9 Customer Satisfaction Metrics to Track (and How to Calculate Them)

Customer satisfaction metrics are the numbers that show how satisfied customers are, from what they say in surveys and what they do afterward. Most teams track CSAT, NPS and CES from surveys, plus resolution, response-time and churn numbers from support and account data.

Metrics to measure customer satisfaction matter only when each is tied to a decision and checked against what customers do.

Below: nine metrics, each with its formula, a worked example, its best use, its catch and a recent benchmark where one exists. You also get a starter set by business model and the mistakes that skew the scores.

What Are Customer Satisfaction Metrics?

Customer satisfaction metrics are numbers that track how customers feel about, or act after, using a product or service. Revenue-only metrics such as net revenue retention, and team-productivity metrics, fall outside this group.

The table below shows all nine metrics with their group (survey, operations or behavior), what each measures, its formula and when to measure it.

Metric	Group	What it measures	Formula	When to measure
Customer Satisfaction Score (CSAT)	Survey	Satisfaction with one interaction	4s and 5s ÷ all responses × 100	Right after the interaction
Net Promoter Score (NPS)	Survey	Likelihood to recommend the company	% promoters (9 to 10) minus % detractors (0 to 6)	Every quarter or half-year
Customer Effort Score (CES)	Survey	How easy one task was	Average score, or 5 to 7 answers ÷ all responses × 100	Right after a task
First contact resolution (FCR)	Operations	Issues solved without a repeat contact	Resolved on first contact ÷ all issues × 100 (7-day window)	Weekly, by channel
First response time (FRT)	Operations	Wait for the first human reply	Median (and average) time to first reply, in business hours	Weekly, by channel
Average resolution time	Operations	Time from open to resolved	Total resolution time ÷ resolved tickets	Weekly, by issue category
Churn and retention rate	Behavior	Customers lost and kept	Churn: lost ÷ start × 100; retention: (end minus new) ÷ start × 100	Monthly or quarterly
Customer lifetime value (CLV)	Behavior	Gross profit an account brings over its life	Average revenue per account (ARPA) × gross margin ÷ monthly churn	Quarterly, by segment
Customer health score	Behavior	Churn risk from combined signals	Weighted sum of scored inputs (0 to 100)	Weekly or monthly, per account

What Types of Metrics Measure Customer Satisfaction?

Three types of metrics measure customer satisfaction: what customers say, what your team does and what customers do afterward. The first two lead because they move before customers act, and the third mostly confirms what already happened.

• What customers say: CSAT, NPS and CES are the customer satisfaction survey metrics. Attribute ratings belong here too: a rating of one feature or touchpoint, such as onboarding or billing. These scores only cover the people who answer.
• What your team does: first contact resolution, first response time and resolution time are the customer service metrics your support team controls. A slow or unresolved contact shows up here before the cancellation does.
• What customers do: churn, retention and customer lifetime value (CLV) lag, but they prove whether survey scores matter. The health score combines behavior signals to flag risk before renewal.

Customer satisfaction measurement techniques go beyond surveys, and these four methods feed the same three groups:

• Reviews and complaints: public reviews and logged complaints, read for recurring themes.
• Call and chat sentiment: tone and topics pulled from transcripts of support conversations.
• Support ticket data: reopened tickets and repeat contacts about the same issue.
• Product usage data: logins, feature adoption and drops in activity before renewal.

9 Key Metrics to Track, with Formulas and Worked Examples

The nine key customer satisfaction metrics to track are CSAT, NPS, CES, FCR, FRT, average resolution time, churn and retention, CLV and the health score. The revenue side of churn and retention sits in our SaaS retention metrics guide. Each one follows the same template, so you can compare them line by line.

1. Customer Satisfaction Score (CSAT)

What it measures: CSAT measures how satisfied a customer was with one interaction.

Question and scale: “How satisfied were you with [interaction]?” on a 1 to 5 scale.

