Key Takeaways
The support team had a CES target of 2.0 or better on a 5-point scale. They hit it every month. Then the new survey tool arrived with a 7-point agreement scale where higher meant better, and the dashboard showed 78%. Nobody could say whether that was an improvement. Meanwhile, the account that had just escalated to the VP had never answered a survey. Its three reopened tickets and two transfers were sitting in the helpdesk, unmeasured.
Most customer effort score guides explain the formula and quote the same statistics about effort and loyalty. Few mention that the question and scale changed, or what the one independent test of CES found. This post covers where CES came from, what the evidence supports, and how a B2B team can measure effort from records it already keeps.
Where the customer effort score came from
CES was introduced in a 2010 Harvard Business Review article by researchers at CEB. They surveyed “more than 75,000 B2C and B2B customers” about recent service interactions. Their conclusion was that reducing effort mattered more for loyalty than delighting customers.
Three details from that article matter for how CES is used today.
It measured intent. Loyalty was “defined as customers' intention to keep doing business with the company, increase the amount they spend, or spread positive (and not negative) word of mouth.” The widely quoted figures follow from that: “Of the customers who reported low effort, 94% expressed an intention to repurchase, and 88% said they would increase their spending.” Those are stated intentions, not renewals.
It covered service interactions only. The authors wrote that “our research focused exclusively on contact-center interactions.” CES was compared with other measures “in customer service interactions,” not across the customer relationship.
It was meant to be paired with records. This is the part most guides leave out. The authors wrote that teams “can use CES, along with operational measurements of such things as repeat calls, transfers, and channel switching, to conduct an ‘effort audit.’”
What the early adopters did with it
The article's own examples were mostly about records, not scores. Bell Canada “met this challenge by mining its customer interaction data to understand the relationships among various customer issues.” It answered the likely next question in the first contact, and the company reported that this cut “calls per event” by 16% and its customer churn by 6%. Another company started a chat with any customer “who has spent more than 90 seconds in the knowledge center or clicked on the ‘Contact Us’ link.” It reported an 8% drop in calls.
These are company-reported results from 2010, without controls. But they show where the useful signal came from: interaction data and help-center behaviour. One B2B example did use the score. Osram Sylvania lowered its CES “from 2.8 to 2.2, 18.5% below the average we see for B2B companies,” by changing how agents spoke to customers.
The article also compared channels. “2.4 e-mails, on average, are needed to resolve an issue, compared with 1.7 calls.” Contacts per issue was already the unit that mattered.
CES 1.0 and CES 2.0 aren't the same metric
The original question used a 1 to 5 scale where “1 (very low effort) to 5 (very high effort).” Lower was better.
The later version reworded it. One CX consultancy notes the question became “The company made it easy for me to handle my issue”. MeasuringU explains that “the CES scale has transformed into a seven-point agreement/disagreement scale and scored differently.” Higher is now better, and it's often reported as the share of customers who agree.
That's why CES benchmarks rarely compare. A target below 3 on the old scale and a target of 80% agreement on the new one describe different scales pointing in different directions. If you changed survey versions, your trend line broke at the switch.
What CES predicts, according to the evidence
The only behavioural test we found is a 2015 study of 93 Dutch firms. The researchers compared survey metrics against retention the customers reported two years later.
Two findings stand out. First, coverage: “79.5% of the customers had not contacted the firm with a request and thus could not provide a CES.” Second, for the customers who could answer, “the CES is the only CFM that performs worse than the baseline model,” though that subsample was small.
There's a useful counterpoint. The same authors found “the best improvement occurs when we combine the top-2-box customer satisfaction with the CES.” CES adds something alongside satisfaction. It just shouldn't carry the weight alone. Our comparison of NPS vs CSAT vs CES covers the wider evidence.
The study also contains a detail that cuts against a common assumption. In its raw correlations, having contacted the firm at all was positively related to retention. Contact isn't the problem. Contact that takes too much effort is. Count effort per issue, not contacts per account.
The survey's own weaknesses
CES's co-creator has since pointed at the survey format itself. A 2019 press release from a company the co-creator later joined said “the CES is a survey question which means that it suffers from low--and, for most companies, declining--response rates, weak customer verbatim and well-known flaws.” That release was launching a competing product, so read it with that in mind. The replacement it describes predicts “the score a customer would have given on a survey”, not whether the customer stays.
The effort audit, as its authors described it
Here's the audit CEB recommended, mapped to fields most B2B helpdesks already record.
| Effort signal | What CEB said | Where it already sits | What to count |
|---|
| Repeat contact | Pair CES with repeat calls | Ticket reopens; same issue within 7 to 14 days | Repeat contacts per issue |
| Transfer | Pair CES with transfers | Number of groups and assignees a ticket passed through | Tickets with more than one handoff |
| Channel switch | Pair CES with channel switching | Tickets that follow a help-center visit or chat | Switches per issue |
| Wait | Not named by CEB | Requester wait time in business hours | Wait per resolved issue |
| Self-service before a case | Not named by CEB | Help-center session before a ticket | Share of tickets preceded by self-service |
Sources: HBR, 2010; Zendesk ticket metrics documentation; KCS v6 Practices Guide. The last two rows are our additions.
Channel switching deserves attention. Gartner research reported in trade press found 62% of customer service channel shifts are high effort, in a survey of B2B and B2C customers. We couldn't reach the original report, so treat the figure with care.
Self-service needs care too. KCS warns that its measures “by themselves” don't represent the user experience, and that “it is the trends that are most important, not the absolute value.” Our guide to customer journey friction points shows how to pull each of these from your records.
Running the audit
- Pick a quarter of closed requests. Group them by issue, not by ticket.
- Count the five signals per issue. Reopens, handoffs, channel switches, wait time and whether self-service came first.
- Rank topics by total effort. The top five usually account for most of it.
- Read the requests behind each topic. Look for missing articles, unclear ownership or a step customers keep getting stuck on.
- Fix one topic and re-run next quarter. Compare the same five signals for that topic.
Where CES still fits
CES is still useful in a narrow role. Send it after specific service moments where you suspect friction, such as onboarding tickets or billing questions. Read it next to satisfaction, not instead of it. And keep one scale. If you switch versions, restart the trend.
In B2B, remember who's answering. The person who filed the ticket often isn't the person who decides the renewal. B2B research on customer health describes it as combining “objective and subjective data points,” so a survey is one input, not the verdict.
Effort data on one record
The effort audit needs tickets, help-center activity and accounts in one place. In most stacks, they aren't. Reopens and transfers live in the helpdesk. Searches live in the help-center tool. Account context lives in the CRM.
MatrixFlows keeps them together. Customer requests, accounts and knowledge articles are linked tables in Matrix. Conversations in the Inbox record who handled each request and when. App analytics show what people searched for in the help center and what the AI assistant answered before they asked a person. Views can count repeat contacts and handoffs per issue, by account. Forms can still send a CES survey after a request closes, with the answer stored next to the operational data.
Back to the escalated account. Its three reopens and two transfers told the story weeks before the VP called. With effort measured from records, that account would have been on the list.
Create a Free Workspace → and run your first effort audit from records this week.
.svg)
.svg)
.svg)
