Customer effort score (CES) measures how easy it was for a customer to get something done with you. It usually comes from one question asked right after a support conversation: "[Company] made it easy for me to handle my issue," answered from 1 (strongly disagree) to 7 (strongly agree). Your CES is the average of those answers, or the share of people who chose 5, 6 or 7. As a rule of thumb, an average above 5, or 70% or more answering 5 to 7, is a good score.
That rule of thumb comes from vendor guides, not an audited public benchmark, so the more useful comparison is your own score month to month and topic by topic. The rest of this guide covers the question wordings, the math with a worked example, when to send the survey and what a small team can change to bring effort down.
What customer effort score is
CES started with a 2010 Harvard Business Review article, "Stop Trying to Delight Your Customers", by Matthew Dixon, Karen Freeman and Nick Toman of the Corporate Executive Board, which Gartner later bought. Their argument was that going above and beyond in service does less for loyalty than simply making problems easy to solve.
Gartner has kept publishing on the idea. In an August 2020 release it reported that customers who had a low-effort service experience were 61% more likely to stay with the company, and that a high-effort experience dropped that probability to 37%.
The metric picks up something satisfaction surveys often miss. Imagine a customer who writes in three times about the same delivery, gets a friendly agent each time and finally gets the parcel. Asked whether they were satisfied with the last agent, they may say yes. Asked whether it was easy, they will probably say no, and that is the answer that predicts whether they order again.
The CES question and scales
There are three common versions. They are not interchangeable, so pick one and keep it.
| Version | Question | Scale | Higher means |
|---|---|---|---|
| Agreement (often called CES 2.0) | "[Company] made it easy for me to handle my issue." | 1 strongly disagree to 7 strongly agree | Easier |
| Ease | "How easy was it to get the help you needed today?" | 1 very difficult to 5 very easy (or 1 to 7) | Easier |
| Original effort question | "How much effort did you personally have to put forth to handle your request?" | 1 very low effort to 5 very high effort | Harder |
The original wording is the one to avoid. "Effort" reads differently to different people, and because a high number is bad, it is easy to mix up with every other survey you run. The agreement version on a 7-point scale is the most widely used today, and it is the one the benchmarks in the next sections assume.
A few wording choices make the answers more reliable:
- Name the specific thing: "handle your refund" gets sharper answers than "handle your issue" when you know the topic.
- Label both ends of the scale in words, and the middle if there is room.
- Add one optional open question, "What made it harder than it should have been?", and show it only to people who answered 1 to 4. That comment is where most of the value is.
- Ask nothing else. Every extra question lowers the number of people who finish.
Examples you can adapt by channel:
- After a chat: "We made it easy to sort out your question today."
- After an email ticket: "It was easy to get my problem solved."
- After a return: "Returning my order was easy."
- After reading a help article: "This article made it easy to do what I came to do."
When and where to send a CES survey
CES measures a single interaction, so it belongs right after that interaction ends. For support, that means when a conversation is marked solved, not after the first reply. A customer who has just received "we're looking into it" can't tell you yet whether it was easy.
Ask in the same channel the customer used. A rating inside the chat window or at the bottom of the closing email gets far more answers than a link to a separate survey a day later. If you want to see how a short rating looks to customers before you send it, try it in our survey previewer; the layout advice is the same for an effort question.
Limit how often one person is asked. Someone who writes in every week should not get a survey every week. Once a month per customer is a sensible ceiling for most small teams.
Finally, mind the sample size. With small volumes the average moves a lot. If 20 responses average 5.5, a single answer of 1 drops the average to 5.29. Read CES monthly rather than weekly until you collect a few hundred answers a month, and always look at the count next to the score.
How to calculate customer effort score, with a worked example
There are two standard ways to turn the answers into one number, plus a third that some teams like.
- Average: add up every answer and divide by the number of answers.
- Percentage of easy answers: count the answers of 5, 6 or 7 and divide by the total.
- Net effort: subtract the share of 1 to 3 answers from the share of 5 to 7 answers, similar to how NPS works.
Here is a hypothetical month for a small store that sent the 7-point agreement question after each solved conversation and got 40 answers:
| Answer | Responses | Answer × responses |
|---|---|---|
| 7 strongly agree | 12 | 84 |
| 6 | 10 | 60 |
| 5 | 6 | 30 |
| 4 neutral | 4 | 16 |
| 3 | 3 | 9 |
| 2 | 3 | 6 |
| 1 strongly disagree | 2 | 2 |
| Total | 40 | 207 |
- Average CES: 207 ÷ 40 = 5.2.
- Percentage of easy answers: (12 + 10 + 6) ÷ 40 = 28 ÷ 40 = 70%.
- Net effort: 70% easy minus 20% hard (8 of 40 answered 1 to 3) = 50.
In a spreadsheet with the answers in column B, the average is =AVERAGE(B2:B41) and the easy share is =COUNTIF(B2:B41,">=5")/COUNT(B2:B41).
Which one to report? The percentage is the easiest to explain to a team ("seven in ten customers found it easy"), and it doesn't hide a cluster of very bad experiences the way an average can. Whichever you choose, use the same method every month.