Formula: top-2-box, responses of 4 or 5 ÷ total responses × 100. Some teams average the scores instead; never mix the two methods in one report. On a 2-point good/bad survey, compare the “good” share only with its own history, not with 5-point top-2-box or benchmarks built on it.

Worked example: 200 responses: 90 fives, 82 fours, 14 threes, 6 twos, 8 ones. 172 are 4s and 5s, so CSAT is 86%.

What good looks like: Simplesat’s 2026 CSAT benchmarks use vendor data from its own customers in 2025. Nearly two-thirds of those companies score 90% or above. By our reading, a CSAT around 90% sits in the middle of that data, so a high score rarely sets a team apart. Industry averages are in our CSAT benchmarks. Judge yours by trend over 3 or more periods and by touchpoint.

When to use it: right after a ticket, onboarding or delivery. After support contacts, some teams call it customer service satisfaction (CSS).

The catch: CSAT metrics measure the moment, not the relationship, so happy customers can still churn.

2. Net Promoter Score (NPS)

What it measures: NPS measures willingness to recommend you, a read on the relationship.

Question and scale: “How likely are you to recommend [company] to a friend or colleague?” on 0 to 10. Promoters answer 9 to 10, passives 7 to 8, detractors 0 to 6.

Formula: as Bain & Company defines it, % promoters minus % detractors, from -100 to +100. Passives count in the base only.

Worked example: 110 promoters (55%), 60 passives (30%) and 30 detractors (15%) of 200 give NPS 55 minus 15 = 40.

What good looks like: Survicate’s 2025 NPS benchmark puts the median at 42 across industries, 38 for B2B and 29 for B2B software.

When to use it: quarterly or half-yearly as a relationship check, not after each ticket.

The catch: one B2B account can mix a promoter champion and detractor users, so report NPS per role, with buyer NPS beside the pooled score. Score each account once, or report the share of net-detractor accounts (accounts whose own NPS is below zero), so large accounts don’t dominate. Read individual answers where respondents are few.

3. Customer Effort Score (CES)

What it measures: CES measures how easy one task was.

Question and scale: “[Company] made it easy to handle the issue,” rated 1 to 7, where 7 is strongly agree.

Formula: the average score (sum ÷ responses) or the share agreeing (5 to 7 answers ÷ responses × 100). Report only one.

Worked example: 37 sevens, 25 sixes, 16 fives, 12 fours, 5 threes, 3 twos, 2 ones from 100 responses. The average is 5.6 of 7; the share agreeing is 78%.

What good looks like: we found no recent public CES benchmark that states its scale, so track your own trend over 3 or more periods.

When to use it: right after setup, a support case or a billing change.

The catch: CES covers only the tasks you survey, so friction in an unsurveyed step such as renewal paperwork stays invisible.

4. First Contact Resolution (FCR)

What it measures: FCR measures the share of issues your team solves on the first contact.

Formula: issues resolved on the first contact ÷ all issues × 100. An issue counts as resolved only if no repeat contact arrives within a set window, such as 7 days.

Worked example: of 1,000 tickets, 720 were not reopened or repeated within 7 days. FCR is 72%.

What good looks like: SQM Group’s 2024 benchmark of call centers puts a good FCR at 70% to 79% and world-class at 80% or higher. For email and chat, treat it as a reference, not a target.

The catch: closing a ticket is not resolving it, so a fast-closing bot or agent inflates FCR unless repeat contacts count.

5. First Response Time (FRT)

What it measures: FRT measures how long a customer waits for the first reply to a new ticket.

Formula: sum of time to the first human reply ÷ tickets, in business hours; report the median next to the average.

Worked example: replies at 10, 25, 40, 90 and 480 minutes. The average is 129 minutes and the median 40; one slow ticket moves the average, not the median.

What good looks like: we found no recent, sourced benchmark by channel; track email, chat and phone separately.

When to use it: for staffing and channel decisions.

The catch: some help desks count an auto-reply as the first response, so measure to the first human reply.

6. Average Resolution Time

What it measures: average resolution time measures how long an issue stays open until it is solved.

Formula: sum of (resolved time minus opened time) ÷ resolved tickets, in business hours.

