A CSAT score is the share of customers who rate a support interaction as satisfactory, usually a 4 or 5 on a 5-point scale, divided by everyone who answered. It measures one conversation. It does not measure loyalty to your brand. On a small team, sample size matters even more than that distinction. A score built on 25 replies can move 10 points or more from luck alone.
TL;DR:
- CSAT = satisfied responses (4s and 5s) ÷ total responses × 100. Pick that version or the average-rating version and never switch between them.
- Ask one question, on one scale, right after a ticket is solved. Every extra question lowers your response count.
- Ignore outside benchmarks. Your own month-over-month trend and per-store splits are the numbers that tell you what to fix.
- Under about 100 responses, treat a swing of 5 to 10 points as noise until it holds for two or three months.
- Track CSAT first, add customer effort second, and skip NPS until you have real volume.
Table of Contents
- How to Calculate a CSAT Score, and the Two Versions People Mix Up
- Writing the Survey: One Question, One Scale, Sent at the Right Moment
- What Counts as a Good CSAT Score, and Why Outside Benchmarks Mislead You
- Small Numbers, Big Swings: Reading CSAT on 30 Responses a Month
- Splitting the Score by Store and by Person
- CSAT vs. NPS vs. Customer Effort Score: Which to Track First
- Turning Low Scores Into Fixes: A 30-Minute Monthly Review
- Get Per-Store CSAT Reports Without a Per-Agent Bill
- Frequently Asked Questions
- Sources
How to Calculate a CSAT Score, and the Two Versions People Mix Up
The standard CSAT question is some form of "How satisfied were you with the help you received?" The customer answers on a 1–5 scale, where 1 is very dissatisfied and 5 is very satisfied.
From there, teams calculate the score in one of two ways. Both are legitimate. Mixing them is not.
| Method | Formula | Example (50 responses: 30 fives, 12 fours, 5 threes, 3 ones) | Reads as |
|---|---|---|---|
| Percent satisfied (top-2-box) | (4s + 5s) ÷ total × 100 | (30 + 12) ÷ 50 × 100 = 84% | "84% of customers were satisfied" |
| Average rating | Sum of scores ÷ total | (150 + 48 + 15 + 3) ÷ 50 = 4.32 | "Average rating of 4.32 out of 5" |
Use the percent-satisfied version. It is easier to explain to a part-time agent, and it is the version most help desk tools and published discussions mean by "CSAT." It also stops a few furious 1s from dragging down a score that is mostly 5s, which would hide what most customers felt.
The average still has one use. Look at it next to the percentage. If the percentage holds steady but the average drops, more of your satisfied customers are giving 4s instead of 5s. That shift often shows up before the headline number moves.
Whichever method you choose, write it down and keep it. A team that reports 84% in March and 4.3 in April has no trend at all.
Writing the Survey: One Question, One Scale, Sent at the Right Moment
The survey design does more to your score than anything your agents do in a given week. Three decisions matter most.
- Ask one question. Each extra question cuts your completion rate, and you cannot afford to lose responses. Add an optional free-text box ("Anything we could have done better?") and stop there.
- Send it when the ticket is solved, not days later. The customer is rating the conversation, so ask while they still remember it. A survey sent a week later picks up feelings about shipping, the product and the weather.
- Label every point on the scale. A bare row of five numbers lets people read "3" as either "fine" or "bad." Labels make the answers comparable across customers.
The ISO 10004 guidelines on monitoring and measuring customer satisfaction make the same broad point. Decide what you are measuring and how before you collect anything, then keep the method consistent so the results can be compared over time. For a two-person support team, that comes down to this: same question, same scale, same trigger, every ticket.
Decide one more thing up front: whether surveys go out on every solved ticket or only some. Survey them all. At small volumes, you need every response you can get.
What Counts as a Good CSAT Score, and Why Outside Benchmarks Mislead You
Owners always ask what a good score is. The honest answer is that a published number tells you very little about your own.
Benchmarks gathered from other companies use different questions, scales, triggers and customer bases. A store that surveys only tickets marked "resolved happily" will post a better score than one that surveys refunds and lost parcels. Neither number says anything about you.
