Why CarPlay is a dead-end & how Apple can turn it around.

For the benefit of the TL;DR crowd, here’s the gist of this piece: CarPlay is a fundamentally flawed product that should be scrapped to give the engineers in Cupertino the space to come up with an in-car concept that’s actually better than the status quo.

Now, for the Android partisans who’ve landed here in the hopes of finding an Apple hit piece, I believe Android Auto suffers from the same fundamental issues, so: These aren’t the droids you’re looking for … move along … move along.

Okay then, for the four of you who are left, part of the reason I’m calling Apple out on their in-car initiative rather than Google is that Google has not published a manifesto claiming that “every idea we touch enhances each life it touches” (see the video below). Nor has Google CEO Larry Page publicly declared that “We’re not in the junk business,” as Apple CEO Tim Cook has on multiple occasions (suggesting, not so subtly, that others are).

It’s fair to say that I’ve found the tone of Apple’s recent pronouncements more than a little off-putting in their tendency towards narcissism and passive-agression. And yet, I’ve been a longtime fan of the brand and its offerings, so what truly inspired me to write this overlong treatise on CarPlay was my deeply disappointing experience with the product itself. I might even go so far as to call it a junk product, Tim Cook’s protestations to the contrary notwithstanding. Before diving into why it’s junk, however, I feel a quick primer on CarPlay is in order, starting with Apple’s own description:

“CarPlay takes the things you want to do with your iPhone while driving and puts them right on your car’s built-in display. You can get directions, make calls, send and receive messages, and listen to music, all in a way that allows you to stay focused on the road. Just plug in your iPhone and go.”

Note that that last line is meant literally. Even if your car supports Bluetooth audio and hands-free control, CarPlay requires a physical connection via a Lightning-to-USB cable. Now, if you’re a Mac geek, you may have read rumors indicating that the latest beta release of iOS 8.3 includes support for something called “wireless CarPlay,” but that tidbit comes with a caveat the size of an Escalade: By all accounts, wireless CarPlay will only work in vehicles equipped with both CarPlay and in-car WiFi connectivity. So, those hoping for CarPlay over Bluetooth to ever be a thing are almost certainly out of luck. And, given the slow rate of replacement for automobiles, this means the overwhelming majority of CarPlay users will be stuck with a wired connection for years to come.

This image, courtesy of Mercedes-Benz, shows an as yet unreleased implementation of CarPlay that should be available in C-Class models before the end of this year. Note that all incarnations of CarPlay demonstrated to date require a wired connection to an in-car USB port, as illustrated here. I think many will find this to be a step backwards from existing Bluetooth hands-free systems, which are completely wireless. Worse yet, unless the manufacturer of your ride offers a post-purchase upgrade to in-car WiFi (something that’s laughably unlikely), you’ll be stuck using CarPlay with a wired connection for as long as you own your vehicle.

This image, courtesy of Mercedes-Benz, shows an as yet unreleased implementation of CarPlay that should be available in C-Class models before the end of this year. Note that all incarnations of CarPlay demonstrated to date require a wired connection to an in-car USB port, as illustrated here. I think many will find this to be a step backwards from existing Bluetooth hands-free systems, which are completely wireless. Worse yet, unless the manufacturer of your ride offers a post-purchase upgrade to in-car WiFi (something that’s laughably unlikely), you’ll be stuck using CarPlay with a wired connection for as long as you own your vehicle.

Speaking of wired connections, astute readers may have noticed the specific callout for a Lightning-to-USB cable in the paragraph above. That’s because CarPlay requires a Lightning-enabled iPhone, meaning the iPhone 5 or later. Putting these bits together, the minimum requirements for CarPlay are a recent vintage iPhone and a set of wheels with a CarPlay-compatible infotainment system (including a USB port). The shaded region of the Venn diagram representing the intersection of these requirements is pretty small today, but it is growing.

Rough Edges
Right then, now that you know what CarPlay is meant to do, what’s so bad about it? Well, for starters, it has more rough edges than a coal mine. For example, nearly every review I’ve read has criticized CarPlay for lacking the real-time responsiveness we’ve come to expect from our devices—a problem I experienced first-hand when I tried CarPlay on a Pioneer aftermarket head-unit (the Pioneer AVIC-5000 NEX to be exact, which retails for $750). Here’s how Zac Hall put it in his review of CarPlay for 9to5Mac:

“CarPlay can also be sluggish at times. Tapping an icon to launch a function could sometimes result in 5 or 6 second delays, which isn’t long once it works, but seems like ages while you’re waiting and not knowing if a touch was received.”

