Customer Feedback Surveys and the Three Support Numbers to Track
10 min read
A customer feedback survey for a small business needs three questions, a free form tool, and an AI pass over the answers. This lesson sets that up and adds the three support numbers worth tracking by hand.
You've changed a lot in four chapters. Templates, a shared inbox, an FAQ, auto-replies, a calmer way with complaints. How would you know if any of it worked? Most owners go on feel, and feel is dominated by the last angry email. Ten minutes a week of light measurement gives you a truer picture.
The three-question survey
Long surveys get abandoned. Ask three things:
- How easy was it to get what you needed from us today? (1 to 5)
- What's the main reason for your score?
- What's one thing we could do better?
The first question gives you a trend. The second and third give you the words, and the words are where the value is.
Build it free in Google Forms or Tally. Both take ten minutes and both collect answers in a spreadsheet. Then pick one moment to send it: after a support issue is closed, after a job is finished, or a few days after delivery. Put the link in your closing reply template. Don't survey the same customer more than once a quarter.
If you'd like question wording tuned to your business, the free Customer Feedback Survey Builder writes it, along with the message that asks people to fill it in.
A realistic expectation: most people won't answer. One in ten is normal for a small business, more if you ask in person. That's fine. Thirty honest answers over a quarter will tell you things you didn't know.
Let AI read the answers
Reading forty free-text answers one at a time, you remember the rudest and the kindest. An assistant reads them all with the same attention.
Below are customer survey answers for my [TYPE OF BUSINESS]
from the last [PERIOD]. No names included.
1. Group the comments into themes. For each theme: how many
mentions, one or two short quotes in the customer's words,
and whether it is praise, a complaint, or a request.
2. List the three changes that would have prevented the most
complaints.
3. Pull out any phrases customers use to describe what they
like about us. I may use them in marketing.
4. Flag anything that sounds like a customer about to leave.
Count only what is in the answers. Do not pad thin themes.
[PASTE ANSWERS]Item 3 is a bonus most owners miss. Customers describe your value in plainer, better words than you do, and those phrases belong on your website. The Voice-of-Customer Miner prompt does the same across support emails, chats and reviews, and the Customer Insights Copilot skill keeps a running record month to month.
The three numbers
Big support teams track dozens of metrics. You need three, counted by hand from one ordinary week each month.
First-response time. How long until a customer hears something useful from a human. Pick ten messages from the week, note when each arrived and when the first real reply went out, and take the middle value. This is the number customers feel. It should match what you promised in Chapter 3.
One-reply rate. Of those ten, how many were finished in a single reply? A low rate means your replies leave out the next thing the customer needs, or your templates answer a slightly different question than the one asked. Each back-and-forth doubles the work.
Repeat questions. How many of the week's messages asked something your FAQ or auto-replies should have covered? If it's high, the FAQ is missing answers or customers can't find it. If it falls month after month, the work from Chapter 2 is paying off.
| Number | How to count it | What moves it |
|---|---|---|
| First-response time | Middle value of 10 sampled messages | Triage, templates, shared ownership |
| One-reply rate | Finished in one reply, out of 10 | Better templates, asking for details in auto-replies |
| Repeat questions | Messages the FAQ should have caught | FAQ coverage and where you link it |
Write the three numbers in a spreadsheet with the date. That's the whole dashboard. If you use a help desk, it reports the first one for you. The other two still take a human eye.
Take a baseline this week, before the changes from this course settle in. "First reply went from nine hours to two" is the sentence that keeps you doing this when you're busy.
Notice what's not on the list: message volume. Fewer messages could mean your FAQ works or that customers gave up. The number can't tell you which. For a version of these metrics built for a team running AI in tiers, see AI for Customer Support: The Small-Team Playbook.
Check the quality of your own replies
Speed is easy to measure. Quality drifts without anyone noticing. Templates get stale, a new hire improvises a policy, AI drafts get approved too quickly on a Friday afternoon. Once a month, take ten sent replies, remove the personal details, and run them through the free Support Reply Auditor. It grades each reply against your policy and your voice and names where they've drifted. Ten minutes, and you catch the wrong refund window before forty customers have seen it.
Where this breaks
Survey scores from small samples bounce around. Five answers in a month can swing from 4.8 to 3.2 on one unhappy person. Read the comments, watch the trend across a quarter, and don't rewrite your process over a single bad week. And be careful what you reward. If staff are judged on response time alone, you'll get fast, useless replies.
Try it now
Build the three-question form today and add its link to your closing template. Then sample ten messages from last week and write down your three baseline numbers. You'll compare against them in the final lesson's Friday review.