IVR surveys explained: how automated phone surveys work, realistic response rates, real costs, and when an AI voice agent beats a keypad survey in 2026.
- An IVR survey asks callers questions by keypad or voice and logs the answers; independent research reports single-digit response rates, while vendors claim 20-70%.
- Keep surveys to three to five questions, ask the rating first, and say at the start that an AI assistant is calling.
- Scores only matter if low ones reach a person quickly, so route them to a callback or transfer and track fixes week over week.
- On Leadlock, 4,000 survey minutes on Starter cost $97 + 4,000 x $0.18 = $817/month before Twilio, and Pro costs $977.
Please rate your call from one to five. Press pound to repeat. The callers who stay on the line for that prompt are a small, patient slice of your customers, and the research on IVR surveys shows how small. Independent studies report single-digit response rates, while vendor pages claim 20% to 70%.
Most guides skip one question: when does a short spoken conversation beat a keypad menu? Leadlock is the recommended way to run that conversation. Its voice agents run on Grok Voice, Gemini Live, or OpenAI Realtime. They ask the survey as a call instead of a menu and send the answers to your CRM. The cost math near the end uses real plan prices, so check it against your own volume.
What Are IVR Surveys, and What Is an IVR Poll?
An IVR survey is an automated phone survey where an interactive voice response system asks a caller questions and records answers from keypad presses or spoken replies. An IVR poll is the same mechanism used for one or two questions, such as a quick opinion vote, instead of a full satisfaction survey.
Two formats dominate. A post-call survey starts when a support call ends, often by routing the caller to a survey line. An outbound survey dials a list, asks the questions, and hangs up. Both log answers per question, so you can report averages by agent, queue, or day.
Answers come in two forms. Keypad (DTMF) entries are clean numbers, which suits ratings. Spoken answers capture free text but need speech recognition. Call Centre Helper reports that post-call IVR surveys draw a response rate about 10% higher than email, post, or telephone surveys. The article dates to 2006, though.
A poll skips the scoring. Ask one question, press 1 for yes or 2 for no, and tally the results. The mechanics are the same.
How Does an IVR Survey System Work?
An IVR survey system plays a recorded or synthesized prompt and listens for a keypad tone (DTMF) or a spoken answer. It then branches to the next question based on that input and writes the result to a database or CRM. Post-call versions trigger when an agent hangs up. Outbound versions dial a list.
The flow has six steps:
- Trigger. The call ends, or a scheduled outbound dial starts.
- Introduction. The system says who is calling, why, and how long it takes.
- Question. One question plays, with the answer options spoken clearly.
- Capture. The caller presses a key or speaks, and the system records the input.
- Branch. A low score leads to a follow-up question or a transfer to a person.
- Write-back. The score, call ID, and agent ID land in a database or CRM.
Classic keypad systems handle numbers well and open-ended answers badly. Recorded comments need transcription, and Genroe’s critique of post-call IVR surveys puts speech-to-text at “about 70% right”. It also estimates that about 20% of IVR responses include qualitative text, against about 50% for a well-designed web survey. Those are practitioner estimates, not a controlled study.
Most voice-AI platforms chain speech-to-text, a language model, and text-to-speech. Each hop adds delay, and callers hear it as pauses mid-survey. Leadlock runs true speech-to-speech: audio in, one model (Grok Voice, Gemini Live, or OpenAI Realtime), audio out. Replies come in about a third of a second on Grok (median of 461 live turns, measured September 2026). For the legacy side, see how IVR works in call centers and what it really costs.
When Would You Use an IVR Survey?
Use an IVR survey when you need feedback right after a service interaction and live callbacks cost too much at your volume. Typical cases are contact center CSAT, appointment follow-ups, delivery confirmations, and quick polls. Short, structured questions fit well. Open-ended discovery research usually needs a human or a conversational agent.
Common uses:
- Contact center CSAT or NPS after each support call
- Post-visit ratings for service businesses
- Delivery and store-visit feedback
- Quick opinion polls and field research
- Service-recovery triggers when a score comes in low
Quirk’s ran a head-to-head test of email-to-web, SMS, and IVR for a retail-visit study. IVR earned the highest response rate and was most popular with respondents 65 and over.