Consider that a car traveling at 60 mph would cover anywhere from 440 - 528 feet over the course of this 5 - 6 second delay (for reference, an American football field is 360 feet long, including the end zones), and you realize how problematic an unresponsive in-car system can be. But more on distraction and safety in a minute.

Reviewers have also noted that the few third-party apps available today via the CarPlay interface are, at best, problematic and, at worst, wholly unusable. Longtime Apple watcher Jason Snell highlighted this issue on his recently launched Six Colors blog:

“Apple’s site lists eight third-party apps that work with CarPlay, but I could only get two to work: iHeartRadio and Overcast. Both of them provided simple interfaces a la the [Apple] Podcasts app, but I found them to be slow and unreliable. In my testing, both iHeartRadio and Overcast sometimes failed to display any menu at all. Other times, there would be long delays between menu items. And the apps frequently just crashed, stopping playback and taking me back to the home screen.” [Note: You can see this in action at the 5:30 mark in Snell’s video review embedded above.]

Snell goes on to note that, “If one of the apps worked great, I could lay this at the feet of the other app developer, but since both of them are buggy I’m more inclined to blame CarPlay itself.” It’s worth calling out here that an app must be specifically approved for use in-car by Apple before it can appear within the CarPlay interface. For example: I’m a fan of the Downcast podcast player and had it installed on my iPhone when I took CarPlay for a test drive, but the app did not appear on the car’s screen because Apple has not yet approved it for use in-car. At least there’s a good chance that Downcast will eventually earn Apple’s “Works with CarPlay” blessing, but there are classes of third-party apps—maps and messaging being two biggies—that will almost certainly never make it to the CarPlay screen.

So it’s obvious to anyone who spends time with the system that CarPlay suffers from issues with interface lag, stability and third-party app availability. But most reviewers conclude that these foibles will be addressed soon—a fair assumption given the rapid development cycles that are the norm for consumer electronics companies like Apple. Indeed, this is one of the leading arguments mooted in support of systems like CarPlay: While mainstream automakers measure iteration cycles in units of years, tech companies live by a culture of constant iteration that industrial firms simply can’t match. The problem with this assumption, though, is that CarPlay doesn’t actually replace the underlying hardware or operating system powering a car’s infotainment system.

Mirror, Mirror
Apple has been characteristically tight lipped regarding the inner workings of CarPlay, but a few automakers have let slip that it’s essentially a variation on AirPlay Mirroring. In other words, your iPhone uses your car’s built-in display as the equivalent of an external monitor, with all CarPlay-related computing tasks handled by the phone. That sounds pretty straightforward, until you realize that the car’s infotainment system still needs its own OS so that it can, among other things, relay inputs made via the car’s touchscreen and other physical controls back to the phone for processing. This bit—the layer that sits between the car’s hardware and your phone—is not controlled by Apple, and probably never will be for a range of reasons that aren’t worth getting into here.

With each passing model year, cars are becoming less mechanical and more digital. And while purists may bemoan “advancements” like electronic stability control, electric power steering and in-car infotainment systems, there’s no question that the transition to digital has led to improvements in safety, efficiency and convenience. Electronics have become so pervasive, in fact, that it wouldn’t be an overstatement to describe the modern car as a computer on wheels and, like any computer, your car needs an operating system. The in-cabin OS of choice for most automakers today comes from a company called QNX, which is a subsidiary of Blackberry (yes, that Blackberry). Illustrated above is a high-level overview of QNX’s real-time operating system, called Neutrino. As you can see, it’s designed to be highly modular, working atop a range of processing architectures and supporting a wide array of user-facing inputs and outputs. Click the diagram above to see some of the latest advancements in in-cabin tech, many of which are powered by embedded systems designed by QNX.

