Something has changed about being asked for feedback.

It used to feel occasional. Now it’s constant. After a delivery, a support call, a click around a website — sometimes mid-visit, before you’ve even finished what you came to do. The timing is so predictable it’s almost funny: you finish the thing, and two minutes later your phone lights up with “How did we do?”

Most of the time, I ignore it. Not because the experience was bad. Usually, it was fine. I just don’t have the time to reply to every feedback request that lands, and I suspect you don’t either.

That’s the tension I’ve been sitting with for this episode. UK retailers are asking for feedback more than ever. At the same time, response rates have dropped significantly. So we’re asking more. And hearing less.

In episode 20 of Where’s Your Customer, I looked closely at what’s happening inside customer satisfaction measurement in retail right now. Why the survey-led model is under more pressure than the dashboards suggest, and what’s worth questioning before we add another feedback touchpoint to the journey.

This isn’t an argument against surveys. It’s worth understanding what customer satisfaction measurement can and can’t tell you.

Most voice-of-the-customer programmes are listening to only a fraction of what’s happening.

Research from Forrester puts the capture rate for transactional surveys between 2% and 7%, depending on channel and sector. That has always been a thin slice. What’s changed is that response rates are falling further still, meaning the slice is getting thinner every year.

If you’re running a structured feedback programme across store visits, deliveries, returns, service calls, and website journeys, somewhere between 92 and 98% of those moments leave no structured signal at all. The organisation doesn’t know what happened in the vast majority of its customer interactions, because nothing in the process was designed to see them.

And the direction of travel is consistent. According to ESOMAR, overall survey completion rates have dropped by close to 30% over the last decade. Digital fatigue, mobile friction, and what researchers are calling ‘privacy cynicism’ (a growing scepticism about how feedback is actually used) are all pushing participation down.

Running a customer satisfaction measurement and assuming the data is representative is, at this point, an increasingly difficult position to maintain.

The slice you do capture may not be as clean as the dashboard suggests.

Even within that five per cent, the question of data quality matters.

Retail Technology Review reported in January 2026 that up to 30% of online survey responses may be fraudulent or misleading, completed by bots, by people chasing incentives, or by respondents rushing through with no meaningful detail.

There’s also the non-response bias problem. The customers who do respond tend to sit at the extremes, such as those who were genuinely delighted or those who had a bad enough experience to say something. The bulk of customers in the middle, the customers for whom the experience was functional and unremarkable, opt out. So the data doesn’t just undersample: it systematically overstates the emotional peaks and understates the everyday.

When the slice is small and the quality within it is variable, the confidence with which organisations act on VoC data appears disproportionate.

UK customers score differently. Most benchmarks weren’t built for that.

There’s a cultural dimension to this that doesn’t get discussed often enough in UK retail, and it’s probably been distorting how CX teams interpret their dashboards.

Most of the major satisfaction metrics (NPS, especially) were designed for and calibrated against the US market. Under the standard NPS methodology, a score of 9 or 10 makes a customer a Promoter. A seven or eight makes them Passive. A score of 6 or below makes them a Detractor.

But British customers score differently. Benchmarking data from Retently and others consistently shows that UK and European respondents are more conservative with top-of-scale scores. An eight, in a UK context, often means: that was good, I’m happy, I’d come back. It’s essentially a compliment.

Under NPS, the same eight registers are Passive.

So UK retailers can end up systematically misreading positive customer relationships because the framework was calibrated for a different scoring culture. And when those scores get compared against US benchmarks, the business looks like it’s underperforming.

Sometimes it is underperforming, of course. But sometimes the experience is fine, or better than fine, and what the metric is really reflecting is simply that your customers are British.

When the measurement pressure is high enough, people manage the score

There’s another side to this that you only really see in retail environments where scores affect pay or performance reviews.

When a satisfaction score is linked to compensation or rankings, something predictable happens. Team members coach customers on the way out: ‘If you get a message asking about your experience, a ten really helps us.’ Survey invites are sent selectively to customers who seem happy, rather than at random. The mechanism gets tuned toward the results that keep the number where it needs to be.

