CSAT, NPS, and CES Explained: When to Use Each (September 2026)

Sep 22, 2026 by Marcos Dymond, Head of Growth

Your dashboard shows CSAT climbing, NPS flat, and CES trending the wrong way, and now someone wants to know which number to bring to the executive readout. The three metrics measure different things on different timelines, and treating them as interchangeable is how programs quietly lose credibility. Here's a plain read on when each one earns its spot.

TLDR:

• Use CSAT for touchpoint execution, NPS for quarterly loyalty reads, and CES for friction on self-serve flows
• CSAT = top-two-box % of responses; NPS = %Promoters minus %Detractors; CES rates task ease on a 1 to 7 scale
• Cap surveys per customer per quarter and suppress NPS if a CSAT prompt fired in the last 14 days
• Read scores against category benchmarks (retail NPS 40-50, B2B software 30-40, telecom below 30 per Retently 2026)
• Merciv joins survey verbatims with reviews, social, syndicated data, and internal docs into one cited finding with a three-tier confidence score

What CSAT Measures and How It Works

Customer Satisfaction Score (CSAT) is a transactional metric. It captures how a customer feels right after a specific interaction: a support ticket closing, a checkout, a delivery, an onboarding call.

The standard CSAT survey asks one question, usually "How satisfied were you with [the interaction]?" Respondents answer on a 1 to 5 scale, and the score is the percentage who picked the top one or two boxes.

A typical CSAT survey looks like this:

• Trigger: fires within minutes or hours of the touchpoint
• Question: one rating question, often paired with an optional open-ended follow-up
• Scale: 1 to 5, where 4 and 5 count as "satisfied"
• Output: a percentage reported per touchpoint, team, or channel

Because the survey lives close to the event, the score reflects execution quality on that interaction, not how the customer feels about your brand overall.

How to Calculate CSAT (Formula and Example)

CSAT = (satisfied responses / total responses) × 100

"Satisfied" almost always means top-two-box: 4s and 5s on a 1 to 5 scale, or 6s and 7s on a 1 to 7 scale.

If a support team sends 500 post-ticket surveys, gets 500 completed responses, and 420 pick 4 or 5, CSAT = (420 / 500) × 100 = 84%.

• 1 to 5: the common convention, easy for respondents, top-two-box standard
• 1 to 7: adds granularity for tracking small movements
• Top-box only (5s on a 5-point scale): a stricter cut that separates delighted from merely satisfied, useful when 4s and 5s cluster and hide variance

Pick one convention and hold it. Switching scales mid-year breaks customer satisfaction measurement trend comparisons that took quarters to build.

What NPS Measures and How to Calculate It

Net Promoter Score (NPS) is a relational metric. One question: "How likely are you to recommend us, 0 to 10?" Recommendation is a loyalty proxy, not a read on the last touchpoint.

Responses split into three bands:

• Promoters (9-10): loyal, likely to refer
• Passives (7-8): satisfied but unattached, exposed to competitor switching
• Detractors (0-6): unhappy, likely to churn

NPS = %Promoters minus %Detractors, landing between -100 and +100. With 55% Promoters, 30% Passives, 15% Detractors, NPS = +40. Passives count toward the base but not the score, which is why a program can lift satisfaction without moving NPS: you shifted Detractors to Passives without earning Promoters.

What CES Measures and Why Effort Predicts Retention

Customer Effort Score (CES) is a transactional metric asking how easy it was to complete a task or resolve an issue, typically rated 1 to 7 from strongly disagree to strongly agree with a statement like "The company made it easy for me to handle my issue."

CES exists because effort tracks retention better than satisfaction. The original HBR research behind CES found high-effort interactions produced disloyalty at a much higher rate than low-effort ones.

CSAT captures whether a customer walked away satisfied. NPS captures whether they would recommend you. CES captures what it cost them to get there.

CSAT vs NPS vs CES: The Core Differences at a Glance

Read the table as a routing guide. Post-ticket, post-delivery, post-onboarding: CSAT or CES. Quarterly relationship health: NPS. Friction diagnostics on a self-serve flow: CES. Selecting the right voice of customer tool shapes which of these you can run at scale. Friction diagnostics on a self-serve flow: CES.

