Why you need a phase 2 for your voice app

by Peter Nann
20 February 2018
Why you need a phase 2 for your voice app
Launching an app on the Google Assistant is just the beginning. Salmat’s speech expert Peter Nann and his learnings on what you can do post-launch to make your app stand out.

Launching a voice assistant project is an exciting experience. The weeks leading up to the launch are always jam-packed as you finalise the minor details. The big day comes and you press the button to go live. Now your app is out in the world and ready to be used. But, the journey doesn’t stop there. Instead you’re straight into the thick of phase two – continuous improvement and promotion.

Similar to a new website launch, it's important to continue to test your app in the live environment to ensure everything works as it should. Also like a new website, there is the challenge of driving users to your app. However, the ephemeral nature of voice apps is forcing marketers to reconsider how best to do this.

In this blog, I will detail our approach to driving people to our new app on the Google Assistant, Australia Says. The app is a fun, voice-controlled game that lets players test their memory using true-blue, Australian slang words. Featuring over 70 classic Australian words to remember, the game challenges you to memorise and repeat a growing sequence of Aussie lingo, in your best Aussie accent.

When examining our app, understanding how users are engaging with it is a major focus now that the app is live and available. So, where did we start?

The Actions Console

As a developer, we had access to the Google Assistant Actions Console within 24 hours of setting up the app to view our app’s performance. Similar to Google Analytics, the dashboard within the console can be used by developers to get a view into the usage, health and discovery of their app. Analytics that can be tracked on the Actions Console include:


Track the number of the following in any one day:

  • Unique users
  • Conversations (from invocation to exit)
  • Messages (individual speech interactions, aka conversational 'turns')
  • Average conversation length
  • Abort rate (conversations that ended unnaturally)


Track the following:

  • Errors (use to identify what is causing your endpoint to crash or behave unexpectedly)
  • App latency (Delays from your app code – should be small)
  • Latency experience by user (Also affected by other factors, e.g. Networks)


Track various metrics for how people are 'discovering' your app (first time users)

'Discovery' is an active area of change in voice assistant platforms, so expect this area of analytics to evolve over the coming months.

Third-party analytics suppliers

For those looking for a deeper dive into analytics for their app, third-party suppliers such as  Adobe, dashbot.io and voicelabs.co offer solutions that integrate with voice application code to help developers understand such things as usage, user intent and message sentiment.

The theory behind the analytics provided by these third-party suppliers is in line with those we have been using in the speech space for years. Not only can you track how many people are using your app, but you can delve into the inputs and responses from the Google Assistant so you can continue to train it to understand your users, or consider new functionality you might want to add.

For example, a user may ask:

User: "What's a 'Hoges'?"
Google Assistant: "Was that 'burger'? Like a hamburger?"

So a possible extension to our game, Australia Says for example, could be to include explanations of what the slang words mean, if a user asks. This could be quite useful, even educational, for use outside of Australia.

Getting the word out

Setting up tracking for your app’s performance is all well and good, but what if usage is low? The age-old adage stands, what’s the point in having a shiny new car if you do not put fuel into it to power it? The same is true for voice assistant integrations. 

However, being such a new area there is little precedent for promoting apps. Remember, Google only launched the first iteration of the Google Assistant on mobile with Allo in September 2016, and with the Google Home in Australia in July 2017.

Of course, all apps for the Google Assistant have a listing in the Google Assistant Directory, which can be viewed on your phone via the Google Assistant app. However, beyond this, what else can you do to drive people to use your app? After all, the Google Assistant can be accessed not only on devices like Google Home, but also compatible Android and iOS mobile phones.

When seeking inspiration, where better to look than Google itself? In the months following the launch (and in the crucial run up to Christmas) the company launched a video ad campaign showing people using the device. 

Google Mini ad

Not only does it clearly demonstrate how the device works, but it gets the “Ok Google, … “ invocation stuck in people’s heads over time. Nobody knew how to work an iPod when they first came out, the same education will have to take place around invoking apps on the Google Assistant. 

We took a similar approach when promoting our Australia Says app. We created a couple of short, 30-second videos to illustrate how the game works. The videos are currently being promoted across a number of paid channels to get the word out.

For those that want more information, we have created a landing page with instructions on how to access and use the game – we took inspiration from this CNBC page and apps promoting mobile games when creating it. The resulting landing page can be found here

We will be keeping a keen eye on our Actions Console to analyse what impact the paid promotion has had on our user take up and engagement. 

Getting started

While there is no one approach to driving engagement, through our experience we have seen that there are clear steps developers can take for the continuous improvement and promotion of their apps on the Google Assistant. 

We are excited to apply our decades of experience with natural language IVR solutions to this developing new sector. While the voice assistant space is new and different in many ways, the VUI (Voice User Interface) design is the key to success, and this aspect is much the same as for Natural language IVRs. 

It is true, development has become much easier thanks to Google, Amazon and others, but design is also a crucial factor in creating a voice app. In fact, design is the difference between a solution, and a good solution.

We are ready and keen to design some great voice experiences for Australia. Are you ready to join us?

This is the final of our Australia Says blog series. You can read the first (How to speak Aussie with Australia Says) and second blogs (How we buily a game for the Google Assistant) on our blog here

For more information on Salmat’s speech services click here or call 1300 725 628.

Page 1 of 2
  < 1 - 2  > 
About the author
Peter Nann
Tech Lead-Speech & Automation, Salmat

Peter has been living and breathing VUI (Voice UI) development for 23 years, from the earliest commercial forays in this space in Australia in the 1990's, waiting all that time for the future to finally arrive.  Over the years he has worked with many organisations in Australia and abroad realising speech recognition solutions that service multiple millions of real-world user contacts every year.

More articles by Peter Nann