With each passing model year, cars are becoming less mechanical and more digital. And while purists may bemoan “advancements” like electronic stability control, electric power steering and in-car infotainment systems, there’s no question that the transition to digital has led to improvements in safety, efficiency and convenience. Electronics have become so pervasive, in fact, that it wouldn’t be an overstatement to describe the modern car as a computer on wheels and, like any computer, your car needs an operating system. The in-cabin OS of choice for most automakers today comes from a company called QNX, which is a subsidiary of Blackberry (yes, that Blackberry). Illustrated above is a high-level overview of QNX’s real-time operating system, called Neutrino. As you can see, it’s designed to be highly modular, working atop a range of processing architectures and supporting a wide array of user-facing inputs and outputs. Click the diagram above to see some of the latest advancements in in-cabin tech, many of which are powered by embedded systems designed by QNX.

The bottom line is that, while all CarPlay-related computing tasks are offloaded to your iPhone, the overall system must still work within the compute and memory constraints of your car’s built-in hardware—constraints that are typically defined years in advance of a vehicle’s launch and remain fixed over its 10-plus year lifespan. Because of this, it’s far from a sure thing that a customer adopting CarPlay today will see dramatic improvements in the speed or stability of the system over time. Just ask anyone who’s cursed under their breath at the crapshoot that is AirPlay Mirroring to an Apple TV—a platform that Apple fully controls—whether Cupertino deserves the benefit of the doubt when it comes to AirPlay performance.

And yet, all this being said, these rough edges are merely peripheral to CarPlay’s core problem, which is that the platform as conceived today is attempting to answer the wrong question. Macworld’s Susy Ochs hinted at this fundamental disconnect after spending 72 hours with the system, ultimately concluding that “CarPlay can’t do anything my iPhone doesn’t do on its own, and obviously the iPhone has a lot more apps.”

That’s Not What I Asked
It’s clear that the goal for CarPlay was to improve upon existing in-car infotainment systems. And there’s no question that existing in-car systems leave a lot to be desired on the usability front. In fact, according to Consumer Reports’ most recent Annual Auto Survey, “The area that garners the largest number of complaints by far is infotainment systems and associated electronics.” So it might seem that this is a great question to tackle: Build a better in-car infotainment system, and the world will beat a path to your door!

The problem is, a better alternative already exists, and Apple created it: It’s called iPhone (or, more broadly, the modern smartphone). An iPhone in your pocket or purse linked to your car wirelessly via Bluetooth is superior to any built-in system on the market today—it’s always connected to the Internet, offers access to over a million purpose-built apps and already contains all of your personal data and media. Yet, rather than improving on this alternative, CarPlay hobbles it by limiting the availability of apps; delivering its few supported apps via a laggy, unresponsive interface; and forcing users to physically tether their phones to their cars.

For those who’ve been following Apple for a while, this may stir memories of Steve Jobs’ take on the smartphone market prior to the introduction of iPhone in mid-2007. From his celebrated Macworld keynote in January of that year up till the phone’s launch in June, Jobs repeatedly highlighted the limited capabilities of competitors’ devices—particularly in terms of their Internet connectivity. For example, here’s an excerpt from Jobs’ conversation with Walt Mossberg at the D5: All Things Digital conference in May of 2007:

“You’ve used the Internet—or youve tried to use the Internet—on your phone, and it’s terrible. They have lousy browsers ... I mean, you don’t get the Internet, you get the baby Internet or the “mobile” Internet or something bizarre. And what people want is the real internet on their phone.”

To paraphrase Jobs, what Apple is offering today in CarPlay is the baby iOS—underpowered and mystifyingly limited in functionality—when what people want is the real iOS, made better for use while driving.

As an observer looking at this from the perspective of a product manager, my conclusion is clear: Apple needs to shift from their current question, which asks: “How can we improve on today’s in-car infotainment systems?,” to a much tougher—but also much more meaningful—question: “How can we improve on today’s in-car iPhone experience?”

Focus Means Saying No
So, where to begin? Based on a review of the in-car infotainment landscape, if I were calling the shots, I’d focus on two fronts: safety and convenience. And, to enable my team to pour all of their efforts into delivering real benefits on these fronts, I’d shut down any work related to integration with in-dash displays—in other words, I’d kill CarPlay as it exists today.