This is Goodhart’s Law in practice. Coined by economist Charles Goodhart in 1975, the principle states that when a measure becomes a target, it ceases to be a good measure. The moment customer satisfaction measurement is used to incentivise or control, the rational response shifts from improving the experience to improving the score. And in a retail environment where satisfaction scores are linked to pay or performance reviews, that is entirely predictable. Not to blame individuals, it’s the system doing exactly what systems do when you apply pressure to a number.

The result is a dashboard that looks healthy, while the people closest to the customer, if you sat down with them, would tell you something quite different.

This matters in the context of wider UK data. The UKCSI hit its lowest point in nine years in January 2024, at 76 out of 100, despite almost every major retailer having some form of feedback programme running. Mintel’s UK Customer Service Report found that 73% of consumers who contacted retail customer service in 2024 encountered problems with the service (again), in a world where listening systems are supposedly everywhere.

The collection happened. Something else didn’t.

Collecting feedback and acting on it are not the same thing.

This is the part that’s uncomfortable to say out loud, particularly for anyone who has fought to build a VoC programme from scratch.

According to Forrester, only about 30% of businesses that collect customer feedback actually act on it. And from the customer side, Mintel found that 73% of retail consumers who contacted customer service in 2024 encountered a problem, despite the retailer having a feedback system in place. The collection happened. Something else didn’t.

So we’ve got a listening system that’s mainly optimised to collect but not to change. If customers sense that completing a survey changes nothing, they stop completing surveys. And data from Mintel and others suggest that this is exactly what’s happening, leading to an erosion of trust. The implicit promise behind every feedback request (that someone is listening and will do something with what you say) has worn thin.

And this is where the M&S example is interesting. M&S runs two parallel listening programmes: ‘Pulse’, a rapid feedback mechanism sent regularly to large volumes of customers, and ‘The Collective’, a community of around 40,000 food customers used for deeper concept testing. What makes The Collective different is the relationship it creates. Customers who contribute can see their input reflected in what appears on shelves. That’s a different relationship with feedback than the standard post-transaction survey because the loop closes visibly.

The question for any organisation watching its response rates fall isn’t ‘how do we get more people to complete this?’ It’s ‘do our customers have any reason to think completing this survey is worth their time?’

Those are different questions. And they lead to different answers.

AI-driven listening reaches further, but brings its own questions.

A lot of UK retailers are starting to move away from relying purely on asking customers and more towards observing what customers do and say organically.

M&S is an example again. Through a partnership with Ocula Technologies, they’re analysing what customers type into the M&S search bar, because what people type into a search bar is unfiltered, with no question to answer, and no format to follow. It’s what customers actually call things when they’re not trying to answer a question ‘correctly’. It’s listening without asking.

The broader logic behind this shift runs as follows. If surveys capture two to seven per cent of customer interactions, then call centre transcripts, social reviews, search behaviour, complaints, and digital intent signals can tell the rest of the story. AI tools such as natural language processing, sentiment analysis, and behavioural pattern detection can operate across that landscape at a scale no survey programme can match.

That’s a compelling case for the back-office use of AI listening. But there’s a watchout worth noting too.

AI isn’t only being used to listen to customers. It’s also increasingly being used to serve them: chatbots, automated responses, and AI-generated recommendations. And when that customer-facing AI goes wrong, it creates its own feedback problem.

Research from Trustpilot and the Centre for Economics and Business Research, published in October 2025, found that reviews mentioning AI averaged 1.7 stars, which is significantly below average. Negative AI experiences are putting an estimated £8.6 billion of UK e-commerce sales at risk.

The EY CEO Outlook Survey at the end of 2025 found that 76% of retail CEOs were confident their AI solutions would deliver a tangible return on investment. Research from CI&T, also published in December 2025, found that 68% of UK consumers could not name a single standout AI retail experience. That’s a large gap between executive confidence and actual customer perception.

