View profile

Hearing Voices - Issue #32, or "Ok Google, set phasers to kill"

"Ok Google, set phasers to kill" - You, to your Google Home "Ok Matt, I want that premium data" - You

Hearing Voices

October 20 · Issue #32 · View online
This week's news about voice computing apps, delivered directly to your inbox.

“Ok Google, set phasers to kill”
- You, to your Google Home
“Ok Matt, I want that premium data”
- You, deciding to become a hearingvoices member for $18/mo (join here)

Wow, big couple weeks in voice, plus I found another person getting paid for Alexa skills, interview below. Let’s get to it:
To try:
> Pet Translator – This is a fun one. Ask your dog what Alexa just said. You can also submit your own translation.
> Chaptter - Free audiobooks on Alexa, just say “Alexa open Chaptter.”
> Curio - Listen to The Guardian, Financial Times, and Aeon magazine on the go. Also, premium members look out for an interview with the founders about user data, revenue, and licensing costs coming soon.
> Castbox - a new podcast player that raised $16m.  More on it below in the newz section.
> Blockchain Curated - you know how much I’m obsessed with listening to my newsletters, so here’s a great service that selects the top cryptocurrency articles written by industry experts each week and converts them into a free podcast.
Google’s Airpods competitor - not out yet, but apparently they do real-time language translation. Muy increíble.  
Industry Newz:
After raising $16M in funding, CastBox plans to fix podcast discovery - CastBox hopes to become the breakout podcast player for iOS and Android. It will start co-producing and releasing original content this quarter and its monetization plan includes a premium option with extra features.
The rise of the podcast adaptation - Serial, Startup, and Lore (a horror podcast) are all being adapted for the screen. The Economist digs into this phenomenon..
> Libsyn is updating its stats.
Anchors Aweigh - Google Ventures led a $10m investment round into Anchor. Here’s MG Siegler’s take on why he’s so bullish on audio and on Anchor.
Invested in Audio: Life in ‘The Podcasting Capital’ - @pmontee who ran voicecamp at betaworks, wrote this post on investing in audio products.
> Going Big with a Consumer Voice Skill Startup? - @davidbeisel’s take on how developers should react to Amazon’s constraints.
On the technical side:
WaveNet: A Generative Model – WaveNet is the new tech out of DeepMind for Text To Speech. This is a technical description of it, but it also has some audio examples that are pretty cool. 
Q&A: inside track on making $$$$ w/amazon
I found another person who’s making $$$$ from Alexa’s developer program. Joel Wilson is both a Data Architect and Data Scientist who built Question of the Day and Three Questions. He shared some interesting insights here, and also some detailed metrics with hearing voices premium members ($18/mo). Here’s our conversation:
MH: Hi Joel, you work on two popular Alexa apps: Question of the Day and Three Questions. Are these the only two apps you’ve developed, or just the most popular?
JW: These are the two most popular apps. I created a few others: “Paige One” with news content provided by the Associated Press, and “Trivia Streak” just a few weeks ago. “Paige One” was fun to work on as I built out the framework that I had created with the first few apps. Now that I have that framework in place, I can create many more apps quickly - it’s just a matter of curating the right content. The new “Trivia Streak” just utilizes the same trivia content that I already have and seems to be gaining traction in the UK.

MH: How did you get started building alexa skills?
JW: My regular job is as a data scientist and I run a small company that does predictive analytics for marketing. I thought Alexa would be a great experiment in collecting data - literally asking people to respond to questions. And I love staying on top of the latest technology. Amazon makes it very easy to program Alexa skills. I think I had Alexa responding to my commands within a half hour of my getting started.

MH: Can you talk about what your usage looks like for each of the apps
JW: I have had [Premium Members Only] “installs” so far. The majority, [Premium Members Only], are for the Question of the Day app, which has about [Premium] daily active users and steadily growing. About [Premium]% of those daily users are returning users, so retention is fairly strong.

MH: How did you get initial traction for these skills and others? Did you promote them? Amazon? Your Partners?
JW: I didn’t do any promotion. I wrote the initial apps in Jan/Feb of this year and received about [Premium Members Only] users a day for the first several months. It wasn’t until May, when Amazon began featuring them in the store, that things started taking off. And that’s when the payments began.
I have done a few experiments in order to drive retention. Since I collect the email addresses for a lot of my users (I have about [Premium Members Only] users with an email attached their user id), I am able to run FaceBook ads for retention. I am also planning on running email campaigns soon.

MH: Can you talk about your development process? Testing is always difficult since there’s no equivalent to testflight. how do you test your skill prior to launching it?
JW: My development process is very agile and experimental. It’s just one step at a time without a big master plan. I will experiment with an idea, like adding points to the games, or allowing users to link their accounts, etc., and just see how things play out. Testing is actually easy: voice conversations are very constrained, so there are only so many ways to foul things up. And now that I have a dialogue framework in place, coding is really simplified. The bigger challenge now is content curation.

MH: Can you share the math of how payments work? i.e., if I get 1000 people to enable my skill, how much will I make? Or does retention matter? Or is it a black box? If it’s a black box, is there information that you can share with our premium members about the relationship between how many users the skill has and how much you make?
JW: It’s a black box. I have attempted to reverse engineer the payments from my usage data (even to the point of running regression analysis) but with little luck. Best as I can tell, the process is something like:
1. Amazon decides how much they are going to pay for each category.
2. Skills in each category are ranked based on their definition of “engagement", which is based on:
a. Number of users
b. Number of “utterances" per user (that is, the number of times a user responds to Alexa within the app). This is perhaps the most important metric. Even if you have a lot of users, if they are not engaging with the app, your payments will be limited.
c. Customer revues.
3. Your skill allocation of the pie is awarded based on your engagement ranking.
4. There seems to be a max that they will pay for any one skill per month: [Premium Members Only]

MH: Thanks Joel! 

Join Hearing Voices as a premium member to learn more about my conversation with Joel.
Why Voice, Why Now? Trends in A.I. and Voice-Driven Technology - In DUMBO in brooklyn. I’ll be speaking at this one.
Smart Voice Summit - If you’re in Europe, check this out to discuss trends, challenges and the future of the industry. Premium members text me for a discount code ;)
As not-so-subtly hinted at, you can now join hearing voice premium to get timely alerts on voice news, get insider data, and chat with Matt about topics in voice UI. Limited to the first 130 people. Subscribe here:
Thanks so much for reading this week, have a great weekend!
- Matt

Did someone forward this to you? Subscribe at
Follow me on twitter @matthartman
Become a premium member of Hearing Voices by joining here to get timely alerts on voice news, and chat with Matt about topics in voice UI.
Did you enjoy this issue?
If you don't want these updates anymore, please unsubscribe here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue