NPS Is Not the Number You Think It Is

You want one number. One clean line on a dashboard that tells you if customers love you. A CEO can quote it, a board can track it, and a bonus can hang off it. That pull is why so many teams crown NPS as the number and stop asking what it measures. The trouble is a single loyalty score hides more than it shows. Read it wrong and you steer the whole team by a number that means almost nothing.

This brief is about knowing what each metric actually measures before you bet a decision on it. Not throwing NPS out. Knowing its job, its blind spots, and what to pair it with so you are reading customers, not a magic trick.

The deep cut

• A metric measures one moment, not the truth. CSAT rates the last call, CES rates one task, NPS asks about a promise. None is the whole customer.
• Intent is not behavior. Jared Spool showed one shopper who spent the most while scoring an 8 and the least while scoring a 9.
• Read the words, not just the score. Intuit's former CEO learned the open-ended answer taught him more than the rating.

Why the one-number dream falls apart

NPS started with a big promise. In 2003, Fred Reichheld called it "the one number you need to grow." It stuck because it fits what leaders want: easy to measure, easy to track, feels legit. But the math is strange. Move a whole set of scores from 0 to 6 and NPS stays at -100, as if the work never happened. Bump them to 9 and the score leaps to 100. Small real gains show as nothing, then jump for no reason.

The question is asking about the future, not the past. "How likely are you to recommend us" is a guess about behavior, not a record of it. Good research asks what people actually did. Spool calls the wild score swings "Analytics Theatre": drama in the numbers that does not help you build a better product.

What each metric is actually for

Stop treating these as rivals and start treating them as tools with different jobs. Jeremy Watkin lays out the split cleanly. CSAT asks if someone was satisfied, usually right after a support call, on a 5-point scale. NPS asks about willingness to recommend the whole company, on a 0-to-10 scale. CES asks how much effort it took to get something done, on a 1-to-7 scale.

The scope is the difference. CES zooms in on one task, like resetting a password or getting a ticket resolved. CSAT and NPS zoom out to the relationship. As the LogRocket team puts it, CES "looks at the interaction in a specific part of the process," while NPS and CSAT give a wider view. Ask the wrong one at the wrong moment and you get noise. Ask NPS after a routine bank transfer and you have learned nothing worth acting on.

Effort tells you where to fix things

When you want a signal you can act on, effort beats recommendation. The idea traces to the HBR piece "Stop Trying to Delight Your Customers," where researchers found that reducing effort predicts loyalty better than wowing people. A high-effort experience is where customers quietly decide to leave, so CES points you at the exact broken step.

This is why some argue startups should pick CES over NPS. Dhananjay Garg makes the case that CES better predicts loyalty because it tracks how easy a real transaction was, not a mood. Fire a CES survey after onboarding, after a purchase, after a support case. Each one hands you a place to look and something to fix. That is more useful on a Monday than a company-wide number that moves for reasons nobody can name.

Pair the score with behavior and the words

A score alone lies by leaving things out. Spool's example says it best: one shopper's spending had no link to their NPS answers. They spent the most on an 8 and the least on a 9. So watch what people do. Netflix did this in its early years by asking new subscribers if a friend referred them, and asking existing ones if they had referred anyone in the last six weeks. Both questions track real past behavior, and both tied straight to growth.

Then read the open text. Intuit's former CEO admitted the rating mattered less than the answer that followed it. Watkin's four-step loop is worth copying: collect the verbatim comments, sort each one into People, Policy, Product, or Process, close the loop with the customer and the agent, then act on the pattern. Josh Seiden pushes the same instinct further, asking which user behaviors lead to a high score so you know what to change instead of just chasing a target.

Build a small stack, not a single number

One survey question cannot carry a business. Zendesk lists a wider set worth watching: conversion rate, retention, churn, and lifetime value sit next to satisfaction scores. Lifetime value and churn are outcomes with money attached. A rising LTV is a real result. An NPS of 100 is not, unless it moves those numbers too.

Jorge García-Luengo frames this as families of KPIs across financial, sales, experience, and operations, so a survey score connects to revenue instead of floating alone. Keep your stack small. One survey metric matched to the right moment, one behavior metric to check if people act on what they say, and one money metric to confirm it matters. Do not tie bonuses to any survey score, or someone will game it and you will lose the truth you were paying for.

Three questions for your team

• Which metric matches each moment in our journey, and where are we asking the wrong one at the wrong time? Map CSAT, CES, and NPS to real touchpoints before the next survey ships.
• Where in our flow is effort highest, and what would it take to cut it? Put a CES survey after onboarding and support, then fix the worst step.
• What behavior actually predicts our high scores, and are we tracking it? Pick one past-behavior signal, like referrals or repeat purchase, to sit next to the survey number.