BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

New iPhone Predictions Are Precise, But Are They Accurate?

This article is more than 10 years old.

I've been reading Nate Silver's book, The Signal and the Noise, and he makes an interesting distinction relevant to the art of iPhone prediction. Silver, who one friend of mine refers to as the Paul Krugman of statisticians, may not be a candidate for the Nobel Prize, but certainly has raised public awareness of things probabilistic more than anyone. I realized from his descriptions of the enquiring frame of mind required to sift the signal from the noise in Presidential politics or climate science that I, too, was in the prediction business.

Following Apple is a form of informed speculation, liberated, in a sense, from certain journalistic conventions because Cupertino only rarely comments to the press on its future plans. If under Steve Jobs this led to punitive secrecy and spectacular reveals, then Tim Cook wields a considerably leakier ladle. The number of leaked parts and production line photos in this iPhone cycle has been unprecedented, but to what extent unscrupulous? How much is signal and how much is noise?

The distinction I mention in the lead, is between precision and accuracy, well-illustrated in a diagram in Silver's book. The diagram shows targets with concentric circles. The first target shows a spread of bullet holes nowhere near the center with the legend, "not accurate, not precise;" the second shows a spread clustered closer to the center with the legend, "accurate, not precise;" the third shows four shots close together, but not near the center with the legend, "not accurate, precise;" and the final shows four shots, dead center with the legend, "accurate, precise." The fact that there have been so many leaks and that those leaks have described very consistent attributes of the new iPhones, suggests accuracy but only proves precision. We may be in the third target, but not the fourth. In fact, a majority of the recent leaked iPhone parts have come from a single source, a young tech blogger in Melbourne named Sonny Dickson with some sort of connections to the shadowy world of Chinese manufacturing.

The problem, as journalists know, with single sources, is that they can be wrong. For all we know, Sonny Dickson is being fed a very precise diet of inaccurate spoof iPhone parts. The same Chinese factories that churn out actual iPhone parts also make cheap knockoffs. If Dickson's source(s) should turn out to be turning out fictitious iPhone parts, we would see how highly leveraged the world of Apple coverage has become. At the height of the financial crisis, Silver reminds us, the speculative markets had placed $50 in side bets for every $1 of actual real estate investment. So although each citing of a Sonny Dickson leak is vetted (to different degrees) by a host of news organizations and bloggers, these citations are all highly correlated with Dickson's information. This correlation is analogous to how risky mortgages turned out to be dependent on one another and their simultaneous failure led to the burst of the housing bubble. If there are 50 blog posts to every Dickson leak (I think there are probably more!) these don't actually represent 50 data points with independently verifiable veracity, but rather fifty leveraged bets on the same piece of unverifiable data.

And the business of prototyping future iPhones, as 3D designer Martin Hajek well knows, is intimately tied to the manufacturers of iPhone cases and accessories. Hajek not only publishes beautiful, highly detailed images of iPhones-to-be, but shares the NURBS (Non-uniform rational B-spline) models for others to use, as well, for making such cases or  other "leaked" images. His fake of the new gold iPhone was so good (with its faux depth of field effect) that the Chinese site Weibo.com took it to be authentic (see below).

The important thing to me, as an Apple watcher, is to constantly measure the accuracy of the sources against what they imply about Apple's design decisions. Ultimately I bring other data sources to bear on my assessment. As a designer, I ask myself, "What would Jony do?" if I see something that seems an outlier. A couple of recent cases made me ask just that question.

The first is illustrated by the array of three possible color choices for the iPhone 5S following the leak of a gold backplate, below. This leak was then followed by a graphite and black 5S combination, which led me to deduce that "The Graphite iPhone 5S Likely Is The Black Option, Not A Fourth Color." New black and white backplates have not been leaked, so everyone's assumption has been that they would stay the same and the gold would be added as a third option. There is a pleasant duality to the original black and white models, but once the gold is added to the set, the black ends up looking harsh and brutal to me. One thing we learned from the iOS 7 beta release is how much sway the marketing department has in Ive's design decisions, and the disharmony of this proposed triad didn't look right to me. It could be that the new graphite option bridges the gap between black and silver, but it doesn't seem like enough of a distinction to be meaningful. And it would push the 5S to four distinct color choices.

I could be wrong here, of course. Silver quotes John Maynard Keynes as saying, "When the facts change, I change my mind. What do you do, sir?" In that sense, I am a Keynesian! But most important to Silver's method is that you have a model of the world that you can test new information against. In this sense you can be opinionated and accurate as long as you understand your opinions to be among the variable you are testing!

The second instance is more of a challenge. In Silver's terms I definitely lean towards the fox, but we all have the capacity to be hedgehogs. The leaked photo below is a purported shot of iPhone 5C packaging but could just as well be a pile of knockoffs. The packaging itself befits the plastic iPhone, but this is the only instance of a leak showing a pink casing. Could this be a badly lit (or badly color-corrected) version of the reddish melon color demonstrated by Sonny Dickson's leaks of physical backplates? Is it a sign of a new color? Or is it a fake? Further, there's the question of the faceplate for the 5C. Will it be white, as has been widely reported, or black as shown here and in some other recent leaks?

This would be a great case for some Silver-style analysis of data sources weighted by past acuracy, but the data in question is highly unstructured and the sources are really just noisy amplifiers of someone else's signal. Will be three or four 5S color choices? Will be six 5C color choices, as shown in Hajek's rendering below, or could there be white and black faceplate versions, bringing the total as high as 12? Stocking Apple stores with even three 5S colors times three memory capacities and six 5C colors times (perhaps) two memory capacities is already quite a lot of SKUs (27, but who's counting?) Would Apple really double the number of 5C models by having a choice of faceplate? Or do that in lieu of a choice of memory capacity? The permutations are multiplying wildly!

There were some earlier suggestions that the 5C would not come in black, but that seems odd considering that black is the most popular iPhone color in general. Many guys will simply not go for such "fruity" colors!

It's easy to imagine the faceplate of the black 5C being black, but the black face with the bright backs feels jarring and very un-Apple-like. As someone attempting to read the tea leaves (or apple cores) and make a reasonable prediction about what we will see next week, the faceplate question is really a conundrum. Am I being a hedgehog and holding to an idea about Apple's aesthetics in the face of material evidence to the contrary? Or is Apple's amazing historical consistency of aesthetics an important form of contrasting data itself? I prefer to think the latter and (for now) give the highest probability to there being three 5S colors (sliver, graphite and champagne) and six 5C colors (black, white, melon, cyan, lime and banana.) And I will venture that the faceplates for the bright colors will most likely be white—because they will look better in the advertisements that way.

– – – – – – – – – – – – – – – – – – – –

To keep up with Quantum of Content, please subscribe to my updates on Facebook, follow me on Twitter and App.net or add me on Google+.