# How to Calculate Customer Effort Score: Formula and Example

Learn how to calculate Customer Effort Score with a worked example, explicit response coding and denominator, and clear limits on comparing results.

- Canonical URL: https://www.harmate.com/en/blog/how-to-calculate-customer-effort-score-formula-and-example
- Author: Harmate Team
- Published: 2026-10-28
- Updated: 2026-10-28T17:07:06.370422+00:00
- Language: en

## Content

# How to Calculate Customer Effort Score: Formula and Example

Customer Effort Score (CES) summarizes an answer about the effort involved in a customer experience. To calculate it clearly, keep the exact question, response labels, numeric coding, aggregation rule, and number of valid answers together. The acronym CES alone does not determine which way a scale runs or which formula to use.

Published guides do not all choose the same scale direction or aggregation. This article gives one fictional calculation that readers can reproduce, then explains what to check before comparing results. It offers no universal “good CES” threshold and does not claim that a score explains customer loyalty.

## What does a CES question measure?

CES captures the effort a person says they had to make to complete a task or resolve a request. It can complement a broader satisfaction measure by focusing attention on one step in a journey. The question should name that task: “How easy was it to resolve your request today?” asks about a different experience from “How much effort did you have to put in?”

Timing matters too. A question sent after an interaction may capture an assessment of that interaction; a general question about a brand relationship has a different scope. Qualtrics describes CES as a measure of effort around an interaction and discusses question timing and a follow-up comment ([Qualtrics’ Customer Effort Score overview](https://www.qualtrics.com/articles/customer-experience/customer-effort-score/)). That helps define the subject, but it does not by itself prescribe one calculation formula.

Before counting, write down what one row represents: an answer about an interaction, one answer per person over a period, or something else. Keep that choice with the measure. If one customer answers after three support requests, you may have three documented interactions, but not three distinct customers.

## Check the scale direction and coding

Two plausible question wordings can make higher numbers mean opposite things. These examples are illustrative, not standardized questions:

| Fictional question | Coded responses | What a higher code means |
|---|---|---|
| “How easy was it to resolve your request?” | 1 = very difficult; 2 = difficult; 3 = neither difficult nor easy; 4 = easy; 5 = very easy | More reported ease |
| “I had to put in a lot of effort to resolve my request.” | 1 = strongly disagree; 5 = strongly agree | More reported effort |

Do not copy codes from an example without checking your questionnaire’s labels. A 5 can indicate an easy experience in one question and substantial effort in another. Preserve the original response and write down the recoding table before calculating.

Vendor guides illustrate different conventions. SurveyMonkey, for example, presents an average where lower values are favorable; Zendesk describes variants where higher values or a percentage of responses are used ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/how-to-use-customer-effort-score/), [Zendesk](https://www.zendesk.com/blog/customer-service/satisfaction/customer-effort-score/)). These pages describe their own choices. They do not make scores built from different questions, scales, or rules directly comparable.

## Worked example: calculate an average and a favorable share

A team asks after each interaction: “How easy was it to resolve your request?” It uses a scale from 1 (“very difficult”) to 5 (“very easy”), where a higher number indicates more reported ease. Its eight valid answers are **5, 5, 4, 4, 4, 3, 2, 1**.

Under this coding, the descriptive arithmetic mean is:

**(5 + 5 + 4 + 4 + 4 + 3 + 2 + 1) ÷ 8 = 28 ÷ 8 = 3.5 out of 5.**

The categories are ordered, but the distance between “difficult” and “neither difficult nor easy” is not necessarily equal to the distance between other choices. The mean is therefore an explicitly chosen summary convention, not a measurement of equal intervals. Show the count in each category too; here, the distribution includes one answer at 1 and two answers at 2 or 3.

For the same responses, the team could also describe the share rated 4 or 5. It should define this convention before inspecting the results and label it clearly as “favorable responses under the chosen threshold”:

**(5 favorable responses ÷ 8 valid answers) × 100 = 62.5%.**

These figures do not say the same thing. The mean uses each response code; the favorable share counts only answers above a chosen threshold. A different threshold or scale would produce a different proportion. Here, 3.5/5 is a mean ease rating. It is not a universal formula that every tool or questionnaire necessarily calls “the CES.”

Show the wording, scale, coding, formula, and denominator alongside the number.

## Which denominator should you use?

The denominator should match the question you are answering. For the mean above, it is the **8 valid answers to the CES item**. Missing, incomplete, or “not applicable” answers do not become zeros and should not automatically be replaced with the scale midpoint.

Suppose 10 eligible interactions were invited, but only 8 produced a usable answer to the question. You can describe CES across the 8 valid answers and report coverage separately: 8 usable answers out of 10 invited interactions. Do not say that 80% of customers responded unless you know each customer received only one invitation or the 10 units are known to be 10 unique customers.

Distinguish answers, interactions, and people. If one person can answer after several interactions, state whether the denominator counts valid answers, eligible interactions, or identified people; do not mix these bases in a time series.

## Add an open question without confusing it with the score

An open question can help locate the specific step that took effort: “Which step took the most effort, and what happened?” Also allow the person to say that no step was difficult, the situation did not apply, or something else mattered.

A comment describes what someone chose to explain in that context. It may add detail or an exception the scale did not show, but it does not turn the score into a diagnosis or prove the cause of a result. Keep the context and cases that do not support the most common explanation. For help drafting a neutral prompt and reviewing answers in their original wording, see our guide to [open-ended survey questions](/en/blog/open-ended-questions-get-actionable-data-without-bias).

## When can you compare two results?

A comparison is easier to interpret when the question, options, and coding stay the same; the timing in the journey is comparable; the unit and population are defined; and the same rule is used for missing answers. If one of these elements changes, explain the change before describing a trend.

Do not directly compare a 3.5 mean on an ease scale with a 3.5 mean on an effort scale coded in the opposite direction. Nor should you compare an average with a share of favorable answers as if they were the same indicator. For a reminder about selecting and coding response scales, see [Likert scale examples and analysis errors](/en/blog/likert-scale-examples-response-options-and-analysis-errors).

A CES score describes the answers collected under your protocol. By itself, it does not provide a universal threshold, explain the reported effort, prove future loyalty, or estimate the effect of an intervention. For other questions that can support a customer-satisfaction survey, see [25 questions to adapt to a customer decision](/en/blog/customer-satisfaction-survey-25-ready-to-use-questions).

## Frequently asked questions

### What is the Customer Effort Score formula?

The formula depends on the question, scale direction, and convention you choose. A mean of response codes is one option; the proportion of responses within a defined range is another. Name the method and do not present one convention as a universal rule.

### Is a higher CES always better?

No. Read the response labels and coding direction. A high number can represent more ease or more effort depending on the question.