Customer surveys are still dead!
Three years ago I wrote a post arguing customer surveys had had their day. The argument then was that response rates were falling, customers were busy, and the data systems coming online would soon do a better job. I’d say the point is even stronger now. Surveys aren’t just past their useful life. There’s a risk they’re distorting how companies serve customers.
Three things have changed since 2023.
Firstly, response rates have got worse, not better. Customer survey response rates in 2025 sat at 20-30% on a good day, and email channels are now routinely in the teens. The UK’s Labour Force Survey (the data the ONS uses to produce official unemployment figures) has fallen sharply to around 13%. If government statistics with real consequences can’t get people to respond, your post-purchase NPS prompt is not winning.
Secondly, AI has flipped the economics. Real-time analysis of unstructured customer interactions, the thing your call and email transcripts and chat logs already contain, was aspirational in 2023. It’s a basic and standard capability in 2026. Companies that haven’t picked it up are choosing the harder route.
Finally, and this is the one that made me stop in my tracks and move my thinking from “surveys are tired” to “surveys are harmful”. The survey economy has started corrupting the operation itself. Two experiences over the past year highlight what I mean.
A courier company couldn’t deliver a parcel to me for three months. The advisor I spoke to couldn’t help, the business processes blocked them, and they couldn’t fix the underlying problem and the parcel was never delivered. The agent, while apologetic, still had to ask me for a top rating (it was part of the process).
An Amazon seller delivered the product on time. Inside the box was a card offering me an 80€ gift card if I posted a 5* review and sent them a photo of it.
No one wins here. The customer gets a worse experience, because being asked to rate a service that failed you is insult added to injury. And the score is meaningless, because the operation generating it has every incentive to game it.
It is worth calling out one defence of surveys: they give you the customer’s words. They do, but only in response to your questions. Reviews and transcripts give you the customer’s words on their terms, when they actually had something to say, in the context of trying to get something done. That’s a stronger form of customer voice, not a weaker one.
So what should you do instead? Same answer as 2023, but easier now. Spend the energy on the data you already hold. Effort scores from interaction logs, sentiment from transcripts, abandonment patterns, repeat contact rates, complaint themes. The signals are there. Stop asking customers things you could find out yourself. And stop asking your own people to force superficial reviews.
If your operation doesn’t have this data structured yet, that’s where to start. You don’t need a big programme. You need three to five signals you can read every week. If you have the data but not the capability to make sense of it, you don’t need a data science team either. You need someone who can ask the right operational questions of the data you already have, and translate the answers into changes the business can act on.
That’s the kind of work I do. If you’re in either of those positions, get in touch.
Surveys made sense when getting customer data was hard. It isn’t hard anymore. Anyone still running a survey-led VoC programme in 2026 is choosing the more expensive, less timely and less accurate route.