Worked example: 400 tickets took 3,200 business hours in total, so the average is 8 hours.

What good looks like: no universal number fits; billing and technical issues need different work, so track it by category.

When to use it: to spot a process bottleneck, such as one category that keeps getting slower.

The catch: rushed closures that cut resolution time can cut FCR too, so read the two together.

7. Customer Churn Rate and Retention Rate

What it measures: churn is the share of customers lost in a period; retention is the share of starting customers kept.

Formula: churn = customers lost ÷ customers at the start × 100. Retention = (customers at the end minus new customers) ÷ customers at the start × 100.

Worked example: start 200, lose 12, add 30, end at 218. Churn is 12 ÷ 200 = 6.0%; retention is (218 minus 30) ÷ 200 = 94.0%.

What good looks like: we found no reliable benchmark by business model, so compare your own periods on the same basis.

When to use it: in subscription and repeat-purchase businesses; with a small customer base, report churn quarterly or over a trailing 12 months.

The catch: only voluntary churn signals satisfaction. Recurly separates cancellations from failed payments, which say nothing about intent to stay.

8. Customer Lifetime Value (CLV)

What it measures: CLV estimates the gross profit one account brings before it churns.

Formula: (average revenue per account (ARPA) per month × gross margin) ÷ monthly churn rate. Convert quarterly churn first: monthly = 1 minus (1 minus quarterly) to the power 1/3. So 6.0% a quarter is about 2.04% a month. E-commerce teams use average order value × lifetime purchases instead.

Worked example: $500 a month at an 80% gross margin and 2% monthly churn gives a CLV of $20,000. At 3% monthly churn, it drops to $13,333.

When to use it: to decide how much retention and support work an account segment is worth.

The catch: CLV rests on today’s churn: in the example, one point more (2% to 3%) cuts it by a third. Show it as a range.

9. Customer Health Score

What it measures: a health score combines signals into one number that flags at-risk accounts.

Inputs: product usage (adoption and active use, the user satisfaction metrics a product team tracks), tickets, CSAT or NPS, renewal or payment signals and engagement.

How to calculate: score each input 0 to 100, weight it by how well it predicted past churn, then sum to 0 to 100.

Worked example: with equal weights, usage 80, tickets 60, latest survey input (scored 0 to 100) 40 and payment 100 average to 70. On our illustrative bands (0 to 49 red, 50 to 74 amber, 75 to 100 green), that is amber.

When to use it: in customer success teams deciding which accounts to call.

The catch: with few churned accounts, keep equal weights and revisit them once enough have churned to show which inputs moved first.

Takeaway: Read each survey metric next to the behavior metric it should predict. Put CSAT or CES next to repeat contacts and churn, and NPS next to renewals.

Which Satisfaction Metrics to Track First, by Business Model

Start with three satisfaction metrics, one from each group: a survey metric, an operations metric and a behavior metric.

Keep those three as your headline numbers and move the rest to a diagnostic layer you check only when a headline number moves. If no decision changes when a number moves, cut it.

The table below gives a starter set for four business models, with when to ask and who owns the numbers.

Business model	Survey metric	Operations metric	Behavior metric	When to ask	Who owns it
B2B SaaS	NPS by role, across accounts	FCR	Logo and revenue churn	NPS each quarter; CSAT after support	Customer success
E-commerce and B2C	CSAT after delivery or support	First response time	Repeat purchase rate; voluntary cancellations for subscriptions	Right after delivery or the ticket	Support or CX team
Contact center or support team	CSAT after the call or ticket	FCR and average speed of answer	Repeat contacts within 7 days	Right after the call or ticket	Support lead
Agency and services	NPS per client	Resolution time on requests	Client retention and renewal	Each quarter	Account lead

Logo churn is the share of customers lost in a period. Revenue churn is recurring revenue lost to cancellations and downgrades ÷ recurring revenue at the start of the period × 100.

Repeat purchase rate is the share of customers who buy again in a period. Average speed of answer is the average time a caller waits before an agent picks up.