And trust is a thread that runs through all of this. A SAP Emarsys study from early 2025 found that 57% of UK consumers have little to no trust in brands to use AI responsibly, and that British shoppers now rank as the least confident globally about how their data is being used.

Because passive listening doesn’t ask customers questions, it’s harder for customers to see and trust. Which means the question isn’t simply ‘is AI listening better than surveys?’ It’s what kind of relationship with customers are you trying to build, and are your listening tools transparent enough to support that?

Customer satisfaction measurement still works when it’s connected to outcomes.

Before drawing any conclusions, it’s worth being fair. Surveys aren’t useless. NPS and CSAT aren’t wrong.

In some contexts, they give leadership a commercially legible number, something trackable over time and connectable to outcomes. Ocado is consistently ranked among the UK’s highest performing brands for customer experience by KPMG, and they’re one of the clearest examples of a retailer that has tied their customer experience investment directly to commercial outcomes, using NPS as a leading indicator rather than a vanity metric.

John Lewis is worth looking at here. In 2024/25, they brought back Never Knowingly Undersold, a significant strategic commitment after a period of brand uncertainty. In their full-year results, they link that decision to a six-point NPS gain in the second half of the year.

Waitrose achieved a record-high NPS in the same period, alongside four consecutive years of recognition for best customer service at The Grocer awards and year-on-year improvement in operating margin.

These aren’t just good scores. They’re scores attached to something concrete, like decisions customers could feel. When the experience changes, and the metric moves with it, the survey is doing its job. It’s not ‘please rate us out of ten’. It’s a signal that what changed has registered with the people experiencing it.

The survey mechanism doesn’t fail. What fails is the expectations we load onto it: we take a simple measuring tool and ask it to behave like a full listening system, a diagnostic tool, and an accountability mechanism all at once.

If you’re thinking about what to measure alongside or instead of NPS, episode 4 covers the CX metrics that connect more directly to commercial outcomes.

The survey problem is largely a problem of distance.

Almost everything discussed here, declining participation, gaming, metrics that drift from reality, these are more likely to emerge when organisations grow past a certain size. When the distance between the people making decisions and the customers having experiences becomes too great to cross without a system in between.

A small independent retailer doesn’t usually have a survey problem. Not because they’ve solved measurement, but because they haven’t needed to formalise it. The owner of a good independent shop knows what their customers think because they’re having those conversations directly. Someone mentions at the counter that the range has changed, or that parking has got worse, or that a particular member of staff was especially helpful. That’s feedback. It’s immediate, unfiltered, and acted on by the person who heard it.

The survey was designed to replace that direct relationship at scale. When you have hundreds of stores and thousands of transactions a day, you can’t stand at every counter. So you build a programme. You ask questions. You aggregate the answers. You try to reconstruct, at a distance, the signal that can no longer arrive in person.

That substitution has always had limits. What’s changed is that those limits are now more visible.

NatWest research found that a third of consumers actively prefer local businesses over larger counterparts. Richer Sounds, with its 52 stores, has collected over 120,000 Trustpilot reviews at 4.9 out of 5 and has been named by Which? Retailer of the Year for six consecutive years. The attribution for that isn’t survey design. It’s an employee-ownership model where everyone has a stake in doing the right thing every day.

If your organisation sits somewhere in the middle, like past the point where the owner knows every customer, but not yet big enough to require a full measurement infrastructure, the main question now is: which of those two worlds are you actually closer to? And are the tools you’re using built for that?

What we’re all probably not asking ourselves is: when a customer doesn’t reply, what do we do with that silence? Does it even register at all?

In Summary

Most customer satisfaction measurement captures somewhere between two and seven per cent of customer interactions, and the slice is getting thinner. Of that percentage, a portion may be unreliable. In the UK, cultural scoring patterns can distort how we interpret ‘good’. And only thirty per cent of organisations act on what they collect.

Run those together, and a lot of ‘listening systems’ produce a view of the customer experience that covers only a small fraction of what’s actually happening and then act on only a fraction of that fraction.

