From left, the Apple HomePod, Google Home and Amazon's Alexa are displayed for a photograph in San Francisco on Feb. 5, 2018. The $349 HomePod costs roughly three times its competitors and arrives in stores on Feb. 9. (Jason Henry/The New York Times)
On my second day with Apple's new HomePod, I asked the artificially intelligent speaker to play some music. Siri, the virtual assistant that powers HomePod, enthusiastically replied, "OK, let's get going with some Dashboard Confessional."
I cringed. "Hey, Siri," I said. "Nobody likes Dashboard Confessional."
Siri replied, "Sorry, I couldn't find the song 'Nobody Likes Dashboard Confessional.'" Then to my horror, HomePod continued playing a track by the emo rock band.
At the time, I gave Siri a pass. After all, Apple's HomePod, a rival to smart speakers from Amazon and Google, is supposed to study your music preferences over time to create special playlists just for you. I had had only one day with HomePod.
But after a week - during which I asked HomePod to play my favorite tunes from artists like Beck, Talking Heads and David Bowie - the smart speaker still did not learn. Instead, like a stubborn DJ, Siri kept playing music by artists outside my music palette: Taylor Swift and Leroy Francis, to name just two.
That leads to my conclusion: The $349 HomePod, which costs roughly three times its competitors and arrives in stores on Friday, is tough to recommend to you, dear reader.
Apple's speaker is certainly an impressive piece of hardware. Audiophiles will appreciate that it has a woofer with a custom amplifier and seven tweeters. The result is a speaker with a deep bass and rich treble that is loud enough to fill a large room with superb sound. HomePod makes the Amazon Echo and Google's Home sound muffled and tinny in comparison.
But Siri on HomePod is embarrassingly inadequate, even though that is the primary way you interact with it. Siri is sorely lacking in capabilities compared with Amazon's Alexa and Google's Assistant. Siri doesn't even work as well on HomePod as it does on the iPhone.
For Apple, that's unfortunate. The company was the first to bring virtual assistants to the mainstream with Siri on the iPhone in 2011, but it has since fallen behind Amazon and Google with smart speakers. Apple announced HomePod last June - but then delayed its release until this year.
Even now, Apple is shipping the HomePod unfinished. On day one, the device will lack some cool features, like the ability to link several HomePods together to create a multiroom sound system that Apple says will fill an entire home with music. That feature will come in a software update later this year.
And there are other limitations: The HomePod requires an iOS device, like an iPhone, an iPad or an iPod Touch, to set it up. To use your voice commands to play music, you will also need to subscribe to Apple's streaming music service, Apple Music.
So how exactly did I reach my conclusion on HomePod? I tested it side by side with Echo and Home smart speakers, grading them on their ability to accomplish 14 tasks across several categories, including music, productivity, commuting, home automation and cooking. Let me walk you through the process and results.
Choosing the Tests
I started picking the 14 tasks by reading up on research studies that looked at how people use virtual assistants in the home. Activate, a management consulting firm, found the majority of people turned to virtual assistants to play music, get the weather and set a timer.
Apple also provided statistics on smart speaker usage from the research firm Parks Associates. That report also found that playing music and getting the weather were the top uses of smart speakers, while roughly 20 percent of people enjoyed using them for tasks like accessing a calendar and searching for recipes.
Amazon says most of its Echo customers use at least one "skill," or third-party app. So I added the ability to use Uber, the most popular ride-sharing service, as a test.
Now onward to the tests themselves.
Questions and Answers
I set up a HomePod, an Echo and a Home in my house and began with a battery of question-and-answer sessions. First up: traffic.
All the speakers gave a similar traffic estimate for a drive to San Jose, California - roughly a one-hour drive on the freeway. But when I asked HomePod to summon a car from Uber, Siri responded, "I wish I could, but I can't help with rides here."
The other speakers were happy to help - so I followed up with: "Hey, Siri, what gives?" HomePod's colorful touch-screen lit up to show it had heard my question, but Siri remained silent.
Next, it was time for some cooking questions. All the speakers were able to flawlessly set a kitchen timer. But when I asked HomePod how to make pasta, Siri blanked and said, "I can't get the answer to that on HomePod."
Google Home was more responsive - it gave me the steps for making pasta from the recipe site Genius Kitchen. Amazon Echo was even more accommodating. Alexa listed the ingredients for a pasta dish, including noodles, milk and heavy cream, and offered to send the steps to my phone or play them aloud.
Then it was time to move on to some work-related tasks. I asked Siri to schedule a meeting for Tuesday. Siri responded: "I cannot access your calendar from here. Sorry about that." It couldn't look at my calendar for the day, either. Google Home and Amazon Echo, in contrast, managed to schedule new events and read my calendar for the day: a meeting in the morning, followed by a photo shoot and a business dinner in the evening.
When I asked Apple about Siri's hiccups, the company said that for the first version of HomePod, it focused on including tasks that people use smart speakers for the most, like playing music and asking about the weather, and that it would continue to evaluate what other features to add over time.
I wondered whether HomePod would do better with smart home tasks and decided to ask all the smart speakers to turn on a Wi-Fi connected light bulb from the smart light company Lifx. In this test, HomePod got a higher score because setting up the smart light, which involved using the iPhone camera to scan a code on the instruction manual, was seamless and much easier than it was with Google Home and Amazon Echo, which required installing a third-party app.
Bonus: The light turned on when I asked.
One of the most crucial tests for HomePod had to do with audio. That's because Apple has long emphasized that the smart speaker is first and foremost a music player.
Phil Schiller, Apple's head of marketing, said at a gathering at the company's audio lab last week that Apple started developing HomePod six years ago with the intention of making a speaker that specialized in playing music in a unique way. The project started well before Apple introduced other services like Apple Music, which was released in 2015, and HomeKit, the home automation platform that was unveiled in 2014, he added.
Schiller said Apple wanted to stay true to the original goal of the project "without ever compromising that it's a speaker first."
All three speakers did great playing music. Echo, Google's Home and HomePod were each able to play specific songs by artists, generate playlists for specific artists or music genres, and play podcasts. HomePod got a higher score in audio with its superior sound quality factored in.
Yet that's also where Dashboard Confessional came in and where HomePod's biggest shortcomings became apparent. Whenever I asked HomePod to "play some music," it never played music that was relevant to my preferences or listening history.
That wasn't the case with the Google and Amazon speakers. When I asked those speakers to play music, the gadgets simply resumed what I was last playing on Spotify, which was satisfying.
(BEGIN OPTIONAL TRIM.)
Siri also had laughably awkward pronunciations of some artist names. When I once asked HomePod to play songs by Tupac, Siri replied: "Sure, here's Tu," and after a short pause said: "Pac." (I'm not sure Dr. Dre or Jimmy Iovine, whose streaming service was acquired by Apple to develop Apple Music, would approve.)
In response to my concerns, Apple said HomePod studies a customer's music preferences over time. I figured a week should have been enough.
From the tests, I graded the speakers on their ability to accomplish each task on a scale from 0 to 4. (I gave a 0 for tasks that could not be done at all, a 2 for tasks that could be done with some issues, and a 4 for tasks that could be completed flawlessly.)
So how did they all stack up in terms of grade-point averages out of 4.0? The final results:
- Amazon's Echo - 3.4
- Google's Home - 3.1