Post-Call Survey Template
Send this post-call survey right after a phone interaction to track CSAT or CES (and optionally NPS), then pinpoint what drove the score: wait/hold time, transfers, resolution, and agent behavior. Use the built-in follow-up permission question to route low scores and unresolved issues to service recovery, and turn recurring friction into coaching and process fixes.
Trusted by 5000+ Brands
When to Send a Post-Call Survey (3 best moments)
Immediately after a support call
Do this now: trigger the survey as soon as the agent wraps so you can attach call metadata (queue, hold time, transfers, reason code) and keep recall fresh. Default: SMS or IVR transfer-to-survey within 0-60 minutes; responses stay cleaner when timing is consistent and fast (measurement timing and satisfaction signal).
After an escalation or supervisor callback
Do this now: send a short follow-up once the escalation path is complete so the customer can rate the final outcome, not the handoff. Default: SMS within 0-60 minutes of the callback; use email only if your callback notes require a longer comment. What to watch: unresolved rate and repeat contacts for that issue type.
After outbound service calls
Do this now: survey outbound calls to spot script confusion, verification friction, and process gaps that show up as transfers or long handle times. Default: SMS with a web link within 0-60 minutes; use email when you need customers to reference an invoice/order. What to watch: low CES tied to transfers and time-on-hold bands.
Choose Your Primary Metric: CSAT vs CES vs NPS (Post-Call)
Pick one primary metric for weekly reporting by queue and reason for contact. Decide whether you are trying to measure (1) how the call went (CSAT) or (2) how hard the call was for the customer (CES). If you add NPS, treat it as a separate relationship signal, not a per-call scorecard.
| Metric | Best for | Avoid when | Recommended question stem | How to trend it (so it drives action) |
|---|---|---|---|---|
| CSAT | Interaction quality on this call (agent professionalism, clarity, outcome). | Finding specific friction drivers (holds, transfers, repeat explanations) as your primary goal. | "Overall, how satisfied were you with the support you received today?" (5-point satisfied scale) | Trend weekly by queue and reason for contact; add cuts by hold-time band and transfers to find where satisfaction drops. |
| CES | Friction and effort (waiting, repeating info, bouncing between agents, complex IVR). | Comparing teams that handle very different call types without also cutting by reason/complexity. | "How easy was it to get your issue resolved today?" (5- or 7-point ease scale) | Trend by queue and issue type; overlay operational drivers (hold time, transfer count, repeat contact) to target process fixes (see CES background: Harvard Business Review on Customer Effort). |
| NPS | Relationship loyalty at a higher level (brand/account health), not a single call. | Do not use as a per-call KPI; individual calls are noisy and can mislead coaching decisions. NPS also does not reliably track short-term revenue change at the call level (longitudinal NPS vs growth evidence). | "How likely are you to recommend [Company] to a friend or colleague?" (0-10) (NPS origin: Reichheld, HBR) | Run separately (monthly/quarterly) and trend by customer segment and product tier; keep post-call dashboards anchored on CSAT/CES for queue management. |
If you can only track one: default to CSAT (simple readout) or CES (best for friction reduction). Reserve NPS for a periodic relationship program aligned to broader customer satisfaction monitoring guidance in ISO 10004 customer satisfaction measurement.
Who Should Take It + Sampling Rules That Avoid Bias
Set eligibility first so your scores reflect real call experiences. Do this now: define the population as customers who completed an inbound call, outbound service call, callback, or escalation and can reasonably rate the interaction.
- Include: completed calls where the customer reached an agent (or completed a meaningful self-service + agent flow) and the issue context is known (queue/intent tags).
- Exclude: wrong numbers, abandoned calls (unless you are explicitly running an abandonment survey), test calls, and calls where the respondent did not participate (for example, a third-party intermediary) unless they are the decision-maker for that account.
- Control duplicates: if the same customer calls back multiple times in a day, pick the last call or the longest call, but apply the rule consistently.
Sampling comes next when volume is high. Default: sample randomly, or stratify by queue and time band (peak vs off-peak) so you can compare hold time, transfers, and AHT conditions more fairly. If you need a refresher on setup options, use sampling methods for call surveys to choose between random and stratified approaches.
Do this now: compare survey completes vs total calls by queue, day, and hour. If you only survey the "happy path" (business hours, short waits, certain queues), your CSAT/CES will drift upward and you will miss the real friction - use how to avoid response bias as your checklist.
Watch nonresponse patterns by segment (for example, long-hold callers ignoring SMS). Nonresponse rates do not automatically mean bias, so validate by comparing available call metadata for respondents vs non-respondents (nonresponse rate vs bias meta-analysis) and by using consistent field definitions (AAPOR guidance: Standard Definitions).
Owner tip: have your WFM or QA team set minimum weekly completes for priority queues (billing, cancellations, escalations) so your dashboard does not get dominated by easy, high-volume calls.
Customization Tips (Keep It Short, Branch Smart, Stay Comparable)
- Keep it to 3-7 questions: Default to 4-5 (resolved, CSAT or CES, one driver question, open text, follow-up permission). If your IVR drop-off is high, cut to 3-4 and move verbatims to SMS/email.
- Cut demographics first: Remove "nice-to-have" items (age, income, long attribute grids) before you touch your core outcome (CSAT/CES) and resolution question. Your QA and ops teams can usually join needed segments from CRM/call logs.
- Branch on resolved vs unresolved: If resolved, ask one driver (for example, "The agent understood my issue") then take verbatim. If unresolved, route to a short recovery path (what is still needed, preferred contact method, permission to follow up).
- Add an escalation/complaint path: When transfers >= 2 or the call was tagged "escalation," ask a targeted driver like "I was transferred too many times" (agree/disagree) so you can fix routing and knowledge coverage.
