NPS vs. CSAT vs. CES: Which One Actually Predicts Churn

"Which metric should we track — NPS, CSAT, or CES?" is a common question with a mostly wrong framing. The three metrics measure different things at different points in the customer journey, and current research shows the choice matters less than using the right one for the right question. Here's what the data actually says about each.

The false-choice framing, and why it's wrong

The standard framing "NPS vs CSAT vs CES" is misleading because the three metrics don't compete to answer the same question. NPS reads the overall relationship, CSAT reads a specific moment, and CES reads the friction involved in getting something done (customergauge.com). Comparing them head-to-head as if choosing one replaces the others is described directly as unfair — they're connected, complementary, and best used together at different stages of the customer lifecycle (customergauge.com).

That said, when a single number is needed for churn prediction specifically, the data does show real differences in predictive strength by context.

What each metric actually measures

CES: the strongest predictor in support-heavy, service-oriented contexts

The single most consistent finding across current sources: CES is the strongest predictor of churn at the relationship level in B2B subscription businesses, and specifically the most actionable metric for predicting churn in support-intensive product categories (SHNO). Customers who experience high effort to resolve an issue are 4x more likely to churn than those with low-effort resolutions (SHNO).

The underlying mechanism, stated directly in the research: high-effort experiences drive disloyalty far more than delight drives loyalty (SHNO). This is a meaningfully asymmetric finding — it implies that reducing friction in support interactions has more retention upside than trying to create delightful "wow" moments, because the downside risk of high effort outweighs the upside of high delight.

CSAT: stronger than NPS specifically in B2B SaaS

Counter to NPS's reputation as the default "loyalty" metric, multiple Forrester and Harvard Business Review analyses find CSAT correlates with retention 2–3x more strongly than NPS in B2B SaaS specifically (SHNO). This is a fairly specific and counterintuitive finding worth flagging — NPS is often treated as the "real" loyalty benchmark across industries, but the B2B SaaS-specific data suggests CSAT, measured at the right touchpoints, tracks retention behavior more tightly in this category.

NPS: weaker direct correlation, but strong signal at the extremes

Research on NPS-to-retention correlation found only a 54% correlation between NPS and retention likelihood — a fairly modest predictive relationship on its own (customergauge.com). However, the relationship isn't linear or weak everywhere: customers rating 9–10 (promoters) show 90% higher retention than detractors (0–6 raters) (customergauge.com). This means NPS is much more useful as a segmentation tool at the extremes (identifying your most at-risk detractors and most loyal promoters) than as a smooth, continuous predictor across the full 0–10 scale.

NPS also correlates with broader business outcomes beyond churn specifically: companies with higher NPS see roughly 2.5x more revenue growth than lower-NPS companies (customergauge.com) — a business-health signal, not necessarily a precise individual-account churn predictor.

NPS correlation with retention:        ~54% (moderate)
NPS: promoters vs. detractors retention: +90% for promoters
CES: high-effort vs low-effort churn:   4x more likely to churn (high effort)
CSAT vs NPS retention correlation
  (B2B SaaS specifically):              CSAT 2-3x stronger

Why most companies end up tracking more than one

CustomerGauge research found 49% of companies using NPS also track an additional metric — most commonly CSAT (33% of NPS users) or CES (about 1 in 6) (customergauge.com). This reflects the "different question, different layer" reality rather than metric redundancy: a company might use NPS quarterly to gauge overall relationship health, CSAT after specific interactions (support ticket close, onboarding completion) to catch moment-level dissatisfaction, and CES specifically around support/service touchpoints where friction is the primary churn risk.

A practical decision framework

Given the data above, the choice isn't "pick the best metric" — it's "match the metric to the touchpoint":

• Support-heavy, service-oriented product (e.g., a platform where users frequently need help completing tasks) → prioritize CES. The 4x churn multiplier on high-effort experiences is the clearest actionable signal available.
• B2B SaaS with infrequent touchpoints (annual contracts, low support volume) → prioritize CSAT at key moments (onboarding, renewal conversations) over NPS, given its stronger B2B SaaS-specific retention correlation.
• Tracking overall account health / expansion readiness → use NPS, but treat it as a segmentation tool (isolate detractors for intervention, promoters for expansion/referral) rather than a precise churn-probability score on its own.

Actionable takeaway

1. Don't treat this as a single-metric decision — NPS, CSAT, and CES measure different things and the data supports using multiple metrics at different lifecycle stages, which is what most mature companies already do.
2. If your product is support-intensive, prioritize CES — it has the strongest documented churn correlation (4x) of the three metrics reviewed here.
3. In B2B SaaS specifically, weight CSAT over NPS for retention prediction — the correlation is 2–3x stronger per Forrester/HBR analysis, despite NPS's broader reputation as the default loyalty metric.
4. Use NPS for segmentation, not fine-grained prediction — the 54% correlation with retention is moderate, but the promoter/detractor split (90% retention gap) is a strong signal for targeting interventions.
5. Instrument metrics at the touchpoint they're designed for — CES after support interactions, CSAT after discrete moments (onboarding, feature adoption), NPS on a periodic relationship-level cadence.

Sources: SHNO — Customer Satisfaction Statistics for 2026, customergauge.com — NPS vs CES vs CSAT: Which Customer Experience Metric To Use?, customergauge.com — Net Promoter & Customer Experience Benchmarks