How to Interpret NPS, CSAT and CES Together
NPS, CSAT and CES answer three different questions. Here is the reading order for diagnosing a problem, and what it means when the three metrics disagree.
TL;DR
- The three metrics are not competing versions of the same measurement. NPS asks about the relationship, CSAT asks about a moment, CES asks about friction.
- Read them in a fixed order: NPS tells you whether something is wrong, CSAT and CES tell you where and why.
- The useful signal is disagreement. When all three move together you have confirmation. When they contradict each other you have a diagnosis.
- Each metric belongs at a particular point in the journey. Running all three everywhere produces noise and survey fatigue, not insight.
Most CX programmes end up running two or three of these metrics without ever deciding how they relate. The dashboard shows an NPS, a CSAT and a CES, all trending independently, and nobody can say what it means when one goes up while another goes down. That gap is where the value sits, because the contradictions between the three are more diagnostic than any of them alone.
The three metrics answer three different questions
NPS asks: would you stake your own reputation on us? It came from Fred Reichheld's 2003 Harvard Business Review article "The One Number You Need to Grow", and it is a relationship measure. Promoters (9 to 10) minus detractors (0 to 6), expressed as a points score from -100 to +100, never as a percentage. It aggregates everything a customer has experienced with you, which makes it a good alarm and a poor diagnostic. A falling NPS tells you something is wrong. It cannot tell you what.
CSAT asks: did that specific thing meet your expectations? It is the percentage of respondents choosing the satisfied end of the scale about a defined moment: a delivery, a resolution, an onboarding call. CSAT is about outcome quality at a point in time. It is narrow, immediate, and much easier to attribute to a cause than NPS.
CES asks: how hard did we make it? Customer Effort Score measures friction in completing a task, and Numr scores it on the net basis as a Net Easy Score: the percentage who found it easy minus the percentage who found it difficult, on a -100 to +100 range. Not an average of the scale, because an average lets delighted customers cancel out the ones who found you nearly impossible to deal with, and those are the ones about to leave.
Effort earns its place because of what it predicts. In "Stop Trying to Delight Your Customers" (Harvard Business Review, July 2010), the Corporate Executive Board team reported that 96 percent of customers who had a high-effort service experience became more disloyal, against 9 percent of those with a low-effort experience. Effort is the metric closest to the mechanics of defection.
Put simply: CES measures the machine, CSAT measures the outcome, NPS measures the relationship those two things add up to.
The reading order
Diagnosis works top down. Use the metrics in this sequence.
1. Start with relationship NPS. This is your alarm. Track it on a fixed cycle across the base and split by segment, tenure and region. If it is falling, you know a problem exists and roughly whose problem it is. Do not attempt to explain it from NPS alone.
2. Drop to touchpoint CSAT. Look at satisfaction at each measured moment for the affected segment. This localises the issue to a stage. A single touchpoint dragging its CSAT down while others hold steady is the fastest narrowing move available to you.
3. Drop to CES at the failing touchpoint. Effort tells you whether the problem is a broken process or a bad outcome. High effort at that stage means the process is the problem: too many steps, too many handoffs, too many repeats. Acceptable effort with poor satisfaction means the process works fine and the outcome is what customers dislike.
4. Read the verbatims last, filtered to the segment and stage you have narrowed to. Open text is expensive to analyse across an entire base and cheap to analyse once you know where to look. This is the order that keeps text analysis useful rather than overwhelming.
Most programmes run this backwards. They start with the open text, find every complaint under the sun, and end up with a list of issues nobody can prioritise. Scores narrow the search. Text explains what you find. Reversing the order buries the signal.
What it means when the metrics disagree
This is the part worth internalising. Agreement between the three is reassuring but tells you little. Disagreement is where the diagnosis lives.
High NPS, low CES
Customers rate you highly as a relationship, but find you difficult to deal with. This is common where the product is genuinely good, the brand is loved, or switching is expensive. Customers are absorbing friction because the alternative is worse.
The risk here is that people read the high NPS as permission to keep the friction. It is not. What you are seeing is goodwill being spent down. The moment a lower-effort competitor appears, the loyalty that was masking the friction stops masking it. Treat a low Net Easy Score under a high NPS as a countdown, not a contradiction.
High CSAT, low NPS
Individual interactions land well, but nobody would recommend you. Customers are satisfied and not loyal. This pattern shows up strongly in categories where service is competent but the overall value proposition is unloved: utilities, insurance, telecoms, subscriptions with lock-in.
The implication is that the problem is not operational. Fixing service faster will not move the relationship, because service is already fine. Look at price, contract terms, value delivered against expectations set at sale, or the competitive alternative. NPS is telling you about the deal, not the desk.
High CES, low CSAT
It was easy, and the answer was bad. Your process is efficient at delivering a result the customer does not want: a fast rejection, a smooth refusal, a quick escalation that resolves nothing. Effort measurement will not catch this, which is exactly why CES should never be run alone.