When to Use CSAT (and When It Falls Short)

Use CSAT when the question is "did we execute this touchpoint." Post-ticket resolution, post-purchase, post-onboarding, post-delivery, and feature-launch feedback are where it earns its place. The score maps to a specific workflow you can fix.

Where CSAT falls short:

• It captures a moment, not a relationship. A customer can rate a return 5/5 the week they cancel.
• It skews positive, since satisfied and low-effort respondents dominate the sample.
• It does not predict retention or referral on its own.

For context, ACSI data puts the U.S. average near 77-78 on the 100-point scale, with retail landing in the low-to-mid 80s and telecom often below 70, per ACSI benchmarks by industry. Read your number against your sector.

When to Use NPS (and Where It Gets Misused)

Run NPS as a relationship pulse, not a transactional read. The right cadence: quarterly or biannual relationship surveys, a post-90-day tenure check, and annual brand awareness tracking. More frequent turns respondents into a fatigued panel.

Where it gets misused:

• It flattens a full distribution into one number, hiding the shape underneath.
• Small samples make it noisy; a handful of Detractors can swing a score by 20 points.
• Tied to comp, it gets gamed. Reps coach customers toward 9s and 10s, and the number decouples from reality.

For 2026 calibration, B2B software commonly lands around 30 to 40, retail 40 to 50, and telecom often below 30, with a roughly 11-point gap between B2B and B2C, per Retently's 2026 NPS benchmarks.

When to Use CES (and What It Cannot Tell You)

Use CES where friction is the improvement lever. Post-support-ticket, post-self-service attempt (help center, chatbot, returns portal), post-checkout, and post-account-setup are the native spots.

Where CES falls short:

• It ignores emotional connection. A customer can rate a cancellation flow "very easy" on their way out. Low effort does not surface the consumer insights needed to understand why they left — a gap insights teams are often left to close without the right tooling.
• Low effort does not equal loyalty. Easy is table stakes, not a differentiator.
• It has nothing to say pre-purchase or on brand perception.

Rule of thumb: CES answers "did we make this easy," CSAT answers "are they happy," NPS answers "will they stay and refer."

How to Combine CSAT, NPS, and CES in a Single Measurement Program

The layered approach treats each metric as a different lens on the same lifecycle: CSAT and CES fire at the transaction level to grade execution and friction, NPS runs at the relationship level to read loyalty.

A workable cadence:

• CSAT survey after every support ticket and post-purchase
• CES after onboarding, returns, cancellations, or first self-serve attempt
• NPS score at 90 days tenure, then quarterly

Ration what fires per customer: suppress NPS if the account got a CSAT prompt in the last 14 days, cap total surveys per customer per quarter, and rotate open-ends. Survey cadence is only one input into measuring consumer sentiment accurately. Asking less, more deliberately, returns cleaner data than asking everything.

Reading Your Scores in Context: Benchmarks and Category Nuance

A raw score in isolation is close to useless. An 80% CSAT is unremarkable for a luxury hotel and exceptional for an ISP. Read your number against your category, not the internet average.

Three benchmarking anchors practitioners actually use:

• ACSI for cross-industry CSAT calibration on the 100-point scale
• Category-specific NPS reports (SaaS, retail, telecom) that reflect structural score gaps
• Your own historical baseline, quarter over quarter, on a fixed scale and sample definition

Do not compare a 5-point CSAT to a 7-point CES to a 0-to-10 NPS as if they shared an axis. Normalize or do not compare.

One more piece of context: 39% of brands saw a meaningful CX decline in 2024, per Forrester's U.S. CX Index. A flat year-over-year score against that backdrop is quiet outperformance.

Common Mistakes That Break Your Metric Program

Programs break in predictable ways:

• Survey fatigue. Firing the same NPS at the same customer every 30 days trains them to click through. Cap surveys per customer per quarter.
• CSAT tied to agent comp. The moment scores drive bonuses, reps coach customers toward 5s and the number decouples from the experience.
• NPS as a headline number. A +42 hides a rising Detractor base without the verbatims underneath.
• CES on the wrong interaction. Effort surveys on brand touchpoints produce noise.