Why? For starters, as outlined above, this attempt to piggyback on in-car systems shackles Apple to embedded hardware that’s already years out of date by the time it reaches dealers’ lots. But, more importantly, anything that takes drivers’ eyes off the road—such as in-dash displays—increases the likelihood of an accident. I’m sure that sounds terribly obvious, but the statistics are striking: According to data from the National Highway Traffic Safety Administration (NHTSA), distraction was a factor in 10 percent of fatal crashes in the U.S. in 2012, which translates to 3,328 souls lost to entirely avoidable circumstances. Furthermore, related research suggests that 15 - 20 percent of all in-car distractions involve drivers interacting with technology. This is a real problem, and adding another screen-based experience in the form of CarPlay does nothing to solve it.

The obvious alternative is voice control, and those familiar with CarPlay will note that the system already makes extensive use of Apple’s Siri voice recognition platform. It’s true that independent research has shown that “driving performance is better … and there is less time spent looking away from the road when using speech as opposed to manual [i.e. touch based] interfaces to operate an in-vehicle system” (Barón & Green, 2006). The problem is, not all speech interfaces are created equal, and a damning study by the AAA Foundation for Traffic Safety suggests that Siri is at the bottom of the heap when it comes to in-car performance.

The chart above is based on data from a rigorous study published by the AAA Foundation for Traffic Safety in October of 2014, and titled Measuring Cognitive Distraction in the Automobile II: Assessing In-Vehicle Voice-Based Interactive Technologies. I highly recommend downloading the full study if you want to dig into the details, but, in a nutshell, this infographic depicts the levels of in-car cognitive distraction associated with several of the most widely used automotive infotainment systems. It’s important to note that all of the systems, including Siri, were used in fully hands- and eyes-free mode, meaning that they were entirely voice driven. Higher bars are worse and, as you can see, Siri fared very poorly.

The chart above is based on data from a rigorous study published by the AAA Foundation for Traffic Safety in October of 2014, and titled Measuring Cognitive Distraction in the Automobile II: Assessing In-Vehicle Voice-Based Interactive Technologies. I highly recommend downloading the full study if you want to dig into the details, but, in a nutshell, this infographic depicts the levels of in-car cognitive distraction associated with several of the most widely used automotive infotainment systems. It’s important to note that all of the systems, including Siri, were used in fully hands- and eyes-free mode, meaning that they were entirely voice driven. Higher bars are worse and, as you can see, Siri fared very poorly.

Just how bad is Siri? According to the University of Utah researchers who authored the AAA study, even though Siri was configured to work purely hands- and eyes-freethe [cognitive] workload ratings exceeded category 4 on our workload scale—the highest ratings that we have observed for any task  Moreover, there were two crashes in the simulator study when participants used Siri (the only other crash we observed was when participants used the menu-based [touchscreen] systems).

This might seem unbelievable, but the researchers highlighted specific aspects of the Siri experience that contributed to these results:

“Common issues involved inconsistencies in which Siri would produce different responses to seemingly identical commands. In other circumstances, Siri required exact phrases to accomplish specific tasks, and subtle deviations from that phrasing would result in failure … Some participants also reported frustration with Siri’s occasional sarcasm and wit.”

I think that last point captures the core of the problem with Siri—as smart as the system may seem on the surface, it lacks the critical element of contextual awareness. Whereas research has shown that a flesh-and-blood driver and passenger will dynamically modulate the complexity of their conversation based on driving conditions, Siri is as verbose in a car speeding along at 60 mph as it is on a couch doing zero. Both Apple and Google have been using real-time location data pulled from our phones to display traffic conditions in their respective Maps apps for years, so the technological underpinnings for a contextual Siri are already in place—Apple just needs to connect the dots.

Apple has been polling location information from iPhone users for years to support the real-time traffic feature of the Maps app bundled with iOS. This means that, so long as you have the “Routing & Traffic” setting turned on (it’s on by default), your iPhone can determine that you’re in a moving vehicle, and know the road you’re traveling along. As far as I can tell, however, Apple is not using this data to inform the performance of Siri, which is a big miss, as this added layer of context could significantly enhance the utility and usability of Siri while driving.

Apple has been polling location information from iPhone users for years to support the real-time traffic feature of the Maps app bundled with iOS. This means that, so long as you have the “Routing & Traffic” setting turned on (it’s on by default), your iPhone can determine that you’re in a moving vehicle, and know the road you’re traveling along. As far as I can tell, however, Apple is not using this data to inform the performance of Siri, which is a big miss, as this added layer of context could significantly enhance the utility and usability of Siri while driving.