The fix is a policy or product question rather than a process one. Making a disappointing outcome even easier to receive does not help.
Low CES, high CSAT
The customer got what they wanted eventually and rated the outcome well, but the journey to it was punishing. This usually means an agent or a team is compensating personally for a broken process. It is fragile, expensive, and invisible in a CSAT-only view.
Watch for the delayed effect. Satisfaction is scored on the outcome in the moment. Effort is what customers remember three months later when a renewal notice arrives.
NPS falling, everything else stable
Nothing you measure has changed, so look outside the measured journey. Common causes: a price change, a competitor launch, a public incident, a policy change, or a channel you are not measuring at all. It is also worth confirming the instrument has not changed, because a new trigger, a new reminder schedule or a different responding population will move a score without any experience moving. That check is covered in what to track once the survey is live.
All three high, churn rising anyway
The uncomfortable one. Your surveys are reaching the wrong people. Customers who are leaving typically stop responding first, which means a healthy dashboard can sit on top of a shrinking, self-selected sample of the people who still like you. Check response rate by segment and compare responders against the full base before believing good news.
"The most useful review meetings we sit in are the ones where the three metrics disagree. A high NPS with a badly negative Net Easy Score is the pattern I trust least in a client's dashboard, because it looks like health and reads like a warning. Those customers are loyal to the product and tired of the company. That gap closes the moment somebody easier shows up." Amitayu Basu, CEO, Numr
Which metric belongs at which journey stage
Running all three at every touchpoint is the most common design error we see. It burns goodwill, drives response rates down, and produces three numbers where one would have done.
Journey stage | Primary metric | Why |
|---|---|---|
Onboarding or activation | CES | A difficult start predicts early churn better than a satisfaction rating does |
Purchase or checkout | CES | The task is transactional and friction is the whole story |
Delivery or fulfilment | CSAT | Outcome quality is what the customer is judging |
Service or support contact | CES first, CSAT second | Effort diagnoses the process, satisfaction confirms the outcome landed |
Renewal | CES | Renewal friction is a quiet and very reversible cause of loss |
Relationship, on a fixed cycle | NPS | The only place the aggregate relationship question belongs |
One metric per touchpoint, plus a relationship NPS on a cycle. That is a complete programme. Anything more is usually measurement for its own sake, and it costs you response rate at exactly the touchpoints where you most need the data.
Turning the pattern into action
Each disagreement points at a different owner, which is the practical payoff of reading the three together.
- Effort problems belong to operations and product. Count the steps, the handoffs and the repeat contacts at the failing stage.
- Satisfaction problems belong to the team that owns the outcome at that stage: fulfilment, service policy, or the product itself.
- Relationship problems belong to commercial and brand. Pricing, positioning, and the expectations set during the sale.
Route the finding to the owner along with the segment and stage it came from, then re-measure the same instrument to confirm the fix landed. If you want the wider view of how touchpoint measures roll into a single journey picture, see measuring the customer journey score. For what to make of the absolute numbers themselves, see what is a good NPS, CSAT or CES benchmark.
Frequently asked questions
Should we run all three metrics? Most mature programmes do, but not at the same touchpoints. Use one transactional metric per touchpoint, chosen by what that touchpoint is actually about, and one relationship NPS on a fixed cycle.
Which metric should we start with if we can only run one? If you are starting from nothing, relationship NPS gives you the widest early read on whether you have a problem. If you already know where the problem is, CES at that touchpoint gives you the most actionable diagnosis.
Can I combine NPS, CSAT and CES into one index? You can, and we would generally advise against it as your primary view. A composite hides exactly the disagreements that make the three metrics worth running. If you need a single number for reporting, build a journey score that keeps the components visible underneath.
Why is NPS not a percentage? Because it is the difference between two percentages, which produces a points value from -100 to +100. Promoters minus detractors. A source writing "NPS of 30%" has not handled the metric carefully.
Why score CES on the net basis rather than as an average? An average of a 7-point scale lets satisfied customers offset the ones who found you very difficult, and those groups do not behave the same way. The net method, percent easy minus percent difficult, keeps the difficult group visible.
Our NPS and CSAT tell opposite stories. Which one is right? Both, about different things. High CSAT with low NPS usually means service is competent while the overall proposition is not something customers would recommend. That is a commercial finding, not an operational one.
How long should we wait before believing a change? Long enough for at least two full measurement cycles at a stable sample size, with the instrument unchanged. Single-period movements in small samples are frequently noise, especially on net metrics where the range is wide.
Related guides
For related reading, see Demystifying NPS: How to Engage Detractors, Passives, and Promoters Effectively, Detractors vs. Promoters: How to Convert Your Biggest Critics into Brand Advocates, What Are CX KPIs? A Beginner's Guide for CX Teams, Inner Loop vs Outer Loop: The Blueprint for a Truly Actionable CX Program, and What Is Net Promoter Score? The Metric, and What to Do With It.