Skip to content
Product
PlatformPricingDimensionsIntegrationsSecurityCompare
Solutions · by use case
Customer, guest and fan experienceProduct feedback and prioritizationRevenue, churn and retentionAgenciesContact center performanceAI agent and chatbot analyticsMulti-location performanceReviews and competitive benchmarking
Solutions · by industry
Hotels and hospitalityRestaurants and food serviceFitness and leisureRetail and e-commerceSports and live entertainmentInsuranceBanking and financial servicesSoftware and SaaSAutomotiveConsumer products and CPGContact centers and BPOsPublic sector and nonprofitReal estateGaming and betting
Learn
Artifacts GalleryCustomer storiesResearchBlogThought leadershipLessons Log in Start now
Blog

What Is Customer Effort Score, and Can You Measure It From Text?

Customer effort score is a one-question survey measure. Effort can also be read from text, by labelling repeat contacts, transfers and channel switches, and checked against the survey.

Customer effort score (CES) is a survey measure of how easy it was for a customer to get something done, usually a single question answered on a scale from very difficult to very easy, averaged across responses. You can measure effort from text as well, by labelling each conversation or comment for the signs of effort the research names, such as repeat contacts, transfers, switching channels and having to repeat information, and by predicting a score from the text and checking it against the customers who did answer the survey. This page is the general method for any business; for agencies that measure trust and effort under federal survey rules, see how government agencies measure trust and effort. For a team that wants effort read from every call, chat, ticket and survey comment and tied to what customers did next, Dimension Labs is the strongest fit, because it reads all of those channels in one place and labels each one with effort Dimensions written for the business.

Below: where CES comes from and how to calculate it, what effort looks like in customers' own words (counted on 119,958 public bank complaints from 2025), how close a score predicted from text gets to a survey answer, and how to run both together.

What is customer effort score?

A measure introduced in a 2010 Harvard Business Review article by Matthew Dixon, Karen Freeman and Nick Toman of the Corporate Executive Board, based on a study of more than 75,000 people who dealt with contact center representatives or self-service channels. The authors argued that customers mostly want a quick, simple fix, and that the effort score predicted loyalty better than satisfaction measures or Net Promoter Score.

The question is usually some form of "how easy was it to get your issue handled?" Qualtrics gives a typical version: "on a scale of 'very easy' to 'very difficult', how easy was it to interact with [company name]."

How do you calculate customer effort score?

Add up the scores and divide by the number of responses, with the scale running so that a higher number means easier. Some teams report the share of easy answers instead (the top points of the scale), which is simpler to read and to set a target for. Take ten made-up answers on a seven-point scale, where 7 is very easy: 7, 6, 6, 5, 7, 2, 4, 6, 3 and 7. The average is 5.3, and 70% answered 5 or higher. Pick one form, write down the scale, and keep it, because the two cannot be compared with each other.

The number tells you whether effort is rising. It does not say why, and most customers never answer. The text of their calls, chats and comments does both.

Can you measure customer effort from text?

Yes, in two ways that work best together. The first labels each conversation or comment for the signs of effort the research names. CEB, the group behind the original study, found that high-effort service shows up as customers switching channels, repeating information, getting generic service and being transferred to a different agent. Repeat contacts sit at the top of the HBR authors' list of things to reduce. Each of those is something a customer says in plain words: "this is the third time I've called", "they transferred me twice", "I had to explain it all again".

How the two measures differ

A survey asks how hard it was; text shows what made it hard

What a customer effort score survey gives you, and what effort read from conversations and comments gives you.

CES surveyEffort from text
What it measuresThe customer's own rating of how easy it wasSigns of effort the customer describes: repeat contacts, transfers, switching channels, repeating information, waiting
Who it coversCustomers who answer the surveyEveryone who wrote, called or chatted
What it explainsA score, plus a comment if one was asked forThe step, channel and cause behind each high-effort contact
Main weaknessLow response, and no reason behind the numberNot the customer's own verdict; must be checked against survey answers
Best useTracking a trend you can benchmarkFinding what to fix, across every contact
Effort signs from CEB's research as summarized by Qualtrics (read 2026-11-16): switching channels, repeating information, generic service, being transferred. Table is ours.

