A post-call survey asks for feedback while the interaction is still fresh — the caller rates the call on the keypad seconds after it ends. On Devotel Orbit this runs as a self-contained IVR step inside the call itself: CSAT on a 1–5 scale, NPS on a 0–10 scale. This explainer walks the shipped surface — the per-survey configuration, the gates that decide which calls get surveyed, the analytics the answers feed, and how it sits next to the broader voice-of-customer program.
Why the cleanest CSAT/NPS signal in voice still comes from the call
A survey sent an hour later measures a memory. A post-call IVR measures the interaction: the caller is still on the line, the agent hang-up is the trigger, and the answer is one digit — not a link, a login, or a form. Response quality differs accordingly. Digit capture happens inside the same call the score describes, so there is no attribution doubt about which interaction a 4 belongs to, and no send step for a message provider to drop. Where the channel itself is the survey instrument, nothing stands between "the call ended" and "the score is recorded."
That is the complement to the SMS and WhatsApp survey program covered in Voice-of-Customer Surveys over SMS and WhatsApp: those carry richer, multi-question surveys to a defined audience after the fact; the post-call IVR carries the one question that needs to be asked immediately.
The config surface, end to end
Authoring a survey lives under Insights → Surveys; its post-call voice behaviour is configured under Voice → Post-call CSAT or Voice → Post-call NPS. Pick the survey at the top of the page and two panels appear — analytics for that survey, and the post-call IVR configuration for it. The NPS page mirrors the CSAT page field for field; only the scale differs.
A first-run checklist for the configuration panel:
- Enabled — the master switch. Surveys ship off and stay off until you turn one on; nothing plays for any caller while it is off.
- Delivery —
Post-call IVRplays the survey inside the ended call;Messagingfalls back to sending it over your messaging campaign instead, in which case none of the voice knobs below apply. - Survey prompt — the text read to the caller before they press a digit (up to 500 characters).
- Thank-you prompt — played after a valid digit lands.
- Invalid-digit prompt — played when the caller presses a key outside the scale; leave it blank to skip the retry prompt.
- Prompt audio URL — a pre-recorded HTTPS prompt that replaces the spoken text when set.
- Digit timeout (ms) — how long the survey waits for a keypress.
- Max attempts — how many digit tries the caller gets before the call ends normally.
- Sample rate (%) — survey only a percentage of qualifying calls.
Every field is editable inline, and Save writes the whole configuration in one update.
Gates: which calls actually get surveyed
Raw "survey everything" produces noise; the shipped gates exist so the survey only fires where a score means something. All are tenant-owned:
- Only after a human-answered call — skip when answering-machine or fax tone was detected.
- Minimum call duration — skip calls shorter than the threshold, so misdials stay out of the average.
- Only these dispositions — an optional allowlist of wrap-up/outcome codes (for example
resolved,transferred); the survey fires only for those outcomes. - Never these dispositions — an optional denylist that suppresses the survey even when the allowlist above would admit it; the denylist wins.
- Sample rate — the percentage gate is deterministic per call, so a retried call-end event never double-plays the survey.
A call that passes none of the gates simply ends — the caller hears nothing extra.
Where the answers land
The same page carries the analytics panel above the configuration: switch between 7, 30, 90 days, or all time, and read the response count, average score, satisfaction percentage (CSAT) or the promoters-minus-detractors NPS computation, and a per-score breakdown bar. Before responses exist, the panel shows a plain empty-state line; if the load fails, it renders a retry card instead of a dead skeleton.
Beyond the per-survey panel, responses attribute to the queue and the handling agent. The queue agent leaderboard joins responses to calls, so a call's survey counts exactly once; agents with calls but no answers stay listed with a null average rather than dropping out. The scorecards roll up CSAT and NPS together, grouped by queue or by agent, and scored responses also feed the QA loop — a detractor answer with a known agent can queue a QA evaluation against that interaction.
Fitting into the broader VoC program
Post-call IVR is one of three survey patterns on Orbit, each owning a different capture moment: the in-call digit capture covered here, the audience-targeted SMS/WhatsApp/email distribution in the VoC surveys explainer, and the close-triggered auto-CSAT/NPS dispatch on conversation close. They share one analytics and trait-writeback substrate, so a score captured on a keypad at call-end and a score tapped on a phone screen an hour later both land on the same customer profile — and the cross-channel fatigue controls still apply, so enabling post-call IVR does not multiply the asks a contact receives.
The honest posture on defaults
Defaults ship off. No caller hears a survey until a tenant enables one, picks the survey it belongs to, and saves the configuration — the two-minute checklist above is the entire activation path. Every gate — sampling, duration, dispositions, human-answer detection — is a tenant-owned control, and Orbit never imposes a platform-mandated survey cadence. Turn it on because you want the signal; keep it scoped because you want the score to mean something.