % Promoters  = 10 ÷ 40 × 100 = 25%
  % Passives   = 13 ÷ 40 × 100 = 32.5%
  % Detractors = 17 ÷ 40 × 100 = 42.5%

  Check: 25 + 32.5 + 42.5 = 100% ✓

Step 3 — Apply the formula:
  NPS = % Promoters − % Detractors
  NPS = 25 − 42.5
  NPS = -17.5, reported as -18 (rounded to the nearest whole number)

A score of -18 here reflects a batch that combines hired and rejected candidates; on its own it says only that detractors outnumbered promoters in this batch. The next step, splitting the same 40 responses by outcome, is what makes the number useful.

The same data, split by outcome

Same 40 responses, now separated into the 14 who were hired and the 26 who were not:

HIRED (14 responses):
  Promoters: 8   Passives: 4   Detractors: 2
  % Promoters = 8 ÷ 14 × 100 = 57.1%
  % Detractors = 2 ÷ 14 × 100 = 14.3%
  NPS = 57.1 − 14.3 = 42.8, reported as 43 (unrounded: 42.86)

NOT HIRED (26 responses):
  Promoters: 2   Passives: 9   Detractors: 15
  % Promoters = 2 ÷ 26 × 100 = 7.7%
  % Detractors = 15 ÷ 26 × 100 = 57.7%
  NPS = 7.7 − 57.7 = -50.0, reported as -50

  Check against combined: weighted by group size, these two groups
  produce the -18 blended score calculated above.

Split this way, the -18 blended figure stops being mysterious: hired candidates are strongly positive (43) and rejected candidates are strongly negative (-50), which is expected direction for most hiring processes. The number worth tracking over time is less the blended score and more whether the rejected-candidate score moves — a shift from -50 to -30 next quarter says something changed about how rejection is handled, even though almost nobody enjoys being rejected.

What moves the score, and what does not

Reading the open-text reason next to the score

The number tells you the direction; the open-text answer tells you why. Group the reasons by bucket rather than reading them in the order they arrived. An invented set from the rejected group above: three detractors wrote some version of "never heard back, had to email to ask what happened"; two wrote "the recruiter was fine, the role itself sounded disorganized when the hiring manager described it." Those are two different fixes — one is a communication-process gap, the other is a hiring-manager preparation gap — and averaging them into one number would have hidden that they need different owners.

Reporting it without overstating precision

Report the score with its sample size every time: "-18 (n=40)" is honest; "-18" alone invites someone to compare it directly against a different quarter's very different sample size as if the two were equally reliable. Track the trend over at least three periods before treating a single quarter's move as meaningful, since a swing of ten or more points from a handful of responses changing buckets is common at typical recruiting volumes and does not necessarily mean anything changed operationally.

Common mistakes

• Counting passives as promoters. A 7 or 8 is a good sign, not a promoter; including it in the promoter count inflates the score and breaks comparability with any standard NPS benchmark or your own history calculated correctly.
• Reporting the score without the sample size. Makes a rough read look as solid as a real trend.
• Blending hired and rejected candidates without also reporting them separately. Hides the more actionable of the two numbers.
• Comparing your candidate NPS directly to a customer NPS benchmark. Different population, different question context, different expectations going in. Compare your own score against your own history.
• Rounding mid-calculation. Round only the final result; rounding percentages before subtracting them compounds small errors, especially with smaller sample sizes.

Where this fits with the rest of candidate experience measurement

Candidate NPS is one number, not the whole picture. Use the fuller candidate experience survey questions to find out what is actually driving the score stage by stage, and the audit in how to improve candidate experience to fix what the detail turns up. The recommend-a-friend question can also sit inside the complete post-interview candidate survey template if you would rather run one combined survey than a standalone NPS check.