How to measure customer satisfaction (CSAT, NPS, CES)

“Our customers seem happy.” That sentence steers nothing. To improve a service, you first have to measure it, with a method and metrics that stay comparable from one month to the next. Customer satisfaction has been quantified for more than twenty years with three standardized scores. Here is what they actually measure, their exact formulas, when to trigger them, and where a help desk tool comes in.

Why measure customer satisfaction?

An unhappy customer does not always complain, they just leave. Measuring satisfaction first serves to catch dissatisfaction before the cancellation, then to pinpoint specific friction points (a response delay, a billing step, a feature). The stakes are financial: in the foundational Corporate Executive Board study reported by Harvard Business Review, 96% of customers who had a high-effort experience became disloyal, versus only 9% of those who had a low-effort one. Measuring also makes support teams’ work objective and lets you prioritize improvement projects on data, not impressions.

In practice, a solid measurement approach comes down to three steps, to set up before you even pick a score:

CSAT: measuring immediate satisfaction

CSAT (Customer Satisfaction Score) measures satisfaction at a precise moment, right after an interaction: a resolved ticket, a purchase, a delivery. The typical question is simple: “How satisfied were you with this experience?”, with a scale (for example 1 to 5, or “very satisfied” to “very dissatisfied”).

The formula is direct:

CSAT (%) = (positive responses / total responses) × 100

A CSAT of 80% means 8 out of 10 customers said they were satisfied with that specific interaction. It is the most intuitive metric and the easiest to deploy, ideal for gauging a single touchpoint. Its limit: it captures a one-off, emotional reaction, often overstated right after an exchange that went well, and it says nothing about long-term loyalty.

NPS: measuring loyalty and advocacy

NPS (Net Promoter Score), created in 2003 by Fred Reichheld in Harvard Business Review, measures brand attachment through a single question: “On a scale of 0 to 10, how likely are you to recommend our product to a friend or colleague?” Respondents split into three groups:

• Promoters (score 9 or 10): loyal customers and advocates.
• Passives (score 7 or 8): satisfied but unenthusiastic, not counted.
• Detractors (score 0 to 6): at risk of leaving and of negative word of mouth.

The calculation:

NPS = % of promoters − % of detractors

The result is an absolute number between −100 and +100. As a benchmark, according to Retently’s 2025 data, the global average NPS sits around 32, with wide variation by industry (technology and software vendors often exceed 60, banking sits around 41 to 44, telecoms closer to 19). These figures are indicative (July 2026) and only mean something compared with your own industry. NPS is a relational, overall metric, best measured at regular intervals rather than after every ticket.

CES: measuring the effort asked of the customer

CES (Customer Effort Score), introduced in 2010 by the Corporate Executive Board, starts from a counterintuitive finding: reducing a customer’s effort builds loyalty more than trying to “delight” them. It measures how easy an interaction was through a statement to rate, usually on a 1 to 7 scale: “The company made it easy for me to handle my issue.”

The most common calculation is an average:

CES = sum of effort scores / number of respondents

On the 7-point scale, the cross-industry average sits around 5.5: above 6, the experience is genuinely smooth; below 4.5, friction pushes customers toward churn (indicative figures, July 2026). One variant is to calculate the share of “easy” answers (scores 5 to 7), where a “top-box” of 70% or more is considered solid. CES is an operational metric, perfect for tracking the friction points of a journey (self-service, a form, wait times).

CSAT, NPS or CES: which one should you pick?

The right answer is rarely “just one”. Each sheds light on a different dimension:

• CSAT, an emotional, one-off metric: trigger it right after an interaction (ticket, purchase).
• NPS, a relational, overall metric: measure it periodically to track loyalty.
• CES, a functional metric: trigger it after a specific action to spot friction.

By combining them, you get a 360-degree view that is at once emotional (CSAT), relational (NPS) and operational (CES). The trap is not the choice of metric, but the timing of collection: a survey sent too late, or systematically, drives down the response rate and biases the measurement.

What is the role of help desk software?

Measuring by hand, through emails and spreadsheets, rarely holds up beyond a few dozen customers. A help desk tool automates the chain: it triggers the CSAT survey at ticket closure, sends the NPS at regular intervals, and links each score to the agent, the channel and the contact reason. The result: metrics become trackable in a single dashboard, segmentable and usable to act on, rather than plain out-of-context percentages.

The main vendors (Zendesk, Freshdesk, Zoho Desk, Help Scout, Intercom) build these surveys in natively, most often billed per agent per month, or on quote for high volumes. To compare the offers and their survey modules, see our best help desk software 2026 comparison. If you are starting from scratch, our help desk hub breaks down the criteria that matter.