CSAT vs NPS

What is the difference between CSAT and NPS?

CSAT vs NPS is the comparison between two survey metrics support teams report: Customer Satisfaction Score, which scores a single interaction just after it happens, and Net Promoter Score, which scores a customer's willingness to recommend the company overall. They answer different questions and move on different timescales.

The scales differ before the interpretation does. CSAT runs 1 to 5 and reports the share of respondents choosing 4 or 5; NPS runs 0 to 10 and produces a single number between -100 and +100, so the two cannot be averaged or compared directly.

How CSAT and NPS are calculated

Both metrics begin with a trigger, and the trigger explains most of what follows. CSAT fires when a ticket closes, so the respondent is scoring a conversation that ended minutes ago. NPS fires on a schedule or at a relationship milestone, so the respondent is scoring an accumulated impression of the company.

The customer satisfaction score formula is top-two-box: count the ratings of 4 and 5 on the 1-5 scale and divide by total responses. Of 240 replies, 186 rated 4 or 5, which is 77.5%. NPS splits the 0-10 scale into promoters (9-10), passives (7-8), and detractors (0-6), then subtracts the detractor percentage from the promoter percentage for a figure between -100 and +100. Of 500 replies, 42% promoters against 26% detractors gives +16; the 32% passives sit in the denominator and never in the result.

Sampling decides what either number describes. Conversations closed by self-service rarely trigger a survey, so a rising deflection rate quietly changes which tickets get rated, and whoever owns the customer feedback loop owns that decision.

What each score counts and what it misses

Each instrument has a defined blind spot, and knowing it is most of the skill in reading the number.

• CSAT counts a moment: The rating covers one interaction, given within minutes of its end, before the resolution has been tested in real use.

• NPS counts an intention: Recommendation intent is a stated willingness to refer, collected away from any specific ticket, and it absorbs pricing and product alongside support.

• Neither counts non-respondents: Response rates on both skew toward the delighted and the furious, leaving the quiet middle of your base underrepresented in every reported figure.

• Neither counts unsurveyed conversations: Sessions that end in self-service, abandonment, or transfer often trigger nothing, so a growing share of contacts stays unrated.

CSAT vs NPS vs CES

Support teams run all three and then argue about which one belongs on the dashboard. CSAT measures whether one interaction satisfied the person who had it. NPS measures whether the whole relationship has produced an advocate. CES measures how hard the customer had to work to get the outcome. The three sit at different distances from the ticket, and CSAT vs NPS is the argument that matters most, because those two are the ones executives compare as if they were the same number.

The scales deepen the confusion. CSAT and CES both report a positive figure on a fixed scale, while NPS reports a difference between two percentages that can go negative, so a dashboard can stack all three as peers when they share no units. The timing differs as well: the first two arrive within hours of a conversation, and the third arrives on a quarterly cycle.

If you own a queue and need to know what to fix this week, CSAT and CES are the two that answer. If you own renewals and need to know whether the relationship is compounding, NPS is the one.

Why CSAT vs NPS matters for customer experience

A team that reports only CSAT can hold a high score for a year while renewals slip, because every individual conversation was handled courteously and none of them touched the reason the customer is leaving. A team that reports only NPS watches a quarterly number fall with no queue, no channel, and no policy attached to the drop, so the follow-up meeting is speculation.

Running both changes what each one is allowed to decide. CSAT drives operational work: coaching, macro rewrites, staffing a queue that scores badly. NPS drives account work, where it sits beside a customer health score in renewal forecasting.

The tradeoff is survey fatigue. Every additional survey lowers the response rate on the others, and a thin response base widens the range of true values a score is consistent with, which is how two consecutive quarters differ by several points for no operational reason.

How are CSAT and NPS benchmarked?

