Maximize productivity with personalized, automated calling

Voice Self-Service for your growing business needs


What is an IVR?

IVR stands for Interactive Voice Response for telephony applications.   It allows you to present audio prompts to callers, and gather their input via touchtone (DTMF – Dual Tone Multi Frequency), or by responding with their voice.

Capture Input

You can gather input from callers in three different ways:

  • DTMF – Callers are guided to press digits (“press 1 for sales”)
  • Press or Say – (“press or say 1 for sales”)
  • Speech – (say “Sales”)

There are pros and cons associated with each method. Digits are simple, but may require long menus.   Speech can frequently resolve a caller’s request quickly, but may require more prompt guidance and error recovery techniques to keep callers on track and engaged.

In many cases it makes sense to mix modes such as using speech to guide caller through a menu, then asking for DTMF input for something such as a credit card number to enhance security and accuracy.

Guidance and Dynamic Data Presentation

Prompts make up the audio information you present to guide callers through a menu, and to respond back as part of a dialog.   Alternatives include:

  • Pre-recorded audio – These are the audio clips recorded by the persona you are presenting.   Prompts are scripted and recorded to match your outline dialog.
  • Concatenated audio – Uses pre-recorded audio for segments of information that can be dynamically linked together such as digits, letters and dates.
  • Text To Speech (TTS) – TTS converts written content to audio output.   TTS is frequently used to present dynamic information such as an account balance.

Pre-recorded audio and TTS can also be mixed depending upon application requirements. A primary benefit is injection of dynamic data into a conversation with a caller. Somewhat of a downside is that the TTS portion will not match your voice talent.

Inbound or Outbound

Automating inbound calls is what most frequently comes to mind with an IVR.   Typical application include call routing, presenting information such as hours and location, or using IVR to get a caller to the right department or agent (all routing).

IVR is also an outstanding solution for proactively reaching out to customers or prospects. You can define a calling list and automatically dial out to callers with built in features such as answering machine detection, redial rules and time of day restrictions.   Just make sure that you are strictly abiding by all TCPA directives for automation dialing.

How Does Nimblevox IVR work?

Nimblevox makes IVR deployments easy by hosting all of the complex telephony connections and speech recognition services in their the cloud.   All you need is a web connection and telephony access to implement your IVR.

The following will provide you with a foundation of what is happening “behind the scenes” to make your IVR application engaging and cost effective:

Nimblevox IVR platforms

Servers in the Nimblevox data center include network and telephony interfaces, data storage, and CPU horsepower to process audio in real-time.

Key components include:

  • Nimblevox Build – the Nimblevox Build tool is a graphical drag-and-drop web service you can use to build and deploy a telephony an IVR application.  It converts the call flow you have created to data that will be absorbed at run-time to execute the application logic.
  • Application Layer –  The application layer interprets the data you have built and deployed with the Build tool.  It executes logic blocks within the application as well as generates VoiceXML actions and issues CCXML events that will be handled by the interpreter within the IVR.   The Build tool also allows you to pull in your own custom VoiceXML code, which will be passed to the VoiceXML browser for execution.  In either case the browser “reads” your data and accesses the appropriate resources to deliver an interactive experience.
  • VoiceXML browser – VoiceXML is an industry standard protocol that defines how your speech application with will operate.   The browser handles the VoiceXML passed to it from the Application Layer.
  • Speech Recognition Engine – Speech recognition engines convert audio input into data that the VoiceXML browser can use.   In real-time a speech engine accepts audio input, and converts it into segmented data (“one” = /w/ /ah/ /n/, “two” = /t/ /uw/ ).  That data is then statistically matched to words (vocabulary) and phrases (grammar).

When the VoiceXML browser tells the speech engine to engage with an application it begins to  continually “listen” for input.   This is why a caller can “barge in” over an audio menu.   This is also where speech engines filter out extraneous noise to limit unintended background audio from triggering a false barge in.

Speech engines are continually improving, so you don’t have to worry about accents and other vocal characteristics.  You do have to define within your application what vocabulary and grammar to match to.   You can do that with packaged Nimblevox grammars, or you can define your own based on the VoiceXML standard.

Speech recognition accuracy is heavily dependent on matching your grammar to what callers actually will say.  Research has shown that 4 out of 5 speech application errors are due to something missing in the grammar and not due to background noise or anything to do with how a speech engine models the incredibly diverse human voice.

  • Telephony – Callers can dial in through mobile phones, landlines, VoIP phone and other devices.   This connectivity is managed through  the Nimblevox platform via its Computer Telephony Integration (CTI).
  • Networking – Your Nimblevox Build application does not have to exist as an isolated application.   Network access provides you with an ability to integrate your Nimblevox Build application with your existing systems.  This supports the ability to add personalization to your dialogs, and track activity within your reporting systems