The American Customer Satisfaction Index shows why. ACSI publishes satisfaction scores on a 0–100 scale for whole industries and large companies, built from its own survey model. It measures customers' overall experience with a company, not a single support ticket. If you set your 88% ticket CSAT next to an ACSI industry score, you are comparing two different measurements that happen to look alike.
Here is a working position you can adopt today:
- Below 70%: something specific is broken, such as a policy, a product defect or one queue that never gets answered. Find it.
- 70–85%: normal for most small e-commerce support. Work on your trend and your weakest store.
- Above 90%: good, or a sign that you only hear from happy people. Check your response rate before celebrating.
Your real benchmark is last quarter. Beating your own trailing three-month average is the target worth setting.
Small Numbers, Big Swings: Reading CSAT on 30 Responses a Month
Most CSAT advice is written for teams that get thousands of survey responses. A small store might get 30. At that volume, ordinary chance moves your score around more than most owners realize.
At 20 responses a month, a single unhappy customer moves your CSAT score by 5 points.
The table below shows the approximate 95% margin of error for a score of about 85%, using the standard formula for a proportion. Read it as "the true score is probably somewhere within this many points of what you see."
| Responses in the period | Approx. margin of error (at ~85%) | What you can conclude |
|---|---|---|
| 30 | ± 13 points | Almost nothing month to month |
| 50 | ± 10 points | Only very large shifts |
| 100 | ± 7 points | Sustained moves over 2–3 periods |
| 200 | ± 5 points | Reasonable monthly comparisons |
| 400 | ± 3.5 points | Fine-grained comparisons |
So if your score drops from 86% to 78% on 40 responses, that is not a crisis. It is within the noise. Three practical rules follow:
- Widen the window. Report a rolling 90-day CSAT if a month gives you fewer than 100 responses.
- Wait for a pattern. Act on a drop only when it shows up in two or three periods in a row.
- Read the comments every time. A single complaint can name a real problem, like a broken discount code, long before the score moves.
Response rate matters too. If only 8% of customers answer, you are hearing from the most motivated customers, who tend to be either delighted or angry. Track response rate next to CSAT, and be suspicious of a score that rises while responses fall.
Splitting the Score by Store and by Person
A single blended CSAT number hides the most useful information you have. If you run several stores, one bad brand can sink an average that would otherwise look fine.
By store
Each store has its own products, its own customers and its own recurring problems. A store selling made-to-order furniture will always field more delay complaints than one selling phone cases. Compare each store against its own history, not against the other stores. A drop in one store usually points to something outside support, such as a supplier delay, a confusing size chart or a new returns rule.
By person
Per-agent CSAT is useful for coaching and dangerous for ranking. Agents get different ticket mixes. Whoever handles refunds will score lower than whoever answers "where's my order?" Use per-person scores to find the conversations worth reviewing together, not to build a league table. On small teams, the sample-size problem gets worse here, because one person might see 15 responses a month.
By ticket type
If you tag tickets, also look at CSAT by tag. "Damaged item" scoring 55% while "order status" scores 92% tells you where to spend next month's effort far better than a blended 81% does.
Pro Tip: Split CSAT by whether the ticket met its first-reply target. If tickets answered within your target score 90% and late ones score 65%, your problem is staffing and coverage, not tone. Hire seasonal help or tighten your queue. If both groups score about the same, speed is not your issue, so look at policy and product instead. This one comparison settles more "should we hire?" arguments than any other report.
CSAT vs. NPS vs. Customer Effort Score: Which to Track First
There are three common satisfaction metrics, and they answer different questions.
| Metric | Question asked | Measures | Best for |
|---|---|---|---|
| CSAT | "How satisfied were you with this interaction?" | One support conversation | Day-to-day support quality |
| Customer Effort Score (CES) | "How easy was it to get your issue resolved?" | Friction in the process | Finding broken workflows |
| Net Promoter Score (NPS) | "How likely are you to recommend us?" | Overall brand loyalty | Company-level tracking |
NPS came from Fred Reichheld's 2003 Harvard Business Review article, "The One Number You Need to Grow." It was designed to measure the whole relationship, not one ticket. Ask it after a support exchange and you mostly measure how the customer feels about your product and prices.
