IVR Testing Checklist: What to Check Before Your Phone Menu Goes Live
A practical checklist for testing an IVR before launch: paths, prompts, hours and holidays, errors, load, devices, accessibility and how to sign off.
An IVR that fails on its first day is expensive in the worst way: customers meet the failure before you do. Most faults are not exotic; they are a wrong key that goes nowhere, a prompt recorded for the wrong team, or a holiday setting that points to an empty queue. A methodical test finds them. This checklist gives you a structure you can reuse for the first launch and every change after it. If you have a call-flow diagram, Mapping Your IVR as a Flow explains how to make one, and it becomes your test plan.

Start with a test plan
Write down what you will test before you pick up the phone. A plan from the flow diagram has one row per path, with columns for the action, the expected result, the actual result and a pass or fail.
Building the plan
- List the pathsEvery route from the greeting to an end state: queue, voicemail, callback, hang-up.
- List the conditionsOpen, closed, holiday, emergency, queue full, no agents.
- List the callersDifferent devices, networks, languages and abilities.
- Assign testersAt least one person who did not build the menu.
- Set pass criteriaThe exact wording, destination and timing you expect.
Functional checks: does every path work?
Paths and routing
- Each menu key reaches the intended queue or message
- Keys that are not in the menu are handled and repeat the options once
- Silence is handled: a repeat, then a fallback to a person or voicemail
- The way out to a person works from every level
- Transfers reach the right team and do not drop the call
- Voicemail records, saves and notifies the right people
- Callback requests are captured with the number and the reason
- Queue overflow and no-answer rules send calls where you expect
- The call ends cleanly after the last prompt with no dead air
Content checks: do the prompts say the right thing?
Functional tests tell you the call went to the right place. Content tests tell you the caller understood. Listen to each prompt as a stranger would.
- Is every sentence accurate today: hours, addresses, names of teams, available services?
- Is the wording consistent from prompt to prompt, with the same name for the same thing?
- Is the volume steady, with no clipping, pops or noise at the start and end?
- Is the pacing comfortable, with short pauses between options?
- If you offer several languages, has a native speaker checked each one?
- Does anything repeat, contradict or overpromise?
Time-based checks: hours, holidays and emergencies
The branch tested least is the one most likely to fail. Closed hours, lunch breaks, holidays and emergency messages are all rarely used and easy to break. Test each of them, and do it with real settings rather than only reading the configuration.
Time conditions to test
| What to set up | What to confirm | |
|---|---|---|
| Normal open hours | A call at an ordinary time | Menu plays, calls reach agents |
| Closed hours | A call just after closing and just before opening | Closed message and voicemail or callback |
| Lunch or break | A call inside a break window, if you have one | The intended message or route |
| Holiday | A holiday entry for the day | Closed route or special message |
| Emergency | The emergency message switched on, then lifted | All calls hear it, then normal service returns |
| Time zone | Hours read in the right zone | Open and closed times match your location |
Emergency and Outage Messages in Your IVR and Routing Calls by Business Hours, Holidays and Closures describe how to build these branches in the first place.
Error and load checks
- Wrong inputs: press every key, including star and hash, and press keys during prompts.
- Interruptions: speak or press early, mid-sentence, to see whether the menu copes.
- Busy queues: fill a queue with test calls to check waiting messages, position announcements and overflow.
- No agents: leave every agent unavailable and confirm callers reach voicemail or a callback request.
- Peak volume: place several simultaneous calls and confirm lines, channels and agents hold up.
- Dropped calls: hang up at each stage and check nothing is left in a queue or reported as answered.
Different callers and devices
A menu that works from your desk phone may not work from a mobile phone in a car. Test on landlines, mobiles on different networks, a browser phone and, if you serve them, international callers. Include testers who use screen readers, relay or assistive services, and testers with hearing difficulties, since they will reveal timing and wording problems quickly. Accessible Phone Menus for Every Caller explains what to look for.
Sign-off and ongoing testing
When the tests are complete, record the results, who ran them and the date. A named person signs off the release, and a short list of known issues, if any, is accepted explicitly. Do not launch with unknown failures.
Keeping a useful test log
A test log turns a vague sense that things work into evidence. Keep it simple enough that people will actually fill it in: a spreadsheet or shared document with the date, tester, device, path, expected result, actual result and a note. Record failures with enough detail to repeat them, including the time of the call and what was pressed or said.
When a fault is fixed, retest the same path and mark it with the date. Over time the log shows which parts of the menu break most often, which is where to spend extra care. It also helps new team members understand what the menu is supposed to do.
A word on audio quality
Listen to the whole menu on the devices your callers use, not only through the speakers of the computer that recorded it. Compressed phone audio can make a recording that sounded fine in the studio sound muffled or harsh. Pay attention to the first and last half-second of each prompt: clipped starts and abrupt endings are the most common faults, and they make the whole system feel careless.
If prompts are generated by text-to-speech, listen for mispronounced names, street names and abbreviations, and fix them with a pronunciation entry or spelling change. See Telephony Audio Quality for what affects the sound once a call is connected.
Watch the first week
Even a thorough test cannot imitate every caller. For the first week, listen to a sample of real calls each day, watch abandonment and misroutes in your reports, and ask agents what callers say about the menu. Keep a short list of quick fixes and ship them together. IVR Analytics explains which measures show trouble early.
After launch, test again whenever something changes: hours, teams, prompts, numbers or phone provider. A smaller regression test with the main paths and the closed-hours branch takes ten minutes and catches most damage. Schedule it monthly, and before every holiday period. See Updating Your IVR for Seasons and Campaigns for a yearly routine.
Testing in Call
Call offers several ways to check your setup without disturbing customers. The routing simulator walks a pretend call through the hours, holidays, capacity and skill rules for a queue, shows each rule the call passed and where it ended, and does not dial anything. A test call feature lets you ring a line from its own number to hear exactly what callers hear, and practice calls and test buttons in settings help check AI agents. Use the simulator for the time-based branches and real calls for audio, wording and transfers. For the setup itself, see Setting Up IVR and Business Hours in Call, and for AI-driven lines read Testing Voice AI Agents.
Frequently asked questions
How long does it take to test an IVR properly?
For a small menu, a few hours; for a large one with several teams and languages, days. Preparation from your flow diagram makes it faster.
Who should test the IVR?
The builder, to catch technical faults, and at least one person who has never seen it, to catch confusing wording. Include people with different phones and abilities.
How often should I retest?
After every change, and with a short routine monthly. Always retest before holidays, campaigns and any change of phone provider.
Can I test without calling real customers?
Yes. Use internal test numbers, a simulator if your system has one, and colleagues acting as callers. Avoid testing against live queues during busy periods.
What are the most common IVR faults?
Dead ends, wrong destinations, stale hours and holiday settings, missing handling for silence and wrong keys, and prompts that no longer match the options.
Do I need to test after hours?
Yes. Closed-hours behaviour is rarely seen by staff, so make a test call outside opening hours and on a day set as a holiday.