Unsucking Apple’s Mapping Data

Yesterday, I linked to two articles expressing similar frustration with Apple’s apparently-stalled mapping data. This is in addition to my own commentary:

… pressing and ongoing complaints remain in iOS 7′s Maps application. Most glaring is the persistent inaccuracy of significant amounts of data. Large cities remain poorly mapped, particularly those not in the United States or Canada. Local amenities are poorly plotted or non-existent, even for cities with generally good road data.

Apple’s online services remain their weakest offerings: iCloud is temperamental, Siri can be slow, and Maps contains missing, poor, inaccurate, and incomplete data.

For several reasons, I didn’t expect Apple to talk about the improvements they may have made to Maps in the past year at either of the iOS 7 press events: it’s really unsexy to talk about data acquisition and correction in a public environment, and it’s hard to quantify any improvements they may have made. But it turns out that the reason they didn’t promote this is because the changes they made — if they have indeed made improvements — appear to be insignificant.

The world is enormous, and it’s filled with millions of individual data points: roads, towns, coffee shops, grocery stores, train stations, and so on. Even if every engineer were working solely on Maps for an entire year, I doubt that they’d be able to bring its data up to the quality of Google’s.¹ The reason why that is, though, is predicated on a little bit of informed speculation.

I had originally intended to include this post within the context of my iOS 7 review, as I think it’s a relevant topic. I ended up leaving this out because I felt it was important not to delve into this sort of speculative guesswork in my review. But now? Sure. Carpe diem, ‘n’ shit.

From what I understand, Apple’s internal version of Radar, their bug tracker, sort of works like a giant leaderboard. Any time a bug is duplicated, it gets a “vote” — a dupe. The more dupes on a bug — in combination with its priority ranking and severity — determines the amount of time and effort engineers will spend fixing that bug:

• Is there a gaping security hole in the iTunes Store? That’s a high-priority bug that Apple will try to fix quickly, regardless of the number of reports.
• Did you dupe that weird bug where Safari’s address bar sometimes activates before you press return? If there are a lot of reports, Apple’s going to prioritize it.
• Did you report a spelling mistake in Mail’s help index? That’s going to get addressed at 4:30 PM on a Friday. Maybe.

I anticipate — and this is the speculative part — that Apple’s internal Maps error tool works in a similar way, with problems addressed according to the number of people reporting them. Is Times Square misplaced in New York? That’s probably going to get a lot of reports, so it will be fixed quickly. Is the marker for a convenience store in Milk River, Alberta (population 811) across the street from where it should be? That’s going to get fixed in 2017. Maybe.

The problem with this system is inherent to users reporting errors: since the absolute least expensive way of running Maps costs $229, there is a significant barrier to entry for certain groups of people, especially those in developing nations. As a consequence, reports of incorrect data in these regions will be fewer, and they likely won’t be addressed as quickly. (Google Maps is, obviously, free, as long as you have an internet connection.)

If my thesis were true, areas with higher concentrations of Apple products should have better data than areas that do not. Unfortunately, Apple doesn’t break down their sales data to the necessary granularity, so to assess this, I need to use information from a group of people I dislike: analysts.

On September 4, Kantar Worldpanel released a report stating that the US smartphone market share of iOS is a little over 43%. In a separate report, the same company claims that 70% of smartphone sales in much of Europe and China were Android-based, and “one in ten” were Windows Phones; according to this data, iPhones must therefore represent approximately 20% of smartphone sales in those regions.²

My thesis seems to be largely corroborated by users in those respective regions in a Branch thread:

Jason Snell: I’ve used them for driving directions a couple of times and in general use around San Francisco, and it’s been great.

Andy Newman: It’s been excellent for me in Columbus, Ohio. The turn-by-turn directions are a welcomed addition.

Garrett Murray: Here in LA the POIs have been relatively accurate for my purposes. I’ve used Maps to find a few restaurants (all of which were correctly reflected), but noticed it has a harder time being accurate with generic searches.

On the other hand…

Federico Viticci: It gets worse with POI names. In my town, Viterbo (a relatively small one compared to Rome or Milan, for sure, still one with its own province), I find Apple Maps to have outdated business listings or simply missing stuff.

And my own take, from my review:

For my use, Maps is decent most of the time, and only occasionally maddening. Calgary’s data is generally accurate, but it’s on the sparse side. Some locations remain stubbornly misplaced despite myself and people I know reporting these issues.

There are a couple of anomalies in that Branch thread: Mark Gurman — who lives in Michigan — found Maps lacking, while Karan Varindani — who lives in Ghana — found it better than Google’s. But based on all the reports, tweets, and blog posts I’ve seen, Apple’s mapping data is much stronger in areas with higher iOS device ownership. Therefore, a reasonable question might be “how can Apple improve their data for regions where their products aren’t as popular?”

One way is to acquire data from established services, either by licensing it, or by buying the entire company. Apple is already doing this: amongst others, they bought Hopstop and Embark, and are rumoured to be in talks with Foursquare for a licensing deal. In addition to Foursquare, several of Apple’s confirmed deals are cross-platform, which means a broader potential user base from which to collect data.

Another way might be to compete with Google on the web. I have no indication that Apple is even considering this, but it would be a reasonable way for them to get more people to report more errors in their data. Of course, Apple uses their hardware sales to provide software to their users at a lower cost, or even for free. Providing non-buyers with access to a product which requries a not-insignificant amount of bandwidth, plus a bunch of engineers and QA staff to attend to problem reports is not cheap. But it is perhaps a reasonable way to gain users in places where Apple products are not as popular.

I’m sure there are loads of other ways to get better data. Apple’s putting people on the ground, and they’re beefing up their data through acquisitions of key companies. But I have no doubt that this is frustrating for users, like Michael Tsai:

Back with iOS 5, it was possible to ask Siri or tap on a contact’s address and have it open up in a top-quality map. This is no longer possible today, because although Google Maps is in the App Store, all the OS services are hard-coded to use Apple Maps.

No map is perfect. Apple went through a giant executive reorganization last year, resulting in a version of iOS which has been redesigned and largely rebuilt. But it’s hard to look back on the past year of that progress and not wish that some of that effort had gone into Maps. It’s a beautiful product, but it’s not done yet, and that’s hard to reconcile with Apple’s penchant for releasing great stuff.