Customer effort has strong support of its own. In "Stop Trying to Delight Your Customers" in Harvard Business Review, Matthew Dixon, Karen Freeman and Nicholas Toman argued that reducing customer effort does more for loyalty than going above and beyond. That is a useful reminder that a polite, slow, three-email answer is worse than a blunt, fast, one-email answer.
Our position for a small team is to track CSAT first, because it is the most direct read on support quality and customers understand the question. Add a CES-style question later if your comments keep mentioning repeated contacts or being bounced around. Skip NPS until you have the volume to make it meaningful and someone outside support who will act on it.
Turning Low Scores Into Fixes: A 30-Minute Monthly Review
A CSAT score that nobody acts on is just a number on a dashboard. Here is a review routine that fits a small team.
- Pull every 1, 2 and 3 from the period. At small volumes, that might be 5 to 15 tickets. Read every one.
- Tag each one with a cause. Use a short list, such as policy, product, shipping, slow reply, wrong answer or tone. Most months, two causes will cover most of the complaints.
- Separate what support controls from what it doesn't. Slow replies and wrong answers are yours to fix. Shipping delays and refund policy belong to whoever owns those decisions, so bring them the tagged tickets as evidence.
- Pick one fix. One. Rewrite a macro, change a returns page or adjust coverage on weekends.
- Check it next quarter, not next month. Given the sample sizes above, that is the honest timeframe.
If your help desk vendor charges extra for the reporting you need to run this review, that is worth noticing. Per-agent pricing and report-tier upgrades are a common reason small teams outgrow their tools. Our guide on what to check before your help desk bill goes up covers the warning signs.
Get Per-Store CSAT Reports Without a Per-Agent Bill
Everything above depends on splitting your satisfaction score by store and by person, and on comparing it against reply-time targets. Replaticket does exactly that. It pulls tickets from all your stores into one inbox, and each store keeps its own reply name and address. It tracks first-reply and resolution targets with warnings, and it reports satisfaction scores per store and per person. It costs $299 once, with full source code, 12 months of updates and 30 days of setup support. A seasonal hire costs you nothing extra.
The caveat is that it is self-hosted. You need a server and someone comfortable maintaining it, and that trade-off is laid out in our subscription vs. one-time help desk comparison. It also can't fix a small sample. Twenty responses is still twenty responses, so use the rolling windows above whatever tool you run.
Frequently Asked Questions
What is the formula for a CSAT score?
The most common formula is the number of satisfied responses, usually 4s and 5s on a 5-point scale, divided by total responses, multiplied by 100. Some teams report the average rating out of 5 instead. Pick one method and keep it, or your trend data becomes meaningless.
What is a good CSAT score for a small online store?
For most small e-commerce support teams, a score between 70% and 85% is normal, and below 70% usually points to a specific broken policy or process. Outside benchmarks use different survey methods, so they are poor comparisons. Beating your own trailing three-month average is a more useful target.
How many survey responses do I need for a reliable CSAT score?
Under about 100 responses, your score can move 7 points or more from chance alone, and at 30 responses the margin is roughly 13 points. If you get fewer than 100 responses a month, report a rolling 90-day score. Act only on changes that hold across two or three periods.
Should I measure CSAT or NPS for customer support?
Measure CSAT for support, because it asks about the specific interaction the customer just had. NPS measures overall loyalty to your brand and mostly reflects product and pricing, not the quality of a support reply. Small teams usually get more value from adding a customer effort question than from adding NPS.
Is it fair to rank support agents by CSAT?
Not on its own. Agents handle different ticket mixes, so someone working refunds will score lower than someone answering order-status questions. Small samples per person add even more noise. Use per-agent CSAT to choose conversations to review together, not to build a ranking.
Sources
- ISO 10004:2018 Quality management — Customer satisfaction — Guidelines for monitoring and measuring | International Organization for Standardization
- American Customer Satisfaction Index | ACSI
- The One Number You Need to Grow | Harvard Business Review
- Stop Trying to Delight Your Customers | Harvard Business Review
Leave a Reply