A request for a rating after the call, one figure for the management team once a quarter – that is how customer satisfaction is measured in many service organisations. The figure usually looks respectable. What remains open is whose opinion it actually reflects.
For its 2026 Consumer Trends Report, Qualtrics surveyed around 20,000 consumers in 14 countries. The finding: after a bad experience, only 29 per cent still give feedback directly to the company – markedly fewer than in 2021. 30 per cent tell nobody at all, the highest value in the series. Anyone steering their service by survey scores alone is therefore steering by a minority.
Three metrics, three different questions
CSAT, NPS and CES are often mentioned in the same breath, yet they answer different questions:
- CSAT (Customer Satisfaction Score): “How satisfied were you with this contact?” The score relates to a single contact and is suited to comparisons between channels and types of enquiry.
- NPS (Net Promoter Score): “Would you recommend us?” It measures the relationship with the company as a whole – product, price and delivery all feed into it. As a metric for a service team, NPS is therefore of limited use.
- CES (Customer Effort Score): “How easy was it to get your issue resolved?” It captures the customer’s effort and is closest to what a service team can influence itself.
For customer service, CSAT and CES are the working metrics. NPS belongs at company level and at longer intervals.
Why the number often looks better than reality
Distortions rarely arise from intent; they come from the way the survey is set up:
- Who responds? Mainly the very satisfied and the very annoyed. The broad middle stays silent.
- Who gets asked? When agents trigger the survey manually, it happens more often after good conversations than after difficult ones.
- Which contacts are missing? Anyone who hung up in the queue or never received a reply to their email does not appear in any survey.
- Which channel? Often only phone calls are surveyed. Chat, email and messaging are left out.
Then there are the statistics: with 40 responses a month, three ratings shift the score by several percentage points. A swing like that is chance, not a trend.
Five rules for robust scores
- Survey automatically: according to fixed rules, not at the discretion of the person who handled the conversation.
- Keep it short: one rating question, one optional free-text field. Every additional question costs responses.
- Ask promptly and in the same channel in which the contact took place.
- Report response rate and sample size: a score without any indication of how many customers responded cannot be interpreted.
- Analyse separately: by type of enquiry and by channel. The overall average hides where things are actually going wrong.
A survey cool-off period matters just as much. Someone who writes three times in one week should not be asked for a rating three times.
What the silent majority reveals
Most customers say nothing, yet their behaviour is visible all the same. Someone who calls again two days later, switches channel, hangs up in the queue or abandons a bot dialogue has given a rating – just without a form. In an analysis published in June 2026, Gartner advises customer service leaders to rely less on surveys for the Customer Effort Score, because low response rates limit their reliability, and to draw on indirect signals from customer contacts instead.
For small and mid-sized businesses this need not become a data science project. It is enough to place three or four behavioural metrics alongside the survey score on a regular basis: repeat contacts within a few days, abandonment rate, channel switches and transfers per case. The prerequisite is that the channels do not sit in separate systems. In myContactCenter from ilogixx, telephony, chat, email, WhatsApp and ticketing come together in one interface and one shared contact history, so the path of an enquiry can be traced across channel boundaries.
From the number to action
A metric that leads to nothing is effort without return – and it teaches customers to stop responding. A simple loop has proved its worth: a poor rating is followed by a personal response within one to two working days. Free-text answers are assigned to fixed causes such as waiting time, missing information, process or product. Once a month, the team decides which cause to tackle next.
Caution is advised with targets. If the satisfaction score becomes a personal target for individual employees, the number tends to rise rather than the satisfaction – for example because customers are asked for a good rating during the conversation. The score is better placed as a team metric and as a starting point for root-cause work.
Conclusion
Customer satisfaction can be measured, but not with a single figure from a voluntary survey. The picture becomes robust when CSAT and CES are collected according to clear rules, reported with sample sizes and complemented by behavioural signals. myContactCenter provides the foundation for this, because all channels and cases come together in one place.
Would you like to know which metrics your service can already deliver today? Book a no-obligation consultation via the online calendar at ilogixx.de – we will go through your reports together.
