Apple SVPs

At first glance, it seems a little odd that Angela Ahrendts’s replacement overseeing the Apple Stores, Deirdre O’Brien, will continue on in her position as chief of human resources. O’Brien, who is currently vice president of People, and is an Apple lifer, will move above the horizontal bar on Apple’s Leadership page and become senior vice president of Retail + People,1 a new title for an Apple SVP. What makes it odd is that Retail + People looks like an enormous portfolio for one person, especially given that those duties used to be split between an SVP and a VP.

But if you look at Apple’s most recent Equal Employment Opportunity filing, you’ll see that Sales Workers make up 31% of Apple’s employee base. Only Professionals, a hair higher at 32%, beats it out, and no other employee category is even close.2 So a lot of Apple HR is already devoted to Retail.

Apple 2017 EEO filing excerpt

The duties of Apple’s top people have often been broad and seemingly unfocused. Eddy Cue, for example, was put in charge of the App Store because he’d been running the iTunes Store. This no doubt seemed reasonable at the time, but it was a disaster. Running the App Store included handling app reviews and developer relations, and Cue’s tenure was one long continual complaint of app reviews taking too long and developers being treated unfairly. Putting the App Store under Phil Schiller, which on paper makes no sense for the SVP of marketing, was the solution, for which both Schiller and Tim Cook deserve credit.

I would argue that broadening Jony Ive’s design oversight to include software in addition to hardware was a mistake as big as putting Cue in charge of the App Store. The software side of Apple’s user interfaces—especially on iOS, which isn’t as hardened by long tradition as on the Mac—has become steadily more cryptic under Ive’s control. Some of this is due to Apple’s need to squeeze more functionality into the OS, but Ive hasn’t been up to the task of melding the new functions into the UI in a consistent and discoverable way.

To me, Ahrendts’s five years in charge of Retail has been similar to Ive’s time as Chief Design Officer. The Apple Stores look better than ever, but they don’t work as well as they used to. No one I know looks forward to going to an Apple Store, even when it’s for the fun task of buying a new toy. No doubt a lot of this is due to Apple’s success and the mobs of people milling about,3 but Ahrendts didn’t solve the problem of efficiently handling the increased customer load.

I hope O’Brien’s background in operations will lead to improvements in the flow of people through the Stores. If so, we’ll look back on her appointment the way we look at Schiller’s installation as head of the App Store, an example of Tim Cook’s slow but successful wrangling of a problem that arose out of Apple’s growth.


  1. It’s very Apple-like to call their HR division People. 

  2. Are all of those Sales Workers in Apple Stores? I don’t know, but I bet it’s close to all of them. And I suspect a fair number of the Technicians in that table are in Apple Stores, too. 

  3. No one goes there anymore, it’s too crowded. 


A Hard Day’s Night

A few weeks ago, I was on a podcast with the folks at The Incomparable, talking about two very different musicals: Gold Diggers of 1933 and A Hard Day’s Night. If you’re an Incomparable member, you had a chance to listen to the raw version right away; if not, the edited version was released for the rest of the world yesterday.

Because the focus of the podcast was on the movies, we didn’t spend as much time on the songs as we could have. That’s probably for the best with Gold Diggers,1 but left me wanting to say more about Hard Day’s Night, the title track in particular.

The story of how it got written is pretty well known. Filming of the movie was going well, the songs that John and Paul had written were of their usual high quality, but the movie didn’t have a title. John suggested A Hard Day’s Night to the producer, Walter Shenson, and both he and director Richard Lester loved it.

The phrase was not original with John. It was something he’d heard Ringo say after an all-night recording session, and he’d tucked it away in his wordplay-loving brain, waiting for a chance to use it. In fact, he had used it in the “Sad Michael” story in In His Own Write, published shortly before filming started.

Sad Michael

The book is shown in the movie, and, famously, John isn’t in some of the cathartic “Can’t Buy Me Love” sequence (the sequence all of MTV was based on) because he was off promoting it. Lester covered this up by donning Beatle boots and filming with a hand-held camera from John’s point of view.

