Customer experience metrics: five stars and a cancellation

TL;DR: Customer experience metrics come in three kinds: what customers say, what customers do, and what your operation did to them. Most programs over-collect the first, under-read the second, and never join them up, which is how a company posts rising customer satisfaction while its customers leave. Here's what each measures, where each misleads, and why metrics matter far less individually than in combination.

The survey arrived four minutes after the call ended. How did we do? I gave five stars, because the agent was patient, apologized twice for something that wasn't her fault, and stayed on the line while I crawled behind a cabinet for a serial number.

Instead of getting fixed, my problem kept getting escalated, and after three more calls chasing it I never heard back. Three weeks later I cancelled the account.

My five stars went into a customer satisfaction average and nudged it up. The cancellation arrived a month later in a churn report, owned by a different team on a different cadence, and nobody joined the two facts, because nothing was built to allow that.

More customer experience measurement has not helped

Companies have never tracked customer experience metrics this intensively, and the aggregate result is going backwards.

The American Customer Satisfaction Index has measured US customer satisfaction since 1994, covering 400-plus companies and conducting roughly 200,000 interviews a year. It put the national score at 76.1 in Q2 2026. In Q1 2013, it was 76.6. Thirteen years of platforms and survey invitations, and the number is worse than where it started.

That Q2 2026 reading was a sharp drop—ACSI says a decline that size has been surpassed only once this century, when the pandemic hit supply and prices.

Claes Fornell, who founded the index and is the world's most cited researcher on customer satisfaction, is blunt about why. Many companies, he writes, use performance metrics that are "too noisy or irrelevant." Some of them predict the direction of change in customer experience, stock returns or profit about half the time, which is what a coin toss does, as he points out.

There are more customer experience metrics than most businesses need, and most get collected at the level of a single ticket, read by the team whose bonus depends on the number, and answered by whoever felt like replying.

The three kinds of customer experience metrics

Useful measurement covers three separate questions, and each group of customer experience metrics answers exactly one. CX metrics is the usual shorthand, and only the complete picture tells you anything.

Perception metrics: what customers say

Customer Satisfaction Score (CSAT), Net Promoter Score (NPS), Customer Effort Score (CES) and customer sentiment. These capture how customers perceive the customer experience. They are cheap to collect but silent on what the customer does next.

Behavioral metrics: what customers do

Customer churn rate, customer retention rate and customer lifetime value. Nobody fills in a form to churn. These customer metrics record customer outcomes and resist flattery, which is why they lag.

Operational metrics: what your operation did

First contact resolution, average resolution time and response time. Contact centers already produce this operational data in volume, and these key performance indicators explain the other two groups.

High-performing organizations read all three together. The commonest failure in a customer experience program is reporting one group and inferring the rest.

Perception metrics: what customers say

Customer Satisfaction Score (CSAT)

CSAT asks how satisfied a customer was, usually on a scale of 1 to 5, and reports the percentage of satisfied customers in the top one or two boxes. It's the fastest way to measure customer experience at the level of one interaction, and the survey responses arrive as direct feedback the same day.

Its weakness cost my broadband provider an account. CSAT measures the conversation rather than the outcome, so a kind agent records a positive experience whether or not the problem died in a queue afterwards.

What good looks like. ACSI is the closest thing to an independent benchmark for overall satisfaction, putting quick-service restaurants at 79 in 2026, full-service at 82 and food delivery at 75 on a 0-100 index. Your internal CSAT percentage isn't comparable—worth saying whenever someone claims to beat an industry.

Net Promoter Score (NPS)

NPS asks how likely a customer is to recommend you, from 0 to 10, then subtracts detractors scoring 0 to 6 from promoters scoring 9 or 10. The result runs from -100 to +100. It's the most widely used customer experience measurement in circulation and reads customer loyalty, on the theory that loyal customers recommend while detractors warn people off.

Two cautions. A single number hides its composition, so a 30 built from passives behaves nothing like a 30 from equal promoters and detractors. And no independent body publishes cross-industry NPS data the way ACSI publishes satisfaction, so the benchmark tables come from vendors measuring their own clients.

Customer Effort Score (CES)

CES asks how much work the customer had to do, usually as agreement with a statement like "the company made it easy to handle my issue." Subtract difficult responses from easy ones for a directional read on friction.

