KoltrixMeasuring customer satisfaction in email support
How to measure satisfaction with email support: CSAT timing and wording, reading comments, response bias, signals beyond surveys, and keeping pay out of it.
On this page(9 sections)
Asking customers "How did we do?" after a support conversation seems like the obvious way to measure support quality. It's useful, but only if you understand what the answers can and can't tell you.
This post explains how customer satisfaction (CSAT) surveys work for email support, how to time and word them, how to read the results without fooling yourself, and which other signals belong alongside them.
What CSAT actually measures
A CSAT survey asks a customer to rate a specific interaction, usually right after it ends. The common format is a single question with a small scale, such as "How satisfied were you with this support conversation?" with options from very unsatisfied to very satisfied, or a simple good/bad choice.
It measures how the customer felt about that conversation. That's influenced by:
- Whether their problem was solved
- How quickly and clearly you replied
- How they feel about your product overall
- Whether the answer was one they wanted (a refund denial can get a low score even if handled perfectly)
So CSAT mixes support quality with product and policy. That's not a flaw, but it's important when interpreting results.
When to send the survey
Timing affects both response rate and accuracy.
- After the conversation is resolved, not after the first reply. Rating a half-finished conversation tells you little.
- Soon after resolution. A day later, the customer remembers less and cares less.
- Once per conversation. If a thread reopens and closes again, don't send a second survey.
- Not after every message for frequent contacts. Customers who email weekly will tire of surveys and stop answering, or start answering carelessly.
Some teams include a one-click rating in the final reply itself ("Did this solve your problem? Yes / No"). Others send a separate short email. Both work; the in-reply version tends to feel lighter.
How to word it
Keep it to one required question and one optional comment field.
How did we do on this conversation?
[Great] [Okay] [Not good]
Anything you'd like to add? (optional)
Wording tips:
- Ask about the conversation, not the company. "How was your experience with us?" invites ratings of the whole product.
- Use a short scale. Three to five options is plenty for email.
- Make the comment optional. Required comments reduce responses.
- Avoid leading questions like "How happy were you with our fast support?"
Response bias: who answers
Only some customers answer surveys, and they aren't a random sample. People with strong feelings, either delighted or annoyed, tend to respond more. Customers who were mildly satisfied often don't bother.
This means:
- Your score reflects responders, not all customers. Don't present it as "customer satisfaction" in general.
- Small numbers swing wildly. If a few people respond in a week, one bad experience moves the score a lot. Look at monthly or quarterly trends, not weekly ones.
- Changes in who responds can look like changes in quality. A new survey format, a different timing, or a product incident can shift results without support changing at all.
Track the response count alongside the score. A score without its sample size is easy to misread.
Read the comments, not just the score
The most useful part of CSAT is usually the free-text comments. Scores tell you something went wrong; comments tell you what.
Read every comment on low ratings, and a sample of the others. Tag them by theme:
| Theme | Example comment | Usually points to |
|---|---|---|
| Slow | "Took three days to hear back" | Coverage or triage |
| Unresolved | "Still doesn't work" | Process or escalation |
| Unclear | "Didn't understand the steps" | Writing quality |
| Policy | "Wouldn't refund me" | Policy, not support |
| Product | "This feature should exist" | Product feedback |
The policy and product themes matter: a low score caused by a refund policy is information for whoever owns the policy, not a criticism of the person who replied.
Signals beyond surveys
Surveys are one input. Other signals are often more reliable because every customer produces them, not just responders.
- Reopen rate. How often does a customer write back after you considered the thread done? High reopen rates suggest answers aren't solving the problem.
- Unprompted thank-yous. Customers replying "thanks, that fixed it" is a strong, honest signal. Some teams tag these.
- Repeat contacts on the same issue. The same customer writing in multiple times about one problem is a sign the first answers missed.
- Churn mentions. When customers cancel, check whether support experiences come up in their reasons.
- Escalations and complaints to founders or on public channels.
None of these need a survey, and together they give a fuller picture than CSAT alone.
Don't tie individual pay to scores
It's tempting to use CSAT to evaluate individual team members. Be careful:
- Scores depend heavily on the case, not just the person. Whoever handles refund denials or outage complaints will score lower.
- Small samples per person make individual scores noisy.
- Incentives distort behavior. People start avoiding difficult conversations, granting exceptions to protect scores, or nudging customers to rate.
Use individual results as a coaching input, discussed alongside the actual conversations, not as a target or a pay factor. Team-level trends are a better management measure.
A simple measurement setup
For a small team, this is usually enough:
- A one-question survey sent once per resolved conversation.
- A monthly look at score, response count, and comment themes.
- A monthly look at reopen rate and repeat contacts.
- Low-rated conversations reviewed together in a team meeting, focusing on what could change.
- Product and policy themes passed to their owners.
Bottom line
CSAT is a helpful signal about how customers feel after a support conversation, but it reflects only the people who respond and mixes support with product and policy. Ask one clear question after resolution, read the comments, watch trends with sample sizes, pair surveys with signals like reopen rates and unprompted thanks, and keep scores out of individual pay.
Start with Koltrix
Your domain, one inbox, and an API that sends.
A team inbox where AI sorts and drafts (nothing is sent without your click), plus the transactional API and SMTP relay your product sends with. 7 days free, no card.