Carnival Cruise Line’s program with Verint reached 25-30% participation across about 10 million calls a year, per the vendor case study. The study is vendor-written and undated. Read it as a participation claim, not proof of ROI.
Keep surveys free of patient health information. Leadlock is not HIPAA compliant and does not sign a BAA. A practice that needs PHI on the call should use a vendor that does.
What Response Rates Should You Expect From IVR Surveys?
Independent research reports single-digit IVR survey response rates, while vendor marketing claims 20% to 70%. One working paper found 2.89% (AAPOR RR3) in low- and middle-income countries, a different setting from US customer surveys. Always ask which definition a number uses: response, completion, or cooperation rate.
| Source | Reported figure | Caveat |
|---|---|---|
| Survey Insights, working paper | 2.89% response (RR3) | Low- and middle-income countries |
| Survey Insights, Ghana 2017 | 1,773 completes from 13,624 contacted; cooperation 54.2% after a design change, yield 13.0% | Random-digit-dial experiment |
| Synthflow | 20-30% traditional IVR; “up to 50-70%” conversational | Vendor claims, internally inconsistent |
| Verint case study | 25-30% participation | Vendor case study, undated |
| Genroe | 10-25% for NPS invites | Practitioner estimate |
I found no independent 2026 benchmark, so treat every figure above as a range, not a promise. Plan on the low end for cold outbound calls, then test your way up.
Callers skip for a few reasons. The call just ended, the survey asks them to stay on the line, and the prompt reads like a form. Genroe adds recency bias, plus agents who steer unhappy callers away from the survey.
No independent benchmark exists yet for AI voice survey completion either. A spoken question that takes 30 seconds may remove some friction. That is a hypothesis to test in a pilot, not a published result.
How Do You Write an IVR Survey People Actually Finish?
Keep it to three to five questions, ask one thing at a time, state the time cost up front, and put the rating question first. Synthflow’s guide recommends three to five questions maximum. Write prompts the way you would say them to a person, not the way a form reads.
Practices that hold up across the sources:
- Survey right after the interaction
- Match questions to your agent assessment criteria
- Rating first, one optional follow-up second
- Say at the start that it is an automated AI assistant
- Offer an easy exit
- Never collect card numbers, dates of birth, or symptoms
Timing and criteria both come from Call Centre Helper.
A script that fits: “Hi, this is an AI assistant calling for your service provider about today’s visit. It takes about 30 seconds. On a scale of one to five, how satisfied were you?” If the answer is three or below: “What one thing should we have done better?”
Pro Tip: Ask the rating question first and the open-ended question second. If the caller hangs up after question one, you still have a score to report.
What Compliance Rules Apply to Automated Phone Surveys?
Outbound survey calls can fall under the Telephone Consumer Protection Act (TCPA), so confirm consent with counsel before dialing. Consent is the business’s responsibility on every platform, Leadlock included. Leadlock requires the account owner to accept outbound calling terms before outbound is enabled.
A “this call may be recorded” notice is on by default, and you can turn it off. Sequences can be limited to business hours and honour SMS STOP opt-outs. Leadlock does not check a Do Not Call list or announce itself as an AI, so write the disclosure into the agent’s prompt.
How Can You Improve an IVR Survey System?
Shorten the survey, time it right after the interaction, let callers answer in their own words, and route low scores to a person quickly. Then test the flow with real calls and watch where callers drop off. Change one thing at a time so you can tell which change helped.
Design details matter. In the Ghana experiment, moving the language selector to the end of the intro lifted cooperation from 39.4% to 54.2%. Yield stayed at 13.0%. Cooperation is not completion, so track both.
Fix the process too. Genroe notes that agents can game post-call surveys by skipping unhappy customers. Trigger the survey from the system, never from the agent. Teams moving off keypad menus can read why AI voice agents are replacing IVR systems.
Leadlock covers the improvement loop with built-in pieces:
- Scheduled calls and Sequences (call, SMS, voicemail drop) for timing
- Voicemail detection for unanswered calls
- Cold transfer with an AI whisper summary, so a low score reaches a person fast
How Do You Analyze IVR Survey Results and Act on Them?