To be fair, there are indications that Apple is limiting some aspects of Siri functionality when an iPhone is paired with an in-car Bluetooth interface. But, in my experience, the company has done nothing to limit Siri’s verbosity while driving, nor have they addressed the system’s penchant for inconsistent responses, as highlighted in the AAA study. I think it’s crucial that Apple prioritize these sorts of contextual enhancements to Siri alongside the company’s ongoing work to improve recognition accuracy.

One other Siri-related feature I’d be lobbying for if I had the keys to Apple’s in-car efforts would be the introduction of an API for third-party developers. This is by no means a new idea—the Apple developer community has been pining for a Siri API since the platform was introduced back in 2011. Sadly, anyone who’s been holding their breath has either passed out by now or completely given up hope, but this needs to happen for Siri to be accepted as a go-to in-car companion. Even a limited API reserved for certain classes of apps, such as audio and messaging, would be a huge step forward because it would enable drivers to use their favorite apps (as opposed to a tiny assortment of apps hand-picked by Apple) while keeping their eyes on the road—something that isn’t possible today with an iPhone.

Addressing these two Siri shortcomings is where I’d apply most of my energies as Apple’s in-car czar, because I believe fixing them would deliver the biggest real world safety gains. Now, I say “most” of my energies because, as noted at the start of this section, my efforts would be focused on two fronts: safety and convenience. And, when it comes to convenience, I’d go all in on improved Bluetooth performance (please, hold your applause … no, really, you’re all too kind).

It’s no secret that Bluetooth has been a bugbear for Apple since Bluetooth has been a thing (if you’ve been under a rock for the past few years, just Google “Apple Bluetooth problem” to get up to speed). And to those who say that the fault lies with the automakers, that fails to explain the explosion in Bluetooth issues that always seems to coincide with major releases of iOS and Mac OS X. The bottom line is that the implementation of Bluetooth on Apple devices today is far from great, and improving it is essential to answering the question of an improved in-car iPhone experience.

Fail Forward
There are many other features Apple could implement to enhance the experience of using an iPhone while driving, such as rich integration with the upcoming Apple Watch and its “Taptic Engine.” But I believe the work areas outlined above must take precedence if the company is serious about delivering a “smarter, safer and more fun way to use iPhone in the car,” which is how CarPlay was described in the press release announcing its launch exactly one year ago.

As it stands today, CarPlay is none of those things. This may explain why, 365 days on from its premiere, exactly one CarPlay-equipped vehicle—the Ferrari FF—is available for purchase (assuming you have roughly $300K to spare). Meanwhile, Toyota, which is the world’s largest automaker and was cited in Apple’s original press release as a CarPlay partner, has done an about-face, telling the New York Times that “it currently had no plans to adopt Android Auto or CarPlay in the United States.” Perhaps engineers at Toyota read the aforementioned AAA study, which found that Toyota’s in-house Entune hands-free system was significantly less distracting than Siri?

This is all conjecture on my part, but these signs suggest that CarPlay is running on fumes before it’s even passed the start line. The one thing that gives me hope for the future of Apple’s in-car efforts is the company’s unusual willingness to kill or fundamentally re-conceptualize products that it deems wanting. One example is the Apple TV, which started life in 2007 as a $299 media center PC with its own 40 or 160 GB hard drive (for those who don’t remember, it looked like a slightly smaller Mac mini), but was then completely re-imagined for 2010 as a $99 streaming media puck. More recently, Apple announced the discontinuation of the company’s long-lived iPhoto and Aperture apps for Mac, with both due to be replaced by an entirely new Photos for Mac OS X app within the next few months.

The “Designed by Apple” manifesto I referenced at the start of this piece includes a line that reads, “There are a thousand ‘no’s’ for every ‘yes,’” and I very much hope Apple has the institutional fortitude to attach one of those “no’s” to CarPlay. I don’t claim that the approach I’ve outlined above is necessarily the answer, but it’s clear that the CarPlay of today is headed in the wrong direction. And Apple needs to get it right because, unlike an Apple TV or Apple Watch, getting it wrong can be a matter of life and death.

Agree? Disagree? Have questions? Let’s continue the conversation on Twitter @edotkim.