Skip to main content
Non-conversational Voice Bot Testing

What is a non-coversational Voice Bot, and how to test it?

Nikola Jonic avatar
Written by Nikola Jonic
Updated yesterday

A non-conversational voice bot is a type of automated system you interact with over the phone. Unlike more advanced voicebots that can have complex conversations, a non-conversational voicebot keeps things simple.

When you call, it listens to what you say, recognizes specific words or phrases, and then responds with a pre-recorded message that matches what you're asking about. For example, if you ask about business hours, it might say, "Our business hours are 9 AM to 5 PM."

There are no menus to navigate, like "Press 1 for this" or "Press 2 for that." Instead, you just speak, and the bot responds directly.

This type of system is great for quickly getting straightforward information without needing to go through a lot of steps.

Before starting the voice bot test, some of the best practices are:

  • Review the test scenarios, questions in the assessment report and further documentation provided by the customer (e.g. flowcharts, expected behaviour, test data) before initiating a call and make sure that everything is clear to you.

  • Record a timestamp of the interaction start (once the voicebot picks up the phone) - this will help our customer to identify the phone call internally.

  • Make a recording of the call. You can then use the recording in order to replay and better analyze the interaction. This will help you to fill out the assessment report and even identify more issues.

Testing the Voice Bot

When it comes to deploying a non-conversational voice bot, thorough testing is crucial to ensure it performs accurately and efficiently. Below is a step-by-step approach you can use to test your voicebot once it answers a call:

Note: You should always pay attention and read carefully the test instructions. In some cases e.g. Error Handling is out of scope, or we need to test only specific keywords that are provided by the customer.

  1. Initial Greeting

    • Observe: Listen to the initial greeting. Make sure it’s clear and welcoming.

    • Check for Clarity: Confirm that the greeting message is easy to understand.

  2. Response Accuracy

    • Provide a Query: Ask a straightforward question that the bot should recognize. For example, "What are your business hours?"

    • Verify Response: Ensure the bot responds with the correct pre-recorded message.

  3. Keyword Recognition

    • Test Different Phrases: Use different phrases that mean the same thing to test the bot’s ability to recognize variations. For example, "What time do you open?" vs. "When do you start business?"

    • Evaluate Consistency: Check if the bot provides accurate responses for all variations.

  4. Error Handling

    • Introduce Variability: Say something the bot might not recognize or is outside its scope, like "Tell me a joke."

    • Response Evaluation: Assess how the bot handles the situation. It should ideally respond with a polite message indicating it didn’t understand or provide alternative guidance.

  5. Multiple Queries

    • Ask Sequential Questions: Test how the bot handles multiple queries in a single session. For example, "What are your business hours?" followed by "Where are you located?"

    • Check Transition: Ensure the bot smoothly transitions from one query to the next without confusion.

  6. Accent and Dialect Testing

    • Test Different Accents: Use different accents or dialects to see how well the bot understands.

    • Evaluate Understanding: Ensure the bot can recognize and correctly respond to queries from various accents.

  7. Background Noise

    • Simulate Real Environment: Conduct tests with background noise to assess the bot’s performance in less-than-ideal conditions.

    • Check Performance: Make sure the bot can still accurately recognize and respond to queries.

Did this answer your question?