Should Amazon Alexa Stop Allowing Duplicate Invocation Names? Should Google Assistant Permit Them?
What do Daily Quotes, Math Quiz, Animal Game, Animal Sounds, Cat Facts, Yoga Guru, I’m Bored, Christmas Radio and 5 Card Draw all have in common? These Alexa skill names and their invocation names are not unique. There are duplicates. As a user, you can ask for an invocation name expecting one Alexa skill and actually wind up with the other.
For example, I recently wrote about how I asked Alexa to “talk to Daily Quotes” and a skill session started even though I had enabled none of the skills with that invocation name. There are six different Alexa skills with this exact invocation name and several others with a near match. What do the other five skill developers think about me delivered to a competing skill of the same name?
This was one of the examples I discussed in terms of Amazon’s experiment with Alexa skill auto-enablement. Should Amazon refrain from duplicate invocation names to avoid the twin problems of user confusion and picking winners? I asked several developers to weigh in on this question. One told me privately.
My opinion has always been that duplicates shouldn’t be allowed, but that ship has sailed, a long time ago.
The developer added that changing the current policy would likely not succeed even with a grandfathering approach where no duplicates were allowed going forward, but existing duplicates could remain. I was told this won’t work, “First because the implementation would require programmatic changes to the Alexa software and there are already so many duplicate skill names/invocations that lots of skills would be caught out by the new programmatic rules. Second, because there are existing invocation clashes with native functionality. So, even if your skill is ‘grandfathered’ under some new no-duplicates policy, it can still have its invocation superseded by native Alexa functionality.”
Google Takes a Different Approach
The question is particularly prescient because Google Assistant does not allow duplicate invocation names. This seems like a solution to the problem of duplicates. However, one developer who wishes to remain anonymous for the purposes of this story told me that the Google Assistant approach also has drawbacks:
Sadly, there is already a sort of squatting happening on Google Assistant. Many of my top Alexa invocation names were swooped up on Google Assistant by a copycat, and implemented in a half-baked way which provides nearly zero real value to consumers.
Oscar Merry, CTO and co-founder of Lifebot, has a similar perspective on the no duplicates approach. He would like to be able to use the same name for his voice app on Google Assistant that he currently uses for Alexa. However, Google told him no.
“I understand that there needs to be protection for people from copycats, but I also think that to have unique invocations makes no sense. I think it should be up to users to choose what they want a particular invocation to trigger. I think Amazon have struck the right balance in requiring proof of brand ownership for 1-word invocations but allowing duplicates for invocation names that can’t necessarily be trademarked. The reason I brought it up was because I was going to port my Alexa Skill Inspire Me to Google Assistant but someone has already taken that name and Google said I’d have to pick something different. Annoying!!”
Jo Jaquinta, CTO and co-founder of Tsa Tsa Tzu sees the merits of both approaches from the user and voice app publishers’ perspectives.
“For business, unique invocation names are important. You want to be able to establish and protect brand identity on the platform if you want to create something worthy of value and thus worthy of investment. But, in so doing, the platform takes on the burden of who gets what and when, and also has to manage the homesteaders, and what is and is not acceptable behavior in the field. So, Amazon’s ‘we’re not going to get involved’ approach is much easier for the platform. They don’t have to invest in policing and judging it.”
Amazon didn’t respond to request for comment on this topic, but a spokesperson for Google did have this to say about the no duplicates approach and the platform’s intent to make every Assistant app available without enablement.
As you know, users don’t have to download or enable any Actions with the Google Assistant. We’re focused on continuous updates to improve discovery so users don’t need to memorize the names of all their favorite apps.
The Google representative added, “We’re improving discovery (implicit and explicit) so users can find the right app they’re looking for without any additional steps. We also give developers best practices for improving their discovery with ‘action invocation phrases.’ And, we’re making updates to the app directory over time to help users find new apps.”
Google sees voice app enablement as unnecessary friction in the user experience. The no duplicate invocation name policy surely helps deliver the right voice app without an enablement requirement. It also simplifies discovery since a single name is associated with a single voice app making user confusion less likely. The downside is that the available name space for voice apps is finite.
Amazon’s Approach Has Its Fans
Voice user experience designer Peter Nann took the user’s perspective on whether duplicate invocation names should be allowed and saw the architectural merits of Amazon’s approach.
“At the moment, I lean towards the Alexa method. I prefer to know exactly which apps I have enabled. And then, sure, why couldn’t multiple apps have the same invocation and I pick which one to enable? The Google approach of ‘has to be unique because it is always enabled’ feels like it might run into trouble to me. I’ve always worried that Google’s approach might not scale. As 3rd-party apps hit 10’s and 100’s of thousands, almost any sensible string of a few words is going to sound like an app invocation, potentially. At least Alexa has the narrow gate of enablement that avoids them listening for ‘everything all the time.’ Then again, never underestimate Google’s ability to solve a problem given enough data! I am prepared for Google to amaze me.”
Eric Olson from 3PO-Labs echoed Mr. Nann’s comments and zeroed in on the issue that invocation names that are unique by their textual representation may not even be good enough. They must also be unique in terms of phonetic pronunciation to be truly unique under a no duplicates voice invocation model.
“I much prefer Alexa’s approach to the problem. While I respect that name-squatting is an age-old tradition on the internet – and I spent more than my fair share of time fighting over handles on IRC – the idea of unique naming won’t be sustainable on a larger platform. With voice, there’s an additional requirement that your chosen name be pronounceable. So, a lot of the early tricks people used, like replacing the letter O with a zero, for getting around written name uniqueness won’t work anymore. I understand why folks may believe otherwise, but the first time they find themselves saying something like, ‘Cortana, open x x cat facts 0178 x x,’ it’ll become clear that Google and Microsoft made the wrong choice.”
The Risk of Free Riders and Collateral Damage
The developers I spoke to on-the-record were split on this issue, but more favored Amazon’s current model. Off-the-record comments were more circumspect and raised concerns that Amazon’s approach devalues investment in your Alexa skill. Why? The duplicate naming means you can have free riders and collateral damage.
For example, one skill can work hard to advertise, generate publicity or drive positive word of mouth support from consumers. Other skills with the same name can capture new users based on these efforts simply because the consumer picks the wrong skill by accident. These free riders get the benefit of audience building without investment. However, the duplicate name skills can also compound the problem if they are not very well designed. In those instances, the skill that was mistakenly enabled might create a bad consumer experience that is communicated through word-of-mouth and creates collateral damage in that all of the skills with that name develop a bad reputation overhang. The process puts too much cognitive load on the consumer to make the right choice and sort through duplicate names to find the “right” one.
What is Best for the User, Publisher and Platform?
So, we are left with some bad choices here. Uniqueness would benefit the publisher and consumer, but it also severely limits the options for publishers due to the limitations of spoken word and could lead to some less-than-optimal invocation names for all but the first movers. The no duplicates model places more burden on the platform, more restrictions on the voice app publishers, but a simpler experience for consumers. The duplicates model involves less burden on the platform, some drawbacks for the publisher and introduces likely confusion for the consumer.
It is unclear how this will play out, but we have the experiment in process at the two leading voice assistant platforms. At some point, we will have data on usage and consumer sentiment regarding the two approaches. And, it is not clear these policies won’t change. We are already seeing a slight modification of Amazon’s approach as some Alexa skills are always enabled to reduce the access friction for consumers. Similarly, Google may be able to figure out with more user data how to offer duplicate invocation names while maintaining good user experience as Peter Nann suggests. Let me know what you think on Twitter.