AI-driven listening can address some of this. It covers more of the landscape, surfaces organic signals, and can reveal emerging friction earlier. But it brings its own questions about trust, disclosure, and what it means to listen without asking.

Underneath it all, there’s a simpler question: what does an organisation believe it’s doing when it says it listens to customers? If the honest answer is that it’s tracking metrics that let it claim it’s listening, then the current tools are probably well-suited to that. But if the answer is that it wants to understand what customers actually experience, whether things are working, and what needs to change, then the gap between the tools and that intention is worth looking at more carefully.

Frequently Asked Questions

What is the voice of the customer (VoC) in retail?

Voice of the customer (VoC) in retail refers to the structured process of capturing, analysing, and acting on customer feedback across all touchpoints in the shopping journey. It typically includes transactional surveys (post-purchase, post-service), relational surveys (periodic NPS or CSAT measures), and increasingly, passive methods such as review analysis, social listening, and AI-powered sentiment detection. The goal is to understand what customers actually experience, not just what the business assumes they experience.

Why are survey response rates declining in retail?

Several factors are converging. Customers receive far more feedback requests than they did five years ago, eroding the novelty of being asked for it. Mobile friction (long surveys on small screens) spikes abandonment rates. And growing ‘privacy cynicism’ means customers increasingly question whether their responses will make any difference. Research suggests this last factor is particularly significant: when customers have no evidence that feedback leads to action, they stop giving it.

Is NPS still a useful metric for UK retailers?

NPS remains one of the most widely used customer satisfaction measurement tools. It’s useful as a directional indicator, particularly when tracked over time and linked to specific business decisions or changes in experience.

The limitations are worth understanding clearly, however. NPS was calibrated against US scoring behaviour, and UK and European customers consistently score more conservatively. An eight in a UK context often reflects genuine satisfaction but is classified as ‘Passive’ under the standard NPS framework. Comparing UK scores against global benchmarks without accounting for this cultural difference can lead to misleading conclusions about how the experience is landing.

Episode 4 explores some of the complementary metrics worth considering alongside NPS

What is Goodhart’s Law, and how does it apply to customer satisfaction scores?

Goodhart’s Law, articulated by economist Charles Goodhart in 1975, states that when a measure becomes a target, it ceases to be a good measure. In retail CX terms, when satisfaction scores are linked to pay, rankings, or performance reviews, teams are incentivised to manage the score rather than the experience. This can include coaching customers on how to respond, selectively sending survey invites, or gaming the timing of requests. The score looks healthy. The experience may not have changed at all.

How is AI being used to improve customer listening in retail?

AI tools are increasingly used to analyse the signals customers generate organically; search queries, call centre transcripts, reviews, social posts, returns data, rather than relying solely on structured survey responses. Techniques include natural language processing, sentiment analysis, and behavioural pattern detection. This ‘passive listening’ approach can cover a much larger proportion of customer interactions than transactional surveys. However, it raises questions about transparency and customer trust, particularly in the UK, where consumer confidence in how brands use AI and data is currently low.

What percentage of businesses act on the customer feedback they collect?

According to Forrester, only around 30% of businesses that collect customer feedback actually act on it. This gap between collection and action is one of the primary drivers of declining survey participation. When customers have no evidence that their feedback makes any difference, they stop giving it.

I value your opinion

I started Where’s Your Customer? because I kept having conversations with retail leaders who were asking the same questions I was. About customers. About who they really are and what drives them to purchase things. About how we’re measuring that. About the gap between the data we have and the picture it paints.

I didn’t want to make a show that told people what to do. I wanted to make one that made us think differently.

I’m not entirely sure I’ve always got that balance right.

So I’m asking: if you listen (or read these show notes pages), what’s actually landing? What do you find yourself coming back to? And, what do you want more of?

I’ve put together a short survey to gather your thoughts so far. Would you mind completing please?

It’ll only take about two minutes. Thanks so much!