CES would have caught me. My overall satisfaction was high and I had still been made to do the work repeatedly, and the second reading predicted my customer behavior better than the first. Effort is where customer expectations for the customer experience have moved fastest.

No independent body publishes a good CES benchmark, so treat your own trend as the target and watch which way it moves after each fix.

The difference between CSAT and CES is what each holds the company responsible for. CSAT rates how the interaction felt, CES rates how hard the customer worked, and NPS tracks more tightly with customer loyalty.

Customer sentiment analysis

Everything above depends on people answering surveys, and most don't. Customer sentiment analysis reads the unstructured material instead: chat logs, reviews, emails, transcripts from contact centers. Sentiment analysis tools use natural language processing to sort customer sentiment as positive, negative or neutral, often scoring intensity from 1 to 5. It's the cheapest way to hear customer needs in their own words.

The value is coverage. Surveys come from a self-selecting minority, while customer sentiment analysis reads everyone who wrote something, and natural language processing is cheap enough now to run across all your customer data. Treat the customer sentiment score as a signal to investigate, and mine the verbatim phrases for actionable insights about the customer journey.

Behavioral metrics: what customers do

Perception tells you how it felt. These customer experience metrics record what happened, and your CFO already trusts them.

Customer churn rate

Customer churn rate is the percentage of customers lost over a period: customers lost divided by customers at the start, times 100. It's the most honest number you have.

Customer churn is where your customer experience metrics get audited. My five stars and my cancellation were both true, and only the customer churn number predicted anything.

Customer retention rate

Customer retention rate is churn's mirror, counting the percentage who stay. Fornell argues retention at high levels drives exponential profit and revenue growth, and the index's own ETF of top-satisfaction companies has outrun the S&P 500 year to date, in a year the S&P hit record highs on record profits. High customer loyalty is the cheapest growth available.

Customer lifetime value (CLV)

CLV estimates the revenue an average customer generates before leaving, most simply as average revenue per user multiplied by expected customer lifetime. It converts a customer experience argument into a budget argument, because it prices what a rescued relationship was worth.

CLV also exposes cheap wins as expensive. Cutting handling time looks efficient until customer lifetime value falls in the cohort you rushed.

Customer health score and subscription metrics

Subscription businesses need a leading indicator, because customer churn reports what already happened. A customer health score combines product usage, support volume and customer engagement into one read on whether an account is drifting. Track it beside monthly recurring revenue, active customers, average revenue and new customers acquired, so your customer success team acts in the month that matters. Engaged customers show up here first.

Operational metrics: what happened inside your contact centers

Perception and behavior tell you something went wrong. Operational metrics tell you where. Contact centers generate this operational data whether anyone reads it, and your customer support teams already sit on it: every queue, every transfer, every one of your support tickets. Service delivery becomes measurable here.

First contact resolution

First contact resolution is the percentage of customer issues solved in the first interaction, and among key performance indicators it sits closest to how customers experience competence. My router case never reached first contact resolution, across four contacts and one cancellation.

Average resolution time

Average resolution time measures how long a case takes end to end, from first contact to real closure. Watch the median beside the mean, because a few month-long cases hide inside a healthy-looking average.

Response time

Response time, often reported as first response time, measures how long a customer waits for anything. It's the easiest customer experience metric to improve dishonestly, since an instant acknowledgment that resolves no customer issues moves the number without moving the customer experience.

Making the numbers catch each other out

Here's the part that would have saved my account. Ask of every metric you report: which other number would catch this one lying?

Leading companies tie customer experience metrics to business outcomes rather than reporting the two side by side. That separates a metric predicting revenue growth from one describing last month, and it's where the competitive advantages sit. Done properly you identify trends early enough to matter and identify areas worth funding before churn catches up.

1. Pair every perception metric with a behavior
   Never report CSAT without churn beside it, or NPS without repeat purchase. The pairing is what makes a rising score falsifiable, and it's how you measure customer experience rather than your own team.

2. Roll customer experience metrics up to the customer
   Three closed support tickets for one person is a bad customer experience wearing three green badges. Roll your customer experience metrics up by customer across the entire customer journey, then look at accounts with more than two contacts in a month.

