IVR Stress Test

There are many sip testing apps out there, but I’m researching ways to stress test IVR functionality. I’ve got a pretty basic IVR demo setup and need an automated stress test. IVR flow is as follows:

  1. Caller calls #.
  2. ‘enter your acct #’ message is heard.
  3. acct # is read via dtmf.
  4. db dip to validate acct #.
    5a) if acct # is valid, then…
    ‘please hold for the next avail rep’ msg is heard.
    5b) if acct # is not valid, then… ‘sorry we can’t help you, please try again’ msg is heard

So for the purposes of testing, I can replace the audio messages with dtmf responses, but I’d prefer to keep the test IVR as is. How to interpret the audio success/failure messages? The only thing I can think of is to “train” the test app by making a success call and a failure call (by intentially entering a bad acct #), recording the 2 audio files, and then comparing the audio files on subsequent calls. However, that seems like a pretty hokey way to do it. If there was a solid way to speech to text the message and then do a key phrase text compare, then maybe it would work.

Any ideas? Anyone ever something similar?


It is not clear to me what you want to test with the stress test of the ivr. For a proper reaction you have to be more specific. Asterisk installed on a proper server will be capable of handling lots of incoming calls routed into an ivr and it is not the ivr that set the constraints.

I apologize for being unclear.

We have a 3rd party call center system (ivr, dialer, acd). I’ve been looking into a platform to test the ivr functionality. There are plenty of sip/h323 apps out there that seem to just test the call control signalling. What I’d like to do is actually programmatically step through the ivr, simulating an actual caller – sending dtmf to simulate selecting ivr menu options.

Of course, I can use call files or an agi script to generate calls, but just generating a bazillion calls is not enough for me. I actually want to intelligently interpret the ivr responses to my dtmf events so that I can branch to the appropriate next step.

Does that make more sense?

What methods are available for interpreting an audio message from the ivr system, and branching to 1 or more dtmf responses based on the content of that message? With the end result being that I can stress test the ivr system by simulating real callers?