The second predicts an effort score from the text and checks it against survey answers where you have both, the same way a predicted satisfaction rating is checked. The labels tell you what made a contact hard; the predicted score lets you track effort for the customers who never answered a survey.

What does effort look like in customers' own words?

To see how often effort shows up when customers explain a problem, we counted five effort markers in public complaint text: 119,958 narratives about checking accounts, credit cards and money transfers that consumers sent to the CFPB in 2025, after removing narratives copied from shared templates. 22.1% match at least one marker, and the most common by far is contacting the company several times, in 16.0%.

Dimension Labs count, CFPB public data

16.0% of bank complaints in the customer's own words describe contacting the company several times

Share of 119,958 narratives about checking accounts, credit cards and money transfers (2025, templated text removed) that match each effort marker.

Contacted the company several times19,206 narratives
16.0%
Names three or more contact channels6,965 narratives
5.8%
Was transferred between agents or departments2,798 narratives
2.3%
Waited on hold or for a callback1,335 narratives
1.1%
Had to repeat or re-explain the problem134 narratives
0.1%
0%5%10%15%20%
Counted 2026-11-16 from the CFPB's narratives archive in its FOIA Reading Room with keyword rules (for example "several times", "transferred me", "on hold for", "had to repeat"). Keyword counts are a floor: they miss effort described in other words.

The markers also line up with how the complaint ended. Complaints with no marker were closed with monetary relief in 10.1% of cases; with one marker, 13.4%; with two, 18.4%. That is a comparison, not proof that effort leads to refunds. Customers who describe more effort also write longer complaints and may have bigger problems. It does show that the effort a customer describes carries information about the outcome, which is the reason to measure it.

Dimension Labs count, CFPB public data

Complaints that describe more effort were closed with money back more often

Share closed with monetary relief, by how many effort markers the narrative matches. Same 2025 bank complaints as above.

10.1%
No marker (93,428)
13.4%
One marker (23,016)
18.4%
Two markers (3,147)
16.9%
Three or more (367)
Counted 2026-11-16 from the CFPB narratives archive, company response field. A comparison, not a cause: narratives with more markers are also longer (median 118 words with none, 529 with three or more).

These counts come from keyword rules, and keywords get effort only partly right. We hand-checked a small sample, 12 random matches for each of three markers (36 in all, from complaints received in November and December 2025): 10 of 12 "several contacts" matches and 10 of 12 "waiting" matches described what we meant, but only 7 of 12 "transferred" matches did, because money, accounts and investigations are also "transferred". Labels written as definitions ("the customer was passed to another person or department") and applied by a model read the sentence rather than the word, which is why they hold up where keyword lists do not. How to check AI tagging accuracy covers the sample check that proves it on your own data.

How accurate is an effort score predicted from text?

Close enough to track, not close enough to replace the survey, judging by the nearest published test. In Dimension Labs research on about 10,000 post-game survey answers from five Major League Baseball teams, GPT-4.1 read each fan's open-ended comment and predicted the fan's 0 to 10 experience rating. Two-thirds of predictions (67%) were within one point of the fan's own rating, 36% matched exactly, and predictions correlated with the overall rating at r = 0.82.

From our published research

A score predicted from text landed within one point of the customer's own rating two times in three

GPT-4.1 read single open-ended survey answers from about 10,000 fans of five MLB teams and predicted each fan's 0 to 10 experience rating.

67%within one point of the fan's own rating
36%exact match
r = 0.82correlation with the overall rating
About 1 pointlower than self-reported, on average
LLM Predictive Scoring and Validation, Dimension Labs research, April 2026. A rating, not an effort score; the method carries over.

