Customer Effort Score (CES)

Customer Effort Score (CES) measures how hard a customer had to work to get something done — resolve a ticket, find an answer, complete onboarding, or renew. You ask one question right after the interaction: “How easy was it to handle your issue?” and the customer answers on a scale. The lower the effort, the higher the loyalty — that’s the whole thesis behind the metric.

CES is not a relationship metric and it is not a satisfaction metric. NPS asks how likely someone is to recommend you (a relationship signal). CSAT asks how happy they were with an interaction (a sentiment signal). CES asks something narrower and more predictive of churn: how much friction did this specific interaction create? A customer can be satisfied (high CSAT) with a resolution that still took four emails and two days — and that effort is what makes them quietly start evaluating alternatives.

How you calculate it

The modern form (CES 2.0, the version most teams use) asks the customer to rate agreement with a statement on a 1-7 scale:

"The company made it easy for me to handle my issue."
1 = Strongly disagree  →  7 = Strongly agree

CES is the average of those responses:

CES = (sum of all responses) / (number of responses)

A 1-5 scale and a “very difficult → very easy” wording are both common variants — pick one and never change it, because CES is only useful as a trend. Some teams report a ”% top-2-box” (the share answering 6 or 7) instead of the mean, which is less sensitive to outliers. Whatever you choose, define it once and hold it.

What a good score looks like

On a 1-7 scale, 5.5+ is strong, 5.0-5.5 is solid, and below 4.5 signals a friction problem worth a root-cause pass. On a 1-5 scale, treat 4.0+ as the equivalent of strong. These are directional norms across B2B SaaS support and onboarding — your own baseline matters more than any benchmark, because question wording and scale shift the absolute number by a full point.

When CES beats CSAT and NPS

CES is the right metric when the thing you’re measuring is a transaction the customer wanted to be over quickly — a support resolution, a self-serve task, a billing change. The original CEB research (now Gartner) found CES predicts repurchase and loyalty better than CSAT in exactly these moments, because reducing effort prevents disloyalty more reliably than delighting customers creates loyalty.

Use the other two when:

• NPS — you want a relationship-level barometer across the whole account, untied to a single event. Good for executive dashboards and renewal-risk segmentation, weak for diagnosing a specific broken flow.
• CSAT — you want immediate sentiment on a single touchpoint and the interaction is one the customer wants (a QBR, a feature demo, a CSM call). Effort isn’t the point there; experience quality is.

Rule of thumb: CES for friction-reduction surfaces (support, docs, onboarding, self-service); CSAT for experience-quality touchpoints; NPS for the relationship. Many CS teams run all three and route each to the team that owns the lever — CS Ops watches NPS trend, support watches CES per resolution, the CSM watches CSAT on their own touchpoints.

Common pitfalls

• Surveying too late. CES decays fast; fire the survey within minutes to a few hours of resolution, not in a weekly batch. Guard: trigger it on the ticket-closed event in Zendesk or Intercom, not on a schedule.
• Treating the number as the work. A 4.2 CES tells you there’s friction; it doesn’t tell you where. Always pair the score with a free-text “what made it hard?” field and tag the responses — the qualitative tags are where the fix lives.
• Comparing across reworded surveys. Changing the question or scale resets the trend line. If you must change wording, run both versions in parallel for a quarter before cutting over.
• Averaging away segments. A healthy blended CES can hide a brutal score for one channel (phone vs chat) or one cohort (enterprise onboarding). Slice by channel, segment, and issue type before declaring the number fine.

Related

• NRR vs GRR — the retention outcomes CES is meant to predict
• Delighted and AskNicely — survey tools that run CES alongside NPS and CSAT