3. Map every metric to a stage of the customer journey
   A number with no location is hard to act on. Attach each of your key customer experience metrics to a stage of the customer journey: response time against first contact, first contact resolution against the middle, customer churn against the exit. Read across the entire customer journey and the pain points cluster in one stage instead of looking like a bad month.

4. Segment before you average
   A national average hides the segment that's leaving. Split by plan, tenure, channel and region, and the friction usually sits in one.

5. Close the loop
   Route each piece of customer feedback to whoever can fix the cause, then tell the customer who complained what changed. Continuous improvement means that loop on a schedule, turning customer experience metrics into actionable insights and service strategies rather than a monthly slide.

How to actually run the measurement

Knowing which customer experience metrics to read is the easy half. Here's how to collect them so the numbers are worth reading.

When to trigger the survey

My survey fired four minutes after a call that resolved nothing, so it recorded the agent rather than the outcome. Send the request when the case closes rather than when the conversation ends, because customer expectations are set by outcomes. Give a delivery case another day or two, so the customer rates what arrived.

Transactional versus relational cadence

CSAT and Customer Effort Score are transactional. Fire them after individual customer interactions and read them weekly. Net Promoter Score asks about the whole relationship, so send it quarterly at most, to a sample. Running NPS after every ticket produces a number about the last five minutes.

How many responses before the number means anything

This part is arithmetic. For a large customer base, roughly 385 responses puts you within five percentage points at 95% confidence, and around 100 widens that to ten. Below that, a three-point monthly "improvement" is noise reported as progress. ACSI also warns that suppliers who "focus on responses to a single question" aren't doing adequate measurement, since good technique needs several questions per dimension.

Where the score should live

Trigger from the help desk, because that's the system that knows the case closed. Store the result against the customer record in your CRM. A score living only inside a standalone survey tool can never be joined to churn, and joining them is the entire point.

Where automation changes what you can measure

Most of the friction in my router case was repetition: four contacts, three re-establishing facts the company held.

This is the part Text was built for. Text's AI Agent connects to your order and account data, resolves routine questions in seconds at any hour, and hands the rest to a person with the full history attached. That lifts first contact resolution and cuts customer effort together, the combination these customer experience metrics exist to detect.

It also widens what you can read. Every one of your contact center interactions becomes text your sentiment analysis tools can process, so customer experience insights stop depending on who volunteered to answer a survey. On top of that, good self service options resolve customer issues before they reach a queue.

Back to the five stars

My rating is still in that company's customer satisfaction average, and it is still wrong. It recorded a woman who did her job well and said nothing about a customer who had decided to leave.

That's the risk when you measure customer experience metrics carelessly: they describe your teams instead of your customers. A positive experience and a satisfied customer are different things, and only a set where one number can be proven wrong by another will tell them apart.

Frequently asked questions

What is a good CSAT score?

A good CSAT score depends on the industry, and internal percentages aren't comparable across companies because survey wording, timing and response rates differ. As an external reference, ACSI put US national satisfaction at 76.1 out of 100 in Q2 2026 and quick-service restaurants at 79. Track your own customer experience trend rather than someone else's number.

What is a good NPS score, and is 72 a good NPS score?

On the -100 to +100 scale, anything above zero means more promoters than detractors, and a score in the 70s sits near the top of any published table. Treat 72 as excellent and cross-industry comparisons as unreliable, because no independent body publishes NPS benchmarks the way ACSI publishes satisfaction data.

How do you calculate Customer Effort Score?

Ask how much effort a task required, usually as agreement with a statement like "the company made it easy to handle my issue" on a five or seven point scale. Subtract the percentage giving difficult answers from those giving easy ones. A higher CES means customers did less work.

What are the 5 key performance indicators for customer service?

The five customer experience metrics covering the most ground are CSAT for how an interaction felt, Customer Effort Score for how hard the customer worked, first contact resolution for whether it got solved, average resolution time for how long, and customer churn rate for whether they stayed. Those four explain the fifth.

What is the difference between CX metrics and customer metrics?

CX metrics measure the quality of customer interactions and how customers perceive them, covering satisfaction, loyalty and effort. Customer metrics is the wider category, taking in acquisition cost, demographics, revenue per account and other business metrics. CX metrics tell you how customers feel about the customer experience; customer metrics describe the commercial relationship.