NPS Is Not a Product Strategy: Use It Without Being Misled

NPS can be useful, but only as one signal. Learn how to combine NPS with behaviour and qualitative evidence so product decisions are grounded in reality, not vanity scores.

I've been on a lot of calls where a CEO says, with genuine pride, "our NPS is 42." I've never quite had the heart to say what I'm thinking, which is: that number on its own tells me almost nothing useful about your business. I've seen companies with an NPS of 65 churn customers at terrifying rates. I've seen companies with an NPS of 12 quietly compound into market leaders. The number is a thermometer. It tells you the room is warm. It doesn't tell you where the heat is coming from, or whether you should open a window or call the fire brigade.

When we lifted NPS at True Digital by 35% over a single year, the lift wasn't the goal. It was a side effect of fixing the things underneath. If we'd targeted NPS directly, we would have spent twelve months optimising survey responses and zero months building anything users actually wanted.

What NPS actually measures (and what it doesn't)

NPS measures the gap between people who like you enough to recommend you and people who don't. It's a stable, comparable, simple-to-collect signal. That's its strength. Its weakness is that it conflates ten different reasons someone might be a promoter and twelve different reasons someone might be a detractor, and it gives you none of them. A high NPS can mean you have a great product. It can also mean you have a niche product loved by a small loud group while everyone else churns silently.

If you treat NPS as a strategy, as in "we will lift NPS by ten points this year," you're committing to optimise an output you don't control by manipulating inputs you haven't identified. It's like deciding to lower your blood pressure by buying a new blood pressure cuff.

The four questions that actually move the number

When I run UX strategy engagements, I throw out the NPS dashboard for the first month. Instead, I get the team to answer four diagnostic questions that consistently predict whether the experience is improving or declining.

1. Time to first value: how long from sign-up to the user's first "oh, I get it" moment? If it's longer than 90 seconds in a SaaS product, you have a story problem, not a feature problem.
2. Activation rate by segment: what percentage of new users do the one thing your product is actually for, within their first session? Slice this by source, by plan, by industry.
3. Repeat-week retention: of the people who used the product in week one, how many return in week two? Week four? This is the only retention signal that survives reality.
4. Support volume per active user, segmented by surface: where in the product are people getting stuck enough to ask for help? That's your design backlog, ranked by pain.

Fix any one of those four and NPS will move on its own, with no campaign, no survey design, no incentivised replies. Try to move NPS without fixing those four and you'll spend a quarter chasing your tail.

The True Digital playbook in three moves

When we set out to improve the customer experience at True Digital, we did three things, in order. None of them mentioned NPS. All of them moved it.

First, we mapped every place a customer touched the company across thirty days: the app, the support line, the in-store visit, the bill in the mailbox. We didn't go looking for the highest-rated touchpoint. We went looking for the silently bad ones, the ones nobody complained about loud enough to make it onto a roadmap. We found seven. Six of them were fixable in a quarter. We fixed them.

Second, we redesigned onboarding with a single hard constraint: a new user had to feel value within the first two minutes, or we considered the design a failure. That single constraint forced us to cut three steps that had been there for years because someone, once, had a meeting and decided they were important.

Third, we built a closed loop with the support team. Every week, the top three pain points coming through support became the top three items on the design backlog. Not eventually. That week. The relationship with support went from adversarial to symbiotic, and the data we got from them was the highest-leverage research input we had.

Why "voice of customer" programmes fail

Most companies have a voice of customer programme, and most of them are theatre. Surveys go out, results get collated, a quarterly slide deck goes around, and nothing changes in the actual product. The reason is simple: the people who collect the feedback aren't the people who can act on it, and the action loop takes six months instead of six days. By the time anything happens, the customer who complained has already churned and the team has moved on.

Where to start if you're inheriting a stalled NPS

If you've just inherited a product whose NPS has flatlined for a year, here's the playbook. Stop running new surveys. You already have enough data. Spend a week in the support inbox personally. Pull out the ten most frequent themes by ticket volume. Map those themes against the four diagnostics above. You'll find that one or two areas account for the bulk of your detractor weight. Fix those. Don't tell anyone you're doing it for NPS. Tell the team you're doing it because users are getting stuck. Then watch the number move on its own three months later.

One more thing worth saying out loud, because I see it derail teams every quarter: never let the NPS number become the bonus metric for product or design. The day it appears in someone's compensation is the day the data quietly becomes useless. Surveys get re-timed. Detractors get suppressed. The wording shifts. Six months in, you have a beautiful upward trend and a product nobody actually likes any more. Tie compensation to the underlying behaviours: retention, expansion, support deflection. Then treat NPS as the diagnostic conversation you have with the team afterwards.

NPS is a useful conversation starter and a terrible North Star. The teams that grow consistently treat it as a barometer they glance at occasionally, not a target they chase. Build the product so well that the number takes care of itself. That's the only NPS strategy that's ever worked.

In short

Practical checklist

• Treat NPS as a signal, not a decision by itself.
• Pair score changes with behavioural and qualitative evidence.
• Identify where detractor pain maps to measurable product friction.
• Set one outcome metric that reflects real user value.
• Review whether interventions improved behaviour, not just survey sentiment.

When to use this approach (and when not to)

• Use this approach when decisions carry customer, revenue, or delivery risk.
• Use it when multiple teams need consistent quality standards.
• Use it when you need repeatable outcomes, not one-off output.

• Do not steer roadmaps from score movement without anchoring to behaviour and outcomes.
• Do not ignore qualitative evidence when metrics look fine—segmentation and expectations hide there.
• Do not present NPS as proof without context on reach, timing, and known biases.

If your team is optimising the score

If your team is trapped in score-chasing, begin by reframing one decision with behavioural and qualitative evidence. I can help you set up that operating model with your product and design leads.

Frequently asked questions

Is NPS useless for product teams?

No. It is useful as a directional signal, but weak as a standalone decision tool. Pair it with behaviour data and qualitative evidence.

What is the risk of overusing NPS?

Teams optimise for score movement instead of user outcomes. That can drive shallow improvements while deeper product issues remain unresolved.

What should we do when NPS and behaviour data conflict?

Trust behaviour signals first, then investigate why survey sentiment diverges. This often reveals segmentation or expectation issues.

How do we avoid gaming NPS improvements?

Anchor decisions to outcome metrics and user behaviour, not score movement alone. Treat score gains without behaviour change as weak evidence.

When is NPS most useful?

As a trend signal for where to investigate deeper, especially when paired with qualitative feedback and journey-level metrics.