Will Alexa Destroy Us ?
How your radio newsroom should think about smart speakers and digital assistants, and how two non-developers built a product on one
by Gabriel Spitzer and Brian Edwards-Tiekert
By the end of 2017, there will be an estimated 36 million active users of “smart speakers”: devices that aspire to become voice-controlled middleman between their users, The Internet, and Everything That Can be Connected to The Internet. Among the first applications they steer their users toward: asking for news or music — which also positions them squarely between radio producers and our audiences. In Stanford’s John S. Knight Journalism Fellowship, we’ve been looking into how that changes the producer/audience relationship — and what local newsrooms can do to adapt. Here’s what you need to know:
Who’s using smart speakers?
By year-end, about 15% of US adults will be using smart speakers in their homes. Users skew young: the heaviest penetration and fastest growth is in the 25–35 age bracket. The growth is explosive: Two years after the introduction of the Amazon Echo, smart speaker use is climbing almost as fast (currently about 130% per year) as smartphone use was two years after the introduction of the iPhone. And home speakers are just the beginning: Tech companies are inking deals to install the same voice-controlled digital assistants in the place that has been the bedrock of radio listening since the advent of the transistor: the car.
How are they getting used?
We have to do some educated guesswork, because all the user data belongs to the platforms — and Google and Amazon ain’t sharin’ (more on this below). But here are some basic points:
News is native. Both Google and Amazon’s smart speakers deliver headline news packages in response to basic prompts (e.g., “tell me about my day”, or “what’s news”). By default, both serve NPR headlines first. A user doesn’t have to manually install a skill or app to get news (though they do have to do some customization via their smartphone if they want to change *where* they’re getting news from). Unsurprisingly, “stream news” registers among the top uses in surveys of early smart speaker adopters:
Listeners are (still) busy and distracted. The most common location for an Amazon Echo is in the kitchen, suggesting the use case for smart speakers is the same as for radio: It’s a good way to get information when you need your hands and eyes free for other tasks (e.g., cooking).
Dominant use: morning routine. In our discussions with early adopters of both Google Home and Amazon Echo, the most common use time was in the morning, as part of a daily routine. The platforms design for this: Ask Google Home to tell you about your day, and it gives you the time, the local weather, traffic conditions on your commute, upcoming appointments from your calendar, and then plays news headlines. Expect news products that are updated daily (or more frequently) to perform better in this context than less-frequent long-form products.
News is bundled. Smart speaker platforms bundle headlines from multiple publishers — say, one for political news, another for business news, a third for international news, etc. (there’s no directory category on either platform for “local” news yet — grrrr.)
It’s easy to switch. There’s less friction than ever for a listener who wants to change what they’re listening to. They can tell their smart speaker to skip ahead, or play something else, without even walking across the room — much less putting down the scouring pad and drying off their hands before they touch the dial.
It’s hard to be discovered. In the glory days of FM radio, your station could practically count on local listeners discovering it, because there were never more than 24 other stations on the dial. A user looking for new flash briefings to add to their Amazon Echo is confronted with a barely-categorized list of (at last count) 2,693 options. (Note to Amazon: Want to keep your name off the epitaph of local news? Create a “near you” category for your flash briefing directory).
Listeners only find what they know how to ask for. The live stream of pretty much every US radio station is available through smart speakers — but only if you know how to ask for the station by name. Here are some questions to consider:
When your listeners think about what they already listen to, do they think of your frequency, your call letters, your tagline? Or do they just think they’re listening to “public radio”? What happens when they ask their device for any of those things?
(From our perch at Stanford, a request for “Public Radio” turns up nothing. A request for “National Public Radio” directed at an Amazon Echo delivers the stream for San Francisco-based pubcaster KQED. The same request delivered to a Google Home delivers the stream for San Francisco’s other pubcaster, KALW. Meanwhile, a request for K-A-L-W by name draws a blank on Google (who apparently thinks that word should be pronounced “Cawl-W”).
Can my station find new listeners there?
Probably not. Because of the discovery limitations, you’re not likely to introduce yourself to new listeners on a Google Home or Alexa-enabled device. However, more and more of your listeners will be moving to these devices, and you might lose your connection to that slice of your audience if you don’t make it easy for them to find you there. So, until Amazon or Google build a local news discovery function into their platforms, think of working with these platforms as a play for audience retention rather than a play for audience growth. You’ll need to use your existing communication channels (FM signal, newsletter, website, social accounts) to tell your listeners how to find your flash briefing, request your podcasts, or load your stream.
What *can* we do there?
Once you’ve got people listening to, say, your flash briefing, you can also tell them what else to ask for — so think hard about where you’d like them to go next. For instance, you can sign off by telling your listener how to ask for one of your podcasts by name (“That’s it for our morning report: for more local news, say “Alexa, play “Town Hall Talk”), or by giving your listeners the words they need to get your stream on the first try (‘. . . to tune in live say “Hey Google, play KALW 91.7”’).