What a good customer effort score is
Most published guides land on similar numbers for the 7-point agreement scale. These are rules of thumb repeated across vendor blogs, not results from an independent study, so treat them as a rough starting point:
| Average (1 to 7) | Easy answers (5 to 7) | How to read it |
|---|---|---|
| 6 or above | 85% or more | Very low effort. Check that the survey reaches unhappy customers too. |
| 5 to 6 | 70% to 85% | Good. Work on the topics that score lowest. |
| 4 to 5 | 50% to 70% | Customers are working too hard somewhere. Read every low-score comment. |
| Below 4 | Under 50% | Usually a broken process. Find it this week. |
Two cautions. First, if you use the original effort question, lower is better, so none of these bands apply. Second, a company-wide CES hides the useful detail. Split it by topic and by channel. A store might score 6.3 on order-status chats and 4.1 on returns, and that gap tells you exactly where to work.
CES vs CSAT vs NPS
The three metrics answer different questions, and most small teams only need two of them.
| CES | CSAT | NPS | |
|---|---|---|---|
| Question | "[Company] made it easy to handle my issue." | "How satisfied were you with the help you got?" | "How likely are you to recommend us to a friend or colleague?" |
| Scale | 1 to 7 agreement | 1 to 5, or good / bad | 0 to 10 |
| Measures | How hard the process was | How the customer felt about this interaction | Overall relationship with the company |
| When to ask | After a solved conversation or a task | After a solved conversation | Every few months, not tied to a ticket |
| Best for spotting | Broken processes and repeat contacts | Agent tone and quality | Churn risk and product sentiment |
A practical setup for a team of two to twenty: run CSAT after every solved conversation, since most help desks support it out of the box, and add CES for one or two processes you suspect are painful, such as returns or account changes. Run NPS separately, a few times a year. Our CSAT calculator turns raw ratings into a score if you want the two side by side.
How to lower customer effort on a small team
Effort usually comes from the same handful of places. Each one has a fix that doesn't require a bigger team.
Customers have to write in more than once
A second contact about the same issue is the clearest sign of effort. Track it as a number: the share of conversations reopened within a week, or contacts per resolved issue. The fix is usually a more complete first answer, with the next step and a date included.
Customers repeat themselves
If chat, email and social messages live in different tools, the second person to answer has no idea what the first one said. A shared inbox keeps the history with the conversation, so nobody asks for the order number twice.
Customers are pushed to another channel
"Please email us about that" turns one conversation into two. Answer where the customer wrote, and only move channels when there is a real reason, such as a document that needs to be attached.
Simple questions wait in a queue
Questions like delivery times, password resets or plan limits shouldn't need a person at all. A help center covers part of it. An AI agent on website chat covers more: in Sellio, the AI agent answers on website chat from your own site and documents and hands the conversation to a person when it isn't sure. Messages on other channels still go to the team.
Customers wait without knowing how long
Waiting is tolerable when it has an end. Tell customers when they will hear back, and follow up even if the answer is "still working on it". Measuring first response and resolution time with our support KPI calculator shows whether those promises are realistic.
Policies are harder than they need to be
Low CES on returns or cancellations is often a policy problem. If a return needs a photo, a form and an email, ask which of the three you actually use.
Common mistakes with CES
- Mixing scales. Comparing a 1 to 5 ease score with a 1 to 7 agreement score, or with the inverted effort question, produces numbers that mean nothing.
- Surveying too early. Asking after the first reply measures speed, not ease.
- Ranking agents by CES. Effort mostly comes from processes and policies an agent doesn't control. Use CSAT and reply reviews for coaching.
- Reading the score without the comments. The number tells you whether there is a problem; the low-score comments tell you what it is.
- Reacting to tiny samples. Ten answers is an anecdote. Wait for the month.
If you are still choosing tools for a small team, our Freshdesk vs Help Scout comparison covers two help desks many teams consider.
FAQ
What is a good customer effort score?
On the 7-point agreement scale, an average above 5 or at least 70% of answers between 5 and 7 is generally considered good, and above 6 is very good. These are rules of thumb from vendor guides; your own trend over time is a better benchmark.
What is the formula for CES?
The most common formula is the sum of all answers divided by the number of answers. The alternative is the number of 5, 6 and 7 answers divided by the total, shown as a percentage.
Should CES use a 5-point or 7-point scale?
Either works if you never change it. The 7-point agreement scale is the most widely used, so it is easier to compare with published figures. A 5-point scale is slightly faster to answer on a phone.
Is CES better than NPS?
They measure different things. CES tells you whether a specific interaction was easy and works well for fixing support processes. NPS measures the overall relationship and is better for tracking churn risk across the whole customer base.
How often should you send a CES survey?
After each solved conversation or completed task you want to measure, with a cap of about once a month per customer so regular contacts aren't asked every time.
Can you measure effort without a survey?
Partly. Repeat contacts per issue, reopened conversations, transfers between people and the number of messages before resolution are all effort signals you can pull from your inbox data. They don't replace asking the customer, but they cover the people who never answer surveys.