- Stay consistent on scales: Pick one CSAT scale and one CES scale and keep them stable so your trendline is real. If you are changing formats (stars vs 1-5 vs agree/disagree), standardize using Likert scale and rating scale options before you compare weeks.
- Tailor wording to inbound vs outbound: Inbound: focus on "getting help" and "issue resolved." Outbound service: add clarity and professionalism ("The reason for the call was explained clearly"). Outbound sales: keep it compliant and avoid pressure-language questions that create defensiveness.
- Protect privacy in open text: Add a note like "Please do not include account numbers or sensitive info." Then use a permission-to-follow-up item (yes/no + preferred channel) instead of asking for personal data in the comment box - follow privacy and data minimization basics and, where applicable, the data-minimization principle in GDPR (Regulation (EU) 2016/679).
Scoring, Reporting, and Close-the-Loop Playbook (Weekly)
- 1) Score the metric correctly and avoid small-sample whiplash
Lock your formulas and show them on the dashboard so team leads do not argue about math. Default scoring: CSAT = % top-box (for example, % 4-5 on a 5-point scale), CES = average score or % "easy," NPS = % promoters (9-10) minus % detractors (0-6).
- Cadence: report weekly, plus a 4-week rolling view for smaller queues.
- Minimums: if a queue gets low completes, roll it up biweekly/monthly and stop reacting to single-week swings.
- Uncertainty: include a confidence band (especially for top-box CSAT) so leaders see when movement is mostly noise (primer: NIST on confidence intervals for proportions).
- 2) Cut results by the operational drivers you can actually change
Add these standard cuts to every report so you can tie scores to call-center reality:
- Primary cuts: queue, reason for contact, and time band (peak vs off-peak).
- Friction cuts: hold-time band, transfer count, and repeat-contact flag.
- People/process cuts: agent tenure (new vs tenured) and escalation tag.
Diagnostic note: if a queue drops when holds exceed your internal target (for example, 5+ minutes as a starter threshold you can tune), treat it as a staffing/routing issue first, not just an agent issue.
- 3) Triangulate with KPIs without blaming agents for system problems
Connect survey records to AHT, repeat contacts, and escalation tags, then review patterns with Ops + QA together.
- If you see low CES + high transfers, fix routing and ownership (who should handle what) before you coach tone.
- If you see low CSAT + high unresolved, tighten next-step promises, improve knowledge coverage, and revisit escalation rules.
- If you see good scores + high repeat contacts, check for "polite but ineffective" interactions (resolution definition, policy limits, or missing tools).
- 4) Set alerts and route service recovery with clear SLAs
Create auto-alerts for at-risk responses and route them consistently so customers get help and leaders get root causes.
- Trigger: low score, unresolved answer, or negative verbatims that mention "rude," "cancel," or "supervisor."
- Routing: send to a service-recovery queue with a same-day SLA (for example, 4-24 hours depending on severity).
- Recovery: confirm resolution, apologize once (no script wall), and log a root-cause tag (hold time, transfers, policy, knowledge, tooling) for trendable fixes.
Why it works: service recovery quality is strongly associated with satisfaction and loyalty outcomes (see foundational work on complaint handling: Tax, Brown, and Chandrashekaran (1998)).
- 5) Turn themes into coaching and process fixes
Run a weekly 30-minute review where QA pulls 5-10 low-score calls per top driver (holds, transfers, unclear next steps) and produces concrete outputs.
- Coaching: 1 focus behavior for team leads (e.g., expectation-setting or ownership language).
- Content: 1 script/knowledge update tied to the most common failure mode.
- Operations: 1 IVR/routing/tooling ticket when friction repeats across agents.
- Measurement: keep a change log so you can see if fixes move CSAT/CES in the affected queue over the next 2-4 weeks.
Frequently Asked Questions
How soon should I send a post-call survey?
Send it immediately after the call when you can, and aim for 0-60 minutes for SMS or IVR. Use end-of-day email only for lower-risk interactions (for example, routine billing follow-ups) where you expect slower but more detailed comments. Keep timing consistent week to week so your queue trends do not get distorted.
Which channel works best: SMS, IVR transfer-to-survey, or email?
Default to SMS for speed and reach, especially when you want clean weekly trends by queue. Use IVR transfer-to-survey when immediacy matters and you can tolerate more drop-off; keep it very short. Use email when you need longer verbatims, but expect slower responses and more bias toward highly engaged customers.
How many questions should a post-call survey have?
Keep it to 3-7 questions, with 4-5 as the default for most support queues. A minimum viable set is: resolved today (yes/no), CSAT or CES, an open-text comment, and permission to follow up. If completion drops, cut demographics and extra attribute grids before you cut your core metric.
Should we use NPS on a post-call survey?
Use NPS sparingly on post-call surveys because it is a relationship metric and gets noisy at the individual call level. If you include it, keep CSAT or CES as the primary post-call KPI for queue management and coaching. Run NPS as a separate monthly/quarterly program if your main goal is tracking loyalty.
How do I decide how many responses I need each week?
Plan around the cuts you want to act on (queue, reason for contact, time band), not just an overall total. Aim for enough completes in your priority queues and new-agent cohorts so you are not reacting to a handful of surveys. If a queue is small, roll it up to biweekly/monthly and use a rolling view.
What should happen when someone gives a low score or says the issue is unresolved?
Auto-tag low scores and unresolved answers, then route them to a service recovery queue with a clear follow-up SLA. Follow up, confirm the outcome, and log a root cause (hold time, transfers, knowledge gap, tooling) so the same issue does not repeat. Feed the top themes into coaching and routing/IVR fixes every week.
Related Survey Templates
FREE TO START -- NO CREDIT CARD REQUIRED