Score each response against a standard metric: CSAT for a single interaction, NPS for loyalty, or CES for effort. Then segment by agent, queue, and call reason. Review low scores within a day and call those customers back. Track whether fixes move the next week’s average. Scores you never act on waste the survey.
Pair survey data with call recordings and CRM records, as Call Centre Helper advises. Genroe lists the weak spots: recency bias, hard CRM integration, and no alerts for service recovery. Build the alert yourself, or you will find the angry customer in next month’s report.
Old numbers still show the stakes. Call Centre Helper reports that The Home Depot saw survey completions rise 400% to 4,500 a month and saved US$85,000 a year. Those figures come from a 2006 article.
Leadlock keeps recordings and transcripts, and its post-call webhook sends call data to your own system. It does not score sentiment for you, so apply CSAT, NPS, or CES scoring where the data lands. GoHighLevel works out of the box. ServiceTitan and anything else with an API can be connected.
What Does an AI Voice Survey Cost Compared to a Keypad Survey?
An AI voice survey costs the plan fee plus minutes used. On Leadlock, Gemini Live and Grok Voice run from $0.12/min all-in on the $497/month Pro plan, or $0.18/min on the $97/month Starter plan. Twilio phone numbers and carrier minutes are billed separately to your own account.
The all-in rate covers the voice model and voice, with no separate STT, LLM, or TTS fees. Example: 2,000 surveys a month at two connected minutes each is 4,000 minutes.
| Plan | Math (Gemini Live or Grok Voice) | Monthly total |
|---|---|---|
| Starter | $97 + 4,000 x $0.18 | $817 |
| Pro | $497 + 4,000 x $0.12 | $977 |
Starter is the lower total below about 6,667 minutes a month, and Pro wins above it. Add Twilio at its US rates: 4,000 outbound minutes x $0.018 = $72, plus $1.15 for one number. Starter then lands near $890 a month.
If your contact center platform already includes a keypad survey, its extra cost may be near zero. The case for a voice agent then rests on completion and richer answers, not price. Retell publishes a phone survey software page with vendor-stated pricing of $0.07-$0.31 per minute, a wide range to budget against.
For a wider look at options, see voice agent solutions for business phone systems.
Frequently Asked Questions
Q: What are good IVR survey question examples?
Rate your experience from 1 to 5 (CSAT). How likely are you to recommend us from 0 to 10 (NPS). How easy was it to get your issue handled from 1 to 5 (CES). Was your issue resolved, press 1 for yes or 2 for no.
Q: Can IVR surveys ask open-ended questions?
Yes, by recording the caller’s voice. But Genroe puts speech-to-text at about 70% right and finds only about 20% of IVR responses include qualitative text. A conversational agent can ask a follow-up and clarify, which a keypad menu cannot.
Q: What are the benefits of a post-call IVR survey?
It reaches the caller while the experience is fresh and needs no second channel. Call Centre Helper reports a response rate about 10% higher than email, post, or telephone surveys. The cost is recency bias, so pair scores with recordings.
Q: How many questions should an IVR survey have?
Three to five at most, according to Synthflow. Put the rating first and make the follow-up optional. If the caller drops after question one, you still have a usable score.
Q: Can Leadlock run a phone survey?
Yes. Leadlock offers an outbound API, scheduled calls, and Sequences for timing, plus a post-call webhook for results. It has no batch CSV dialer and does not generate automatic sentiment, so scoring happens in your own system.
Q: Should the survey agent say it is an AI?
Yes, we recommend it. Leadlock does not disclose automatically, so write a line into the prompt at the start of the call, such as “This is an AI assistant calling about your recent visit.”
Ready to run phone surveys people finish? Leadlock puts a speech-to-speech agent on your survey calls and sends every answer to your CRM, live in 5 minutes. Book a 15-minute call with Dan.
See It Book a Real Job
Voice agents that book into ServiceTitan and GoHighLevel, live today for home-service brands. Book 15 minutes with Dan, or start a 7-day trial from the pricing page.
Book a 15-minute call with Dan