Pick the survey metric by the moment you can change. Use CSAT for a single interaction, CES for the effort of one task and NPS for the overall relationship. Then add the behavior metric that proves it.

CSAT and NPS can also disagree. In Simplesat’s 2025 data, Technology tops satisfaction at 92.4% but sits seventh of eight industries on advocacy. So one survey metric cannot stand in for the other.

If leadership asks for one number, report the survey metric that fits your business model, with its response count and margin of error. Put the behavior metric beside it. A B2B SaaS line could read: NPS 40 from 50 responses (±20 points), quarterly logo churn 6.0%.

How to Measure Customer Satisfaction in 5 Steps

Customer satisfaction measurement takes five steps: tie each score to a decision, fix its timing and scale, then store it and act on low scores.

1. Tie each metric to one decision. Before you send a survey, write down what a low score would make you change. CES after onboarding, for example, decides whether you rework the setup flow.
2. Choose the moment and the cadence. How often should customer satisfaction be measured? Send transactional surveys (sent after one event) right after the interaction, and relational NPS (sent on a schedule, about the whole relationship) quarterly or half-yearly.
3. Pick the scale and keep it fixed. Keep each customer satisfaction scale and its direction fixed, and never compare scores across different scales unless you rescale them by range.
4. Connect scores to your CRM. Store the latest score and a rolling average on the company record, so account owners see the trend next to renewal and usage data. Segment scores by channel, plan, account and agent.
5. Close the loop on low scores. Our rule: contact every detractor and every CSAT 1 or 2 within 2 business days.

The table below maps four common journey moments to the metric we’d send; the timing column is our rule, not a sourced benchmark.

Journey moment	Metric	When to send
Onboarding done	CES	Within 24 hours of setup
Support ticket closed	CSAT	When the ticket closes
Quarterly relationship check	NPS	Once a quarter per contact; in a renewal quarter, 60 to 90 days before renewal instead
After cancellation	Exit question (why they left)	At cancellation

We’d cap it at one transactional survey per contact per 30 days and one relational NPS per contact per quarter. Asking more drives response down: in Retently’s 2025 response-rate study of 600 e-commerce brands, Q4 survey volume nearly doubled while response rates dropped 44%.

Tools to Collect and Track Satisfaction Scores

Customer satisfaction measurement tools fall into three groups, but check whether your CRM or help desk already sends surveys before you buy a survey tool.

• Survey tools: custom CSAT, NPS and CES surveys sent by email or shown in-app, with your own questions and timing.
• Help desk and contact-center software: built-in CSAT after a ticket or call, plus FCR, FRT and resolution time from ticket and call data. Help desks such as Intercom also report a CSAT response rate (rated conversations divided by all new conversations), worth watching next to the score. A shared inbox without ticket IDs can’t count repeat contacts, so start with FRT and add FCR once conversations are tickets.
• CRM fields and dashboards: scores next to renewals and revenue, the place to read survey and behavior metrics together.

How Many Responses Before a Score Change Is Real

A score change is real only when it is bigger than the margin of error at your response count. Small samples swing, so a move of a few points on a few dozen responses is usually noise.

The table shows 95% margins for one score and for a change between two waves of equal size, at 50 and 400 responses. CSAT uses a score of about 85%; NPS uses the example above (55% promoters, 15% detractors).

Metric	Responses	Score	Margin of error (95%)	Margin for a change between two waves
CSAT	50	85%	±9.9 points	±14.0 points
CSAT	400	85%	±3.5 points	±4.9 points
NPS	50	40	±20.4 points	±28.8 points
NPS	400	40	±7.2 points	±10.2 points

Comparing two waves needs about 1.4 times the single-score margin. Act on a move larger than the change margin at your response count, or on a trend that holds over 3 periods.

A move beyond that margin makes chance an unlikely explanation, but it does not rule out a change in who gets surveyed. If the channel mix, contact reasons or response rate shifted between waves, the score can move with no change for any customer.