Benchmarking these two is asymmetric, and the asymmetry should shape how much weight each number carries in a board pack. Satisfaction has a neutral public reference: the American Customer Satisfaction Index publishes industry-level scores on a 0-100 scale, where most tracked industries sit in the 70s, the weakest sectors fall into the 60s, and the strongest reach the low 80s. That study uses a different instrument from a post-ticket survey, so treat its figures as directional context for your sector; your own top-two-box percentage will sit on a different footing.

Recommendation scores have no equivalent registry. Published figures come from vendors summarising their own customer bases, and sector norms diverge so widely that a cross-industry average carries almost no information. The defensible comparison for NPS is your own trailing series, measured with unchanged wording, an unchanged scale, and an unchanged sampling rule.

How AI agents change CSAT and NPS

Automation removes the simplest tickets first: password resets, order status, shipping windows. Those were also the tickets that scored highest, because they were short and ended in a clean answer. When they leave the human queue, what remains is longer, harder, and angrier, so agent CSAT can fall several points while nothing about the team's work got worse. Reading that drop as a coaching problem is the standard error, which is why teams are tracking AI CSAT separately from agent CSAT.

Two further effects show up. Surveys frequently never fire on autonomous resolutions, leaving the automated share of volume unrated and the reported score unrepresentative. And an agent that escalates on a low confidence score hands the customer a second wait, which lands in the CSAT of the human who inherits the case. NPS moves too slowly to register any of this inside a quarter.

How to choose between CSAT and NPS

Coverage comes first: work out what share of contacts can actually be surveyed. If much of your volume closes in self-service or an IVR, a post-ticket score describes a shrinking slice of the experience.

Integration surface is next. A score earns its keep when it lands back on the ticket, the agent, the queue, and the account record, so the segmentation you need at review time already exists.

Governance decides whether the trend survives. One owner should hold the question wording, the scale, and the sampling rule, because changing any of the three resets the series and no annotation makes the old and new numbers comparable.

Responses carry personal data and free-text comments, so buyers in regulated sectors ask where they are stored, how long they are retained, and whether the vendor holds ISO 27001 certification.

The constraint teams underestimate is suppression: without a rule capping how often one contact is surveyed across CSAT, NPS, and product research, your best customers receive three requests in a week and stop answering all of them.

CSAT, NPS, and the wider metric stack

Neither score explains itself. The free-text comment attached to a rating carries the reason, and conversational analytics is what turns thousands of those comments into ranked contact drivers with an owner attached.

Effort sits between the two. Customer effort score rates how hard a resolution was to obtain, which often explains a CSAT dip the rating alone leaves unexplained, and it tends to move earlier than a recommendation score does.

What does CSAT vs NPS mean in plain terms?

Think of CSAT as a receipt and NPS as a reference. CSAT stands for Customer Satisfaction Score, and it is the customer signing off on one transaction: that went fine. NPS stands for Net Promoter Score, and it is the customer deciding whether they would put their own name behind you in front of a colleague.

Someone can sign the receipt every time and still withhold the reference. The support conversations were fast and courteous, and the product still crashes on Tuesdays, so the ticket scores well while the recommendation question does poorly.

The tradeoff is speed against weight. CSAT gives you a number this afternoon that points at a queue you can fix by Friday. NPS gives you a number once a quarter that reflects a decision worth real revenue, with almost no instruction about what to do next.

Common CSAT vs NPS mistakes

Averaging the two, or expecting a move in one to appear in the other. They run on incompatible scales and different clocks: a percentage of top-box ratings collected daily against a difference of percentages collected quarterly. A chart that stacks them invites a comparison the arithmetic cannot support.

Setting agent targets on NPS. An individual agent controls none of pricing, roadmap, or account management, so an NPS goal on a scorecard produces gaming and demoralisation, and it teaches the team to distrust every survey number.

Surveying only clean closures. If the trigger fires on resolved tickets and skips escalations, transfers, and abandoned sessions, the score reports your successes back to you, and shifting automation and escalation patterns change that mix month to month.

Rewriting the question. Changing wording, scale, or reminder cadence mid-year creates a break in the series that looks exactly like a real movement, and nobody remembers the change six months later.