Three different questions for three different things
| Metric | The question | What it measures | When to send it |
|---|---|---|---|
| CSAT (satisfaction) | “How would you rate the service you received?” from 1 to 5 | One specific interaction | Right after the case or the service is closed |
| NPS (recommendation) | “How likely are you to recommend us?” from 0 to 10 | The relationship as a whole | Periodically, not after every case |
| CES (effort) | “How easy was it to get this resolved?” from very difficult to very easy | How much effort it took the customer | After a process: a complaint, a change or a return |
You do not need all three. A company that closes many small cases usually starts with CSAT. One with few customers and long contracts gets more out of NPS.
How they are calculated
CSAT = responses of 4 or 5 ÷ total responses NPS = % promoters (9 and 10) − % detractors (0 to 6) CES = % of “easy” or “very easy” responses
Example
200 responses to the recommendation question: 110 give a 9 or 10, 50 give a 7 or 8 and 40 give 0 to 6.
Promoters = 110 ÷ 200 = 55% Detractors = 40 ÷ 200 = 20% NPS = 55 − 20 = 35
NPS is not a percentage: it runs from −100 to 100. The 7s and 8s neither add nor subtract, but they deserve attention: they are satisfied, not enthusiastic. The figures are an example.
Why surveys go unanswered
- They arrive late. A survey about a visit ten days ago asks about something the customer has already forgotten.
- They are long. Every extra question is one more chance to quit. The first question is the one that matters.
- They ask for one step too many. Open a link, wait for it to load, find the question. If the score can be tapped inside the message itself, answering is easier.
- There are too many. Someone who got a survey last week does not want another one today.
- Nobody answers the answer. If the customer gave a 2 and nothing happened, they will not respond again.
When and through which channel
| Situation | When | Channel |
|---|---|---|
| A resolved support ticket | When it is closed, the same day | The channel the customer used: email or WhatsApp |
| A technician visit | As the technician leaves | Text message or WhatsApp to whoever signed the service report |
| A purchase or a delivery | Once the customer has been able to use the product | |
| The overall relationship (NPS) | Once or twice a year, away from any incident | Email, to the decision-maker |
The rule is simple: ask on the channel where the conversation already was, while the memory is fresh. If you use WhatsApp, first review how to use it well in customer service: outside the 24-hour window you need an approved template.
What to do with a low score
This is the part that decides whether the survey is worth anything. A low score is an alert with a name attached, not a number for the monthly report.
- Set the threshold. For example, 2 or lower on CSAT, or 6 or lower on NPS.
- Open a case with an owner. Every score below the threshold becomes work for whoever handled the case or for their manager.
- Call soon, and listen. Within one or two business days, for example. It is not the time to defend yourself.
- Close the case with its cause. Record what was done and why it happened: response time, attitude, technical outcome or price.
- Review the causes every month. If the same one keeps coming back, the problem is not a person: it is the process.
A customer who scores low and hears nothing back is an early warning sign of churn. Checking scores against contracts helps you avoid losing a renewal.
Public reviews are a different matter. If you ask for Google reviews, invite everyone equally: Google does not allow showing the invitation only to customers who scored high.
Three mistakes when reading the results
- Comparing people with few responses. With ten ratings, one bad day moves anyone’s average.
- Looking only at the average. A 4.2 out of 5 can hide a group of very unhappy customers.
- Changing the question every quarter. Then you have nothing to compare against.
One question, at the right moment, on the usual channel, and someone who calls when the score is low. With that, a survey stops being a formality and starts being useful.
Last updated: