What is customer effort score?

Where the idea came from

The research behind it, published in the Harvard Business Review in twenty ten, argued that service interactions rarely build loyalty and frequently destroy it, and that the reliable lever is removing effort rather than exceeding expectations. The practical conclusion was that a customer who had to call twice, repeat their account number three times and wait for a promised callback is at risk regardless of how friendly the final conversation was. That reframing is what makes the metric useful operationally: unlike satisfaction, effort maps onto specific steps in a process that somebody can change. The original question was worded around how much effort the customer personally had to put in, and the more common current form uses an agreement statement instead because it tests better with respondents.

Getting the wording right

The widely used version reads as a statement to agree or disagree with: the company made it easy for me to handle my issue. Anchor the scale from strongly disagree to strongly agree and label every point where you can. Avoid asking how much effort did you have to expend, which sounds unnatural in most English speaking markets and produces a reversed scale that some respondents will answer backwards. Name the specific interaction rather than the company in general, since effort is a property of an episode. If the survey follows a support ticket, say so in the question, because a customer who has had three separate interactions this month will otherwise answer about whichever one they remember.

When to send it

Immediately after the interaction closes, while the friction is still remembered, and once only. Effort decays from memory faster than satisfaction does, so a survey sent three days later measures a reconstruction. Sending it after every single contact will train customers to ignore it, so pick a trigger that matters, such as a closed ticket or a completed onboarding, and sample rather than surveying everybody if volumes are high. One more timing trap: sending an effort survey after an interaction that is not actually finished, such as a ticket closed by automation while the customer is still waiting, produces a low score and an angry open comment, which is useful information about the process rather than about the agent.

Acting on the result

The score is only worth collecting if low scores are routed somewhere. The usual pattern is a threshold that triggers a follow up, plus a weekly read of the open comments grouped by the step they mention. Effort problems cluster in predictable places: authentication, handoffs between teams, repeated information, and any point where a customer has to chase. Because each of those is a process rather than a person, the metric works badly as an individual performance measure and tends to produce defensive behaviour when used that way. Track it by journey step or by queue instead, and report the distribution alongside the average so a small group of very high effort experiences is not hidden.

Building it in a form

A single effort question is a radio field with the agreement scale as options, laid out in one, two or three columns depending on label length. The open follow up is a long text field, left optional. A one question at a time layout suits a post interaction survey, since the whole thing is two questions and a thank you screen. Note that a classic field's description is not shown to the respondent, so any context such as this is about your recent support request belongs in the label or in a paragraph block above the field, or use a conversational form where the description does render.

Reading the comments by process step

The fastest way to make an effort programme useful is to code the open comments against the steps of the journey rather than against sentiment. Authentication, finding the right team, explaining the problem again, waiting for a promised reply, and the resolution itself is a serviceable starting frame for most support processes. Code a hundred comments against it and the distribution usually points at one or two steps carrying most of the friction, which is a far more actionable finding than an average. Keep the frame stable between rounds so the comparison holds, and add a code rather than redefining one when something new appears. It is also worth counting how often a comment mentions a step that the organisation does not consider part of the journey at all, since those are the gaps nobody is measuring and they are where the largest improvements usually sit.

Questions people ask