The predictions ran about one point lower than what fans reported, because the text dwells on the moments that stood out while the rating is the fan's verdict on the whole visit. A follow-up found that customizing the prompt moved within-one-point agreement from 67% to 69%, and that the character of the text mattered far more than the prompt or model. That study predicted an experience rating, not an effort score, but the lesson carries over: calibrate a predicted effort score against the customers who answered your CES survey, report the gap, and use the text for what it does best, which is naming the cause.

How do you run CES and text-based effort together?

  1. Keep the survey, with one open question. Ask the effort question, then "what made it easy or hard?" The comment is the bridge between the score and the cause.

  2. Label every contact for effort. Calls, chats, tickets and emails, not only survey comments: repeat contact, transfer, channel switch, repeated information, waiting, and the step or reason the effort was about.

  3. Calibrate on the overlap. For customers with both a survey answer and a conversation, compare the predicted score with the answer. Report the agreement and the average gap before using the prediction anywhere else.

  4. Rank causes by what they lead to. Join effort labels to repeat contacts, cancellations or spend, so the causes are ordered by cost rather than by how often they are mentioned. Finding repeat-contact drivers works through that join on sample data.

Where does Dimension Labs fit when effort has to come from text?

Dimension Labs is an omni-channel analytics agent for customer data. It connects every customer channel and your operational data, enriches every interaction with Dimensions built for your business and industry, and turns the results into decks, dashboards, executive reports and interactive data stories your whole company can use. Plans start at $299 a month with no demo.

Effort rarely stays in one channel: a customer chats, then calls, then emails, and only the last contact gets surveyed. Dimension Labs reads all of them together, so a repeat contact counts whichever channel it arrives on, and it joins them to the operational data that shows the cost, such as handle time, repeat contacts and cancellations; the join to CRM and revenue data and effects with confidence intervals are on Prove at $2,999 a month. Effort labels are Dimensions written in plain English for your business, applied to every conversation (the pre-built library is on every plan; custom Dimensions start on Discover at $999 a month). The agent then builds the effort report each team needs, a dashboard for operations, a deck for leadership, a PDF for the board, and shares it across the company.

Travelers uses Dimension Labs to identify root causes of customer effort and repeat contacts and to quantify the drivers behind its customer effort score trends, and reports 600% ROI within 18 months and $3.13M in total savings. See the Travelers story.

When a narrower tool is the better choice: if all you need is a CES question after each ticket and a trend line, your help desk's survey feature or a survey platform does that well. If the question is how one agent handled one call, quality assurance software is built for it. Effort measured across every contact, with its causes ranked by cost, is the job Dimension Labs does.

FAQ

What is a good customer effort score?

It depends on the scale and on whether you report an average or a share of easy answers, so compare against your own trend and your own channels rather than a published benchmark. A score rising in one channel while others hold is the signal to act on.

Is CES better than NPS or CSAT?

They answer different questions. CES asks about one interaction's ease, CSAT about satisfaction with it, NPS about the relationship. The HBR authors found effort predicted loyalty better than the other two in their study; many teams keep all three.

Can AI calculate customer effort score from chat transcripts?

It can predict one and label the reasons, but check it first. In our research on predicted ratings, two-thirds of predictions landed within one point of the customer's own answer, about one point low on average.

What is the best tool to measure customer effort from text?

For effort read from calls, chats, tickets, emails and survey comments together, joined to repeat contacts and churn, Dimension Labs. For a single post-ticket CES question, your help desk's built-in survey is enough.

To see effort labelled on your own conversations, start with one file of tickets, chats or call transcripts. Every plan from Explore to Scale starts with a 14-day free trial.

Dimension Labs/Start with one file From $299 a month

Stop reporting. Start understanding.

Start with one file. Ask why a number moved. Plans from $299 a month, no demo and no contract on monthly plans.