Curves don’t lie: the real story of Apple Watch sales
Last week, analytics firm Slice Intelligence released some charts that caused headlines that Apple Watch sales were collapsing to well under 5,000 watches per day. I posted the article on Facebook, generating nearly 50 comments. One of which linked to an Apple Insider item, and asked: “Duncan, what’s your take on yet another article refuting the claims of poor Apple Watch sales?”
The Apple Insider piece will thrill the heart of any data geek: it is 2,000 words of closely reasoned argument buttressed by numerous charts. But it was NOT a refutation – it failed to present new data of its own that showed the Slice data to be false. Instead, it did an amazing job of pointing out the multiple problems with relying too heavily on the Slice information.
Slice already admitted many of these limitations: their data is based only on eReceipts from a relatively small sample population, it is US-only, online only, and excludes resale sites like eBay. After I finished reading the Apple Insider article, I agreed that the deficiencies are so material that no one should rely on the Slice data to try to predict Apple Watch sales. You CANNOT say “Apple only sold ~3,000 watches on July 2, and there are 365 days per year, so the Watch run rate is only about one million units.” You just can’t do it! But I wasn’t trying to. I actually have no view on whether Apple will sell 5 million or 50 million watches in 2015. Instead, what I am really interested in is “what kind of buyer is buying the Apple Watch?” There are two camps out there today.
Camp One says that all previous smartwatches have been mediocre successes at best, kind of like tablets were before 2010 and the iPad. And, like the iPad, the Apple Watch is the smartwatch done right. The iPad version one wasn’t perfect, but people bought it, liked it, showed it to their friends, version two was better, and Apple now sells over 60 million iPads per year, and the tablet market as a whole is 230 million units in 2015. In the same way, the Apple Watch will be big for Apple, but also transform the wearable space and turn it into the Next Big Thing.
Camp Two agrees with most of that, but worries that the time is not yet right for wearables, especially watches. Maybe one day they will be big, but in 2015 the mass market of consumers isn’t interested. Further, since Apple does produce generally great products, has a very loyal customer base, and has enjoyed very favourable media interest (the narrative of “Apple will do to watches what they did to tablets” is pretty hard to resist!) means that people in BOTH Camp One and Camp Two agreed that the opening weeks of Apple Watch sales were likely to be large, and the Apple Watch would easily become the most successful smartwatch launch ever. But where Camp Two diverges is that although initial sales will be big, relatively quickly we will see sales drop sharply.
Although we’ve been talking about watches and tablets so far, the phenomenon at work is common to technology devices, movies, music, games, and so on. The entire movie business is built around two numbers: how big is your opening weekend box office…and what happens after that? Whenever the latest Marvel Universe movie (or Woody Allen film, or Tom Cruise vehicle, or James Bond, etc.) opens, the studio can count on a certain number of millions of super-fans to pack the multiplexes’ biggest screening rooms. But some movies fall off a cliff in the second weekend, while others might never match the opening day success, but still generate sufficient box office interest to stay in theatres for weeks or even months. Whether long term success, staying power, or “it has legs”, they all mean the same thing…and they all look the same on a graph.
And here is where I think the Apple Insider article doesn’t come close to “refuting” the Slice data. Although the Slice methodology has multiple weaknesses and deficiencies, and is possibly useless for calculating annual sales – it is CONSISTENT. To coin a phrase, the chart at the top is “comparing apples to apples.” The absolute numbers shouldn’t be relied upon, but I will tell you that I barely noticed them when I first saw the shape of the curve.
The chart was properly done and reminded me of other charts I have seen: the time period across the bottom with a uniform interval; there was a seven day rolling average to eliminate one day glitches (see the daily chart below to see what the unsmoothed version looks like); and (most importantly) it was log scaled. Anyone who follows media, technology, or any of the natural sciences could look at this graph and say “this is a hyperbolic decay curve. It follows a reverse S-curve shape. There is a big drop at the beginning, then a plateau, then another drop. The two most important things this chart tells me comes from a) the duration of that plateau; and b) the steepness (or slope) of the second drop.”
Once again, I ignore the absolute value of the numbers on the right hand axis. Instead, the chart nearly screams the following: After initial enormous sales on launch, daily transactions fell about 90% and stabilised in mid-April. They stayed highly range bound for seven (7) weeks, until the second week of June, where they declined another ~90% over a three week period.
There are two interesting things about the shape of the chart. The first is that Camp Two is more likely to be correct. Regardless of how many millions sold, the first generation Apple Watch seems to have very strong launch sales, a relatively brief but strong plateau, and (once the early adopters are finished buying) the second decline is steep and to a low level. The product doesn’t appear to be popular outside the early adopter crowd, and not many people appear to be looking at Watches on the early adopter wrists and saying “Man, I need to get me one of those.” If true, this is NOT like the iPad history, where you could see the number of devices ‘in the wild’ growing pretty steadily over time.
The second thing is maybe even more important. I have no skin in the game, and it doesn’t matter to me whether Camp One or Two is right. But as a long-time data guy, the Slice chart “looks right.” The shape of the Slice curve isn’t “close” to what Camp Two would have predicted…it is almost an exact match. The robustness of that fit, both to predicted curves and to other decline curves we see in tech and media, makes me believe the Slice data is useful at some level, and likely reliable.
New data might change my mind, but for now the “Watch is mainly about early adopters” looks to be the more probable hypothesis.