What Is Customer Effort Score (CES) and When to Use It

A customer leaves five stars and still doesn’t buy again next month. That happens constantly in cross-border ecommerce. CSAT measures whether the customer was satisfied with this interaction, and they’ll say yes because the problem eventually got solved. What they won’t tell you is that getting there took three transfers, a two-day wait, and repeating their order number four times. Customer Effort Score (CES) measures exactly that: how much work it took to get resolved. The more work it took, the less likely that customer comes back, even if this ticket had a happy ending.

What CES is: one question, one score

CES collection is deliberately simple. Right after a conversation closes, you ask one question: “How easy was it to resolve this issue?” and the customer rates it on a 1-5 or 1-7 scale (1 being “very difficult,” the top of the scale being “very easy”). Some teams phrase it as “How much effort did you personally have to put in to get this resolved?” Same idea, either way — you’re measuring friction, not emotion.

The metric traces back to CEB (now part of Gartner), which introduced it in 2010 with a core finding: reducing customer effort retains customers better than delighting them does. Compared to feel-good tactics like enthusiastic service or surprise perks, simply not making the customer jump through hoops delivers a much better return. A low CES (high effort) usually correlates with lower renewal and repeat-purchase rates; a high CES (low effort) tends to mean customers don’t bother shopping around — they just keep using you.

How CES differs from CSAT and NPS

The three metrics often get deployed together, but they answer different questions:

These aren’t mutually exclusive — they’re complementary. A customer can rate the outcome five stars (high CSAT) while finding the process exhausting (low CES). That combination is precisely the one worth worrying about: it means you won this round, but there’s no guarantee they’ll give you another shot next time.

Why effort predicts churn better than satisfaction

The logic is fairly plain: people remember hassle longer than they remember satisfaction. A customer won’t remember you for resolving something cleanly, but they will remember getting transferred three times and repeating their order number four times as “support at this store is a pain.” Next time something goes wrong, they’re unlikely to give you another chance — they’ll switch to a competitor or just stop buying.

In cross-border ecommerce, time zones and language gaps amplify effort further. A customer messages at 3am and waits eight hours for a reply; or asks a question in Spanish and gets bounced between agents until someone who speaks Spanish picks it up. The wait and the transfers themselves are what drive effort up — independent of whether the final answer was even correct.¹

How AI support cuts effort at the source

Effort’s enemy is repetition — repeated transfers, repeated information, repeated waiting. AI support is built to attack exactly those three:

• Resolve in one pass, cut transfers. The AI agent answers shipping timelines, return conditions, and discount rules straight from the knowledge base — when it has an answer, the customer is done, no bouncing between agents. When it doesn’t, it hands off once, instead of the customer looping through a bot maze before ever reaching a person.
• Shared workspace, cut repetition. AI and human agents share the same conversation context and customer profile. When a chat hands off from AI to a human, the agent sees the full history — order number, what’s already been asked, what the AI already answered — so the customer never has to repeat themselves.
• 24/7 response, cut waiting. A customer messaging at 3am doesn’t wait until business hours open. The AI picks it up immediately, resolves what it can on the spot, and for anything that needs a human, has the details already organized so the agent can start working the moment they’re online instead of from zero.

For how this AI-first, human-backed division of labor is drawn in practice, see AI answers first, humans back up: where exactly the line goes. And whether the AI can actually resolve things in one pass comes down to how well-fed the knowledge base is — more on that here: Build a knowledge base your AI can actually use.

How to actually collect CES

The default approach is a one-question survey right after the chat ends, but that’s just a starting point. A few details matter:

• Word the question specifically — don’t ask “were you satisfied,” ask “was this easy to resolve.” The two questions lead to very different answers.
• Keep the scale consistent across channels. Mixing 1-5 on one channel and 1-7 on another makes cross-channel comparison meaningless.
• Pair it with objective data — number of transfers, average wait time, how often customers repeated themselves. This fills the gap left by customers who never bother filling out a survey.
• Break it down by channel. WhatsApp, email, and the website widget usually score differently, and lumping them together hides a channel with a real problem.

A single survey question alone rarely gets enough responses — the customers willing to fill out a post-chat survey are already a minority. Pairing survey scores with system-native behavioral data like transfer counts and wait times gives you the fuller picture.

Low effort doesn’t mean zero human involvement

One misconception worth clearing up: cutting effort doesn’t mean routing every conversation to AI — it means not making customers jump through hoops. In some cases, what the customer actually wants is to talk to a person: complaints, refund disputes, complex custom requests. Forcing AI into those moments raises effort instead of lowering it, because the customer feels brushed off by a bot.

What actually lowers effort is matching the right handling to the right situation — simple questions get an instant AI answer, complex ones get handed off once and cleanly instead of getting punted around. High-risk actions like refunds, compensation, and price changes should still route through human approval. That’s not adding effort — it’s protecting the customer’s money and trust, and it doesn’t conflict with cutting effort elsewhere.

CES matters precisely because it isn’t measuring “was this interaction fine” — it’s measuring “will they come back.” Satisfaction can be bought back with one great gesture; effort debt sticks around a lot longer. Rather than waiting for NPS to slide before you look into it, find the friction points you can fix directly right now — transfer counts, wait times — and grind them down one at a time.