NPS vs. CSAT vs. CES: picking the right metric

NPS vs. CSAT vs. CES: what each metric actually measures, where it falls short, and how to pick the right one for your research.

Net Promoter Score, Customer Satisfaction, and Customer Effort Score measure different things at different moments. NPS measures relationship-level loyalty, CSAT evaluates a specific interaction, and CES measures effort. Each produces a number but requires an open follow-up to explain why a respondent chose that score.

What each metric measures

Net Promoter Score (NPS)

NPS asks how likely someone is to recommend a product to a friend or colleague, usually on a 0-10 scale. Respondents split into detractors (0-6), passives (7-8), and promoters (9-10). The score is the percentage of promoters minus the percentage of detractors. NPS measures loyalty rather than satisfaction with one interaction. In Versive, NPS is a dedicated question type with the standard 0-10 scale and customizable end labels. Keep the standard "Not at all likely" and "Extremely likely" labels when comparability with outside benchmarks matters.

Customer Satisfaction (CSAT)

CSAT asks how satisfied someone was with a specific experience, such as a support ticket, checkout, or onboarding step. It is typically scored on a short scale, often 1-5, and reported as the share of respondents who selected the top one or two options. Because it is tied to a specific moment, the score can be attributed to the interaction that just occurred. Versive does not have a standalone CSAT question type because satisfaction has no single standardized scale. Build it with Rating Scale, a Likert-style question with a default 1-5 range and labels for the low, middle, and high points. Star Rating provides a more casual, consumer-facing format.

Customer Effort Score (CES)

CES asks how easy it was to complete a task, often through agreement with a statement such as "this made it easy for me to resolve my issue," rated from strongly disagree to strongly agree. Satisfaction and effort can diverge: a customer may be satisfied with the outcome after doing substantial work to reach it, and that friction can be more predictive of churn than satisfaction alone. Like CSAT, Versive has no standalone CES type. Build it with a Rating Scale question and effort-specific end labels.

NPS vs. CSAT vs. CES, side by side

Where each one falls short

NPS is a single number that averages away everything interesting about it. Two products can post the same score for completely different reasons, and the 0-6 detractor band is wide enough to lump someone mildly annoyed in with someone about to churn. It also measures an intention, whether someone would recommend you, rather than a behavior, so it can drift from what people actually do.

CSAT is so reactive that it's easy to over-collect: ask after every interaction and you get survey fatigue instead of signal, and a high score on one ticket doesn't tell you whether the underlying product still has real friction elsewhere. It also has no fixed industry standard the way NPS does, so scores aren't reliably comparable across companies, or even across your own surveys if the wording or scale changes.

CES is the narrowest of the three by design. It only covers effort, not whether the outcome was actually good, so a low-effort interaction that solved nothing will still score well. It's also the least familiar of the three outside product and support teams, which can make it a harder number to rally a whole organization around.

When to use which

Reach for NPS when you're tracking overall relationship health over time, benchmarking against competitors, or segmenting your base into detractors, passives, and promoters for account-level follow-up. It belongs on a quarterly or relationship-level cadence, not after every ticket.

Use CSAT for a specific, well-defined interaction such as a support conversation, onboarding step, or purchase flow. It supports operational monitoring rather than relationship-level loyalty measurement.

Reach for CES around effort-heavy moments specifically, such as support resolution, self-service flows, or onboarding, where the real risk isn't dissatisfaction so much as quiet friction that pushes people toward a competitor with less hassle. Most teams that use CES don't run it alone; they pair it with CSAT or NPS to see effort and satisfaction side by side.

The metrics can work together: NPS as a periodic relationship check, CSAT after key interactions, and CES around effort-heavy tasks. Each captures a different failure mode.

The blind spot all three share

The score records what happened but does not explain why. A 6-out-of-10 NPS response and a 2-out-of-10 response both count as detractors, although the underlying problems may differ. An open-ended "why did you give that score" field helps, but a static text box often produces short answers with no follow-up.

An AI-moderated follow-up can probe the reason behind a score. In Versive, routing based on NPS, CSAT, or CES can trigger an AI Question that asks an open-ended follow-up and adapts subsequent questions to the response. If a detractor says the product was "too slow," the interviewer can ask what "slow" meant in that context. See Skip logic and branching, explained for routing, and AI follow-up questions: probing without a moderator for probing behavior.

Combine scores and follow-ups in Versive

One study can collect all three metrics. Use the dedicated NPS question type for NPS, and a Rating Scale with context-specific wording for CSAT or CES. Conditional logic can route specific score bands into different AI Questions, so detractors and promoters receive relevant follow-ups. The results include chartable scores and transcripts for review or automated thematic analysis. See Survey question types, and when to use each for other structured types, and Thematic analysis, automated: from transcripts to themes for analyzing open-ended answers.

Choose the first metric

Choose NPS for loyalty, CSAT for a specific interaction, or CES for effort on a task. Add an open or AI-moderated follow-up when the score alone cannot explain a change.