Survicate set a 30-response minimum per survey in its 2025 survey completion benchmark, because below that a few respondents could shift the result. Our rule: show the response count next to every score, so readers can judge the move themselves.

In B2B, a quarter may bring answers from only a few dozen accounts. Read the comments and the accounts behind a change, not only the score.

How we worked this out: a CSAT share’s 95% margin is 1.96 × the square root of (p × (1 minus p) ÷ n). NPS uses 1.96 × the square root of ((p + d minus (p minus d) squared) ÷ n), times 100 for points. Use p for the CSAT or promoter share, d for the detractor share (both decimals), n for responses.

Where Surveys Fall Short (and What to Track When AI Handles the Conversation)

Surveys hear from a minority of customers, and AI-handled contacts change who answers and what they rate.

Who answers: customers with strong views answer most. In Retently’s 2025 e-commerce data, surveys customers opened themselves (a feedback button, QR code or signature link) reached a 97.61% response rate. That shows who wanted to answer, not how many would.

Resolved is not loyal: Gladly’s 2026 Customer Expectations Report surveyed 1,000 U.S. consumers who use AI customer support. Of them, 88% said AI, alone or with a human agent, resolved their issue. Only 22% said the experience made them prefer the company.

A bot “contains” a conversation when it closes it without a human, but that customer can still come back. Vendor case studies such as Zendesk’s Vagaro story report +5 CSAT points with no word on repeat contacts.

When a bot takes the easy contacts, blended CSAT can rise with no change for any customer. Split CSAT by bot vs human and by contact reason, then compare before and after launch on the same mix.

The fix: next to CSAT, report repeat contact within 7 days, reopen rate and escalation to a human. Keep them as separate lines rather than blending them into one weighted score.

Count repeat contact per customer across all channels, including bot chats abandoned without a survey. A customer who quits the bot and phones later never shows up as a ticket reopen.

Count an account as low if it is a detractor or gave a CSAT of 1 or 2. With a few dozen churned accounts a year, the churn gap between low-score accounts and the rest needs several quarters of data to mean anything.

Takeaway: In B2B, compare churn for accounts that scored low with everyone else’s; if it is no higher, the score isn’t predicting anything. For B2C subscriptions, compare cancellations within 30 to 60 days (our rule) for customers scoring CSAT 1 to 2 against those scoring 4 to 5.

Common Mistakes That Skew Satisfaction Scores (and How to Fix Them)

Five common mistakes skew satisfaction scores without any change in how customers feel. The table applies each mistake to the customer satisfaction metrics examples from the nine metric sections, then gives the fix.

Mistake	What it does to the number	Fix
Averaging CSAT in one report and top-2-box in another	The same 200 responses read 86% top-2-box but 4.2 out of 5 averaged (84% of the maximum), two numbers that look comparable and aren’t	Use one method and name it in every report
Flipping the CES scale between surveys	A 5.6 of 7 average reads as 2.4 when 1 is assumed to mean “easy”, an apparent collapse with no change in customers	Fix the direction and label it on every chart
Doubling a 5-point score to compare with a 10-point one	4.2 of 5 becomes 8.4 of 10, while rescaling the 1 to 5 range onto 0 to 10 gives 8.0	Compare like scales only, or rescale by range
Counting new customers in the churn base	12 ÷ (200 + 30) = 5.2% instead of 12 ÷ 200 = 6.0%	Divide by customers at the start of the period
Calculating CLV without gross margin	$500 ÷ 2% = $25,000 instead of $20,000, overstated by 25%	Multiply ARPA by gross margin first

How we worked this out: we recomputed every distortion from the article’s own worked examples with the same inputs. The arithmetic is ours, with no outside data.

Pick Your Three Metrics and Read Them Together

Start with three customer satisfaction metrics, one each from surveys, service operations and customer behavior, and judge them together. A survey score can rise while customers quietly leave, so trust no single number until the other two agree.

Your last survey’s rating counts are enough for the CSAT calculator to show the score with its margin of error. Which three would you start with, and which of them do you already have data for?