How you measure CX is Preposterous

2002 Nobel prize laureate Daniel Kahneman said

"For some of our most important beliefs we have no evidence at all, except that people we love and trust hold these beliefs. Considering how little we know, the confidence we have in our beliefs is preposterous - and it is also essential."

And then I thought about NPS.

And I laughed. Then felt sad.

It was hilarious and all too real at the same time.

Nobody woke up one day and decided CSAT and NPS were the twin pillars of customer experience. They didn't come down from a mountain on stone tablets.

They came from a boss. Who got them from their boss. Who read a Harvard Business Review article decades ago and became a true believer.

We trusted these people. We respected them. They were good at their jobs. So we inherited their beliefs the same way we inherit everything else we don't consciously choose: uncritically.

Ask most CX leaders why they track NPS and you'll get one of three answers:

• "It's the industry standard."

• "It's what our leadership expects."

• "It's how we've always measured loyalty."

Which is better than not having a metric.

But, what’s missing is evidence. Evidence that a single number, meaningfully predicts growth, retention, or the quality of the experience you're delivering.

The honest answer (in the very best case) is: sometimes. In some industries. Under some conditions. With a lot of caveats.

That's not nothing. But it's also not the bedrock certainty with which most organisations treat it.

The CSAT problem is even more preposterous.

At least NPS is trying to predict something forward-looking.

CSAT asks someone how satisfied they were immediately after an interaction. Which is a snapshot of a feeling, in a moment, scored on a scale that different customers interpret completely differently.

A 4 out of 5 from one person is reluctant disappointment and genuine happiness from another.

We then aggregate those scores, average them, put them on dashboards, and make decisions based on them.

And we do this because the leaders who shaped our careers believed in it. Because it feels like evidence. Because having a number, any number, is more comfortable than sitting with the messiness of not knowing.

So why haven't we changed?

Because the belief is also essential.

You can't run a CX function in a state of total uncontrolled chaos. You need to track something. You need to report something to the business. You need a shared language with your peers, your stakeholders, your board.

CSAT and NPS, whatever their flaws, give you that. They're a common language. And the confidence we place in them, even the slightly inflated, is what allows organisations to move, to set targets, to hold people accountable.

Preposterous, but essential. Kahneman was right.

Now add AI to the mix.

The question is, now, what you do with this realisation?

Here's where CX gets interesting, and tricky.

AI is transforming how customers interact with businesses faster than most teams can track. Automated conversations, intelligent routing, predictive service, real-time sentiment analysis, predictive escalation detection. The technology is moving. And fast.

But the metrics aren't.

You'll have seen this in every AI vendor pitch deck: “How do we use AI to improve our NPS?" Or: "Can AI help us drive CSAT up?"

The risk is AI accelerates our ability to optimise for the wrong things. If your belief system says CSAT is the goal, AI will help you chase CSAT with extraordinary efficiency. You'll move faster in the wrong direction.

The opportunity is AI actually gives us, for the first time, the ability to build better belief systems. Genuine signals such as conversation patterns, effort indicators, churn predictors, sentiment trajectories that could replace the proxy metrics we've been relying on with something closer to truth.

But only if we're willing to question the beliefs we inherited in the first place.

This isn't a call to tear up your dashboards on Monday morning. That would be reckless. (The essential part of his quote matters as much as the preposterous part.)

It's a call to hold your metrics with a slightly looser grip. To ask:

• What behaviour does this metric drive? And is that the behaviour I actually want? What are the second order impacts of this behaviour?

• If I were starting from scratch today, with everything AI can now give me, would I build the same measurement system?

The CX leaders who will navigate this transition well aren't the ones who adopt AI fastest. They're the ones who use the AI transition as a forcing function to finally challenge beliefs they never chose.

The confidence we have in our metrics is preposterous. But the leaders who acknowledge that are the ones who get to make it essential again. This time, based on their own beliefs.