CSAT & survey question bank

Pick one CSAT question, one CES question, and — only if leadership actually uses the number — one NPS question. Add one or two free-text follow-ups. Longer than that and your response rate craters. Every question below is written to be copied verbatim; the notes tell you what you can change and what you shouldn't touch.

The core questions

CSAT — the satisfaction question

Ask this after a single resolved interaction, not about the company at large.

How satisfied were you with the support you received? Very dissatisfied · Dissatisfied · Neither satisfied nor dissatisfied · Satisfied · Very satisfied

• Scale: a 5-point labelled scale. Label every point, not just the ends — unlabelled middle points get skipped or guessed at.
• Score: CSAT % = (top-2-box responses ÷ total responses) × 100. Decide top-2 vs top-1 once, write it down, and never quietly change it.
• Anchor it to the interaction ("the support you received"), not to the product, the price, or "us" — otherwise you're measuring brand sentiment and filing it under support quality.

Wording to avoid:

• "How happy were you with our amazing support team?" — leading; "amazing" primes the answer before it's asked.
• "Did we resolve your issue quickly and professionally?" — double-barrelled (speed and professionalism); one yes/no can't answer two things.

CES — the effort question

The single best predictor of whether they'll come back. Ask it in the customer's frame — effort, not delight.

How much do you agree with the following statement? "[Company] made it easy for me to handle my issue." Strongly disagree · Disagree · Somewhat disagree · Neither · Somewhat agree · Agree · Strongly agree

• Scale: the 7-point agree/disagree scale (the CEB/Gartner CES 2.0 standard). Keep it at 7 points if you ever want to compare against published CES benchmarks.
• Score: CES = the average numeric response (1–7), or the % who agree (top-3-box). Higher = less effort = better.
• Say "handle" or "get help with" rather than "resolve" — a customer whose issue is still open can still tell you whether the process was easy or painful.

Trap: don't flip the polarity mid-survey. "How much effort did you personally have to put in?" scores backwards on a 1–7 scale (high = bad). Pick one direction, label the scale, and don't make the customer work out which end is the good one.

NPS — optional, and never per-ticket

Only add NPS if someone above you genuinely acts on it. It measures loyalty to the brand, not the quality of one support interaction — firing it after every ticket irritates people and pollutes the number.

How likely are you to recommend [Company] to a friend or colleague? 0 — Not at all likely … 10 — Extremely likely

• Scale: 0–10, label only the two ends.
• Score: NPS = %Promoters (9–10) − %Detractors (0–6). Passives (7–8) count toward the base but not the score.
• Cadence: a relationship survey — quarterly or twice a year, to a sampled slice of customers — not a post-ticket trigger.

Follow-up & free-text prompts

One open-ended follow-up, shown conditionally on the score, earns you more than three generic boxes. Branch it on the rating:

More prompts to pull from — use at most two, and keep every box optional:

• "Was your issue actually resolved?" — Yes / No / Partly. A cheap reopen/FCR signal that the star rating hides.
• "Is there anything you expected us to do that we didn't?"
• "If you contacted us more than once about this, what made you come back?"
• "Anything you'd change about how we communicated?" (tone, clarity, updates along the way)
• "In a word or two, how did dealing with us leave you feeling?"

A required comment field is the fastest way to kill your completion rate. Every free-text prompt stays optional.

Wording & design traps

• Every question asks exactly one thing — no "quick and friendly", no "easy and fast".
• No adjectives that pre-load the answer ("amazing", "great", "world-class", "friendly").
• The subject is this interaction, not the brand — except NPS, which is deliberately about the brand.
• One scale direction throughout: the "good" end is always the same side, and every point is labelled.
• Scale points match each metric's standard — CSAT 5-point, CES 7-point, NPS 0–10. Don't invent a 4-point CSAT to "force a decision"; you lose comparability against every benchmark.
• A labelled neutral midpoint exists — don't manufacture a fake positive by deleting it.
• No "rate [Agent name]" framing if it makes customers soften scores to protect a person.
• Order: the easy scored question first, the optional free-text last.

Timing traps

• Trigger on resolution, not on ticket close — a bot auto-closing a stale ticket hasn't resolved anything, and surveying it drags your score down.
• Wait long enough that the customer knows whether the fix held: same-hour for a simple answer, next-day for anything that needed a change to take effect.
• Don't survey the same person for every ticket in a burst. Cap it (e.g. once per customer per 7 days) or you train them to ignore you.
• Match the channel to the interaction: email ticket → email survey, chat → in-widget, phone → post-call IVR or SMS. Cross-channel asks get abandoned.

Response-rate tips

The real cost of a low response rate isn't a smaller sample — it's a bent one. The people who bother to answer skew to the delighted and the furious. Lifting the rate narrows that gap.

• Embed the first question in the email or chat (one-click rating) instead of linking to a separate form. This one change often doubles response.
• Ask one scored question up front; reveal follow-ups only after they've clicked. Perceived length kills completion more than actual length.
• Keep the whole thing under ~30 seconds: one or two scored questions, one optional comment.
• Send from a real human or team address, not "noreply@". A reply you can act on is a feature, not a leak.
• Put the response rate next to the score on every dashboard — a 92% CSAT on 12% response is a different claim than 92% on 55%.
• Occasionally sample your non-responders (a short outbound to a random slice of silent tickets) to check which way your bias leans.

---

Benchmarks to keep nearby: CSAT good = 75–84% (top-2-box), world-class 85%+. CES averages around 5.99/7. Typical post-ticket response rate is 10–20% — treat anything you report against that denominator with appropriate suspicion.