Then sampling. Roughly 1 in 26 unhappy customers complains; the rest churn quietly. Treating support tickets as a consumer intelligence feed closes that gap. An 88% CSAT on a 12% response rate is a survivorship artifact, not a health signal.

Beyond the Score: Why the "Why" Behind Each Metric Matters Most

A CSAT of 82%, an NPS of 45, a CES of 5.2: on their own, none tell you what to do Monday. You need to know which attribute, channel, or competitor moved the number before it becomes a decision.

Getting to the "why" takes three habits:

• Tag open-ended verbatims against a stable code frame (price, fit, support wait, packaging) so themes stack quarter over quarter.
• Pair survey data with unsolicited signal: review verbatims, social conversation, support transcripts.
• Triangulate against behavioral data: repeat purchase, churn cohorts, ticket volume, self-serve drop-off.

Most programs stall here. The score is instrumented, the dashboard ships, and no one has built the workflow connecting a 4-point NPS drop to a defensible cause. The number moves, someone asks why in the QBR, and the answer is a guess in a slide. A customer insights strategy built around decisions prevents that outcome.

How Merciv Connects Your Score Movements to the Consumer "Why"

A score moves. The question is why, and whether you can defend the answer.

Merciv joins survey verbatims with cross-retailer reviews, social conversation, licensed syndicated data, and internal documents into one cited finding. Every claim carries a three-tier confidence score (High, Directional, Exploratory) and a clickable audit trail to the source verbatim, so a CSAT drop traced to a reformulation lands as evidence.

When a threshold trips on ingredient claims, complaint clusters, or hero-SKU sentiment, the brief routes to the metric owner the morning it happens.

Final Thoughts on Building a CSAT, NPS, and CES Program That Holds Up

The metrics are simple; the discipline around them is where most programs break. Fire your CSAT survey where execution matters, keep NPS on a relationship cadence, and use CES where friction is the lever you can pull. Once the instrumentation is clean, the harder question becomes tracing every score movement back to a cause you can defend. Merciv's enterprise workflow is built for that step, joining verbatims with reviews, social, and syndicated data into a cited finding.

FAQ

Should you use CSAT, NPS, or CES to measure customer experience in a CPG program?

Use all three, mapped to different moments: CSAT after a specific touchpoint (post-ticket, post-delivery), CES where friction is the lever (self-serve, returns, onboarding), and NPS as a quarterly or biannual relationship pulse. Treating one score as a universal health signal is where most programs stall, because each metric answers a different business question.

What's a good NPS score for CPG and retail brands in 2026?

Category context matters more than the raw number. Retail commonly lands around 40 to 50, B2B software around 30 to 40, and telecom often below 30, per Retently's 2026 NPS benchmarks, with a roughly 11-point gap between B2B and B2C sitting inside those ranges. Read your score against your sector and your own historical baseline, never a cross-industry average.

Can you tie CSAT scores to agent compensation without breaking the metric?

Not without breaking it. The moment scores drive bonuses, reps coach customers toward top-box responses and the number decouples from the experience. An 88% CSAT on a 12% response rate is already a survivorship artifact before comp pressure enters the picture. Use CSAT for workflow diagnosis, and measure agents on resolution quality signals that are harder to game.

How do you figure out why a CSAT or NPS score moved instead of guessing in the QBR?

Tag open-ended verbatims against a stable code frame so themes stack quarter over quarter, then triangulate survey data with unsolicited signal (reviews, social conversation, support transcripts) and behavioral data (repeat purchase, churn cohorts, self-serve drop-off). Merciv's synthesis layer joins survey verbatims with cross-retailer reviews, social conversation, licensed syndicated data, and internal documents into one cited finding, so a CSAT drop traced to a reformulation arrives with a three-tier confidence score and a clickable audit trail to the source verbatim.

When does a CES survey give you the wrong read?

CES misfires on brand touchpoints and pre-purchase moments, where friction isn't the improvement lever. A customer can rate a cancellation flow "very easy" on their way out the door. Keep CES on post-support-ticket, post-self-service, post-checkout, and post-account-setup, and use CSAT or NPS where emotional connection or loyalty is the actual question.