Remember: if your relationship with your audience on a smart speaker platform doesn’t go any deeper than a three-minute headline package, then your work will mainly produce value for the platform — so try to steer your audience to the places where you know how make money. (Note to Amazon and Google: You guys have already connected your devices to your payment systems — the least you could do it make *possible* to donate using those systems).
How hard is it to actually build for these platforms?
Not that hard! We are two journalists with no actual coding skills, and we did it without acquiring any such skills. Specifically, we built a flash briefing for Amazon’s Alexa assistant (which currently dominates the smart speaker market). Here’s what you do:
- Sign up as an Amazon developer through their developer portal.
- Select “Alexa” in the top menu, then select the option called “Alexa Skills Kit” (this is also how to access your skill once you’ve made it — which is not obvious).
- Select “Add a new skill,” and choose “flash briefing API.”
- Follow the form to name the skill, add a description, customize messages, and link it to an audio feed.
Here we encountered one major hitch. Previous posts on building flash briefings led us to believe it was as simple as creating a free podcast through a service like Soundcloud, then pasting a URL for the RSS feed into one of Amazon’s form. We tried various free hosting services (Soundcloud, Podbean, Podomatic, etc.) — without success, and with little explanation from Amazon beyond an error message.
It turns out Amazon has recently made some security upgrades, and now has fairly specific requirements for RSS feeds driving flash briefings — requirements that no off-the-shelf service we experimented with could meet.
If you have some coding experience, those requirements may pose no problem: Amazon provides boilerplate code that you can incorporate into your RSS feed to make it compliant. Since we don’t edit code, this was lost on us. (You can also use something called a JSON format, which may mean something to you if you know something about coding, but was another option wasted on us.) Since we wanted to know how a small newsroom with no in-house developers would fare, we intentionally didn’t seek out assistance from a developer.
We eventually tried hosting our RSS feed on the web site of an NPR-affiliated station that uses NPR’s Core Publisher platform. That feed turned out to work fine.
- Finally, submit your skill to Amazon for approval. (Note: You can’t make changes to the skill while approval is pending). They are pretty prompt: every time time we sought approval for a prototype skill, we got an answer in about 24 hours.
Amazon’s technical support, especially for lay developers, is quite skimpy. There is no phone number or live chat that connects you with a real person. Questions are channeled into a public forum, where it may or may not get an answer (Gabe posted a question about our RSS feed woes that got no response whatsoever).
How do you know if you’re finding an audience?
With flash briefings, you don’t. Amazon doesn’t have any kind of analytics — you won’t know how many people are subscribed, actually listening, and/or when they’re tuning out. (Note to Amazon: you should really build that.)[UPDATE 6/7/17: Amazon did just build it.] And since Amazon caches the audio of your flash briefing, you can’t count on metrics from the RSS feed that’s sending the audio to Amazon either. The only things you can do, at this point, are survey your listeners and monitor the other things you’re hoping to funnel flash briefing listeners into: for instance, podcasts and live stream listening. Both get served by TuneIn by default — and TuneIn offers a free analytics service here.
Bottom line?
There’s a lot of potential here for public newsrooms, even local ones. Smart speakers are expanding the parts of the day in which listeners have easy access to your content. The platforms they’re running seem well-suited to hard news. Although there’s currently no location-based news discovery on any smart speaker system, it wouldn’t surprise us if one comes out soon. (Google and Amazon are both engaged in a feature war as they race for market share — and Apple’s likely to enter the fray with a competing product by year’s end.) [UPDATE 6/6/17: Apple just announced a December release for its HomePod]
“Voice-activated devices play directly to the strengths of public radio.” — Thomas Hjelm, Chief Digital Officer, NPR
For small independent stations, the barriers to entry are pretty manageable. Your stream and podcasts are probably already accessible on both platforms — you just have to tell your existing audience how to find you there. With a little more effort, you can build a flash briefing (or the Google equivalent — which doesn’t seem to have a name) to improve your chances of staying part of your listeners’ morning routine. While it’s also possible to build your own skill or voice app outside the Flash Briefing API, it’s unclear what you’d gain for the added work and expense (if you have success stories from your newsroom, please write us!).
As smart speakers proliferate and their digital assistant platforms develop, the upsides of having your news there are going to grow. NPR Chief Digital Officer Thomas Hjelm sent us this bullish assessment via email:
“Voice-activated devices play directly to the strengths of public radio. Not only are we a source of excellent and relevant content — whether it’s headlines or involved storytelling — but the voices on our airwaves are the original audio companions to your day. As the medium evolves, we intend to be there with it, and integrating the local expression of public radio alongside the national content is central to that strategy. We believe the national/local blend of content, branding and sponsorship is an essential and defensible asset of public radio, and as these devices get smarter, localization tools sharper, the rendering of conversational interfaces more sensitive, it will create only more opportunities to deliver the national and local conversation in more integrated ways.”