With the title of the movie chosen, Shenson wanted a new song to go with it, something to be played over the title sequence. John wrote it pretty much overnight, something that’s always presented as amazing, but it really isn’t. Music was simply pouring out both John and Paul during those early years.

To me, “A Hard Day’s Night” is just about the peak of this period of Beatle songs. The opening chord, and the mystery of exactly how they got it, has been the basis for hundreds of blog posts and YouTube videos, but I’ve always felt that what makes the song great is how well it uses the different strengths of John and Paul’s voices.

Paul’s part (chorus? middle eight? the Beatles were always playing around with structure) is so perfectly written for his voice that you’d think he wrote it himself, and yet I’ve never seen this listed as anything other than a fully John-written song. I guess it comes from the two of them living in each other’s pockets for the previous six years or so. Anyway, this part absolutely soars and is something John couldn’t have sung with anywhere near the same effect.

And it’s not just the notes, the words fit Paul’s persona, too. Unsurprising, I suppose, as some of them were a direct steal from the mostly Paul-written “Hold Me Tight,” recorded just four or five months earlier. But even the unstolen words—everything seems to be right—are just so Paul.

And John’s part is so John. His profession of love is not unalloyed as Paul’s is; it comes tinged with complaints. Working like a dog to get you money to buy you things. He takes advantage of the stringent, almost whiny quality he can get out of his voice to give the song an edge. I also love the way he uses “work” and “worth” to get an extra rhythm—an entirely verbal rhythm, no relationship to the beat of the song—in the second verse.

Where it all comes together is at the end of Paul’s second chorus. His voice rises up those eight bars to that intense tight… tight, yeah! ending, and on the yeah, John comes in underneath with that groaning cross between oh and uh, bringing the song back down to earth and kicking off a repeat of the first verse.

It’s just perfect.


  1. Apart from “We’re in the Money,” which is a truly great song, especially the seldom-remembered lyrics that put it squarely in the Depression. 


Shut it

I got a Raspberry Pi 3 Model B for Christmas and finally started using it. If “finally” seems like the wrong word, I will clarify: I’m not talking about the Christmas a month ago, I’m talking about the Christmas 13 months ago.

I bought a power supply, case, and microSD card for the Pi right away, but I never got it up and running. I dithered so long because I couldn’t decide what to do with it. My first thought was to set it up to run the Wolfram Language, which can be installed and run for free (for non-commercial use) on the Pi. That’s still kind of appealing, but I’m more interested in learning R and ggplot (to that end, Kieran Healy’s Data Visualization was recent Christmas gift to myself).

Then I thought about using the Pi as a Homebridge server, but that would mean committing to a bunch of smart devices in my house, and although I’m kind of interested in hearing other people’s experiences in home automation, I’m not sure I want to do it myself. I’ve never felt it burdensome to flip a light switch.

I could always set up the Pi as a general-purpose server to do command-line things from my iPad via Prompt. I already have that, though, because both of my iMacs (home and office) run continuously and have the tools I like already installed. Rosemary Orchard has an interesting variation on this: a Pi she travels with as a portable server for times when she doesn’t have good internet access. But for that I’d need a Pi Zero W; mine doesn’t have WiFi.

Update Feb 2, 2019 4:01 PM  Apparently my Pi does have WiFi, even though most of the setup guides I’ve seen for it discuss Ethernet connections only. Even so, if I were going to travel with a Pi, it’d be the tiny Zero W.

A couple of weeks ago, Nathan Alderman was on a Slack channel I visit, talking about installing an internet ad-blocking service called Pi-hole on his Mac mini. He was having some trouble with its configuration because Pi-hole is more easily installed in its native environment, a Raspberry Pi. I’d never heard of Pi-hole before, but it seemed like a perfect fit for my dust-gathering Model B. Last weekend, I spent about an hour and a half setting it up on my home network, and it’s been successfully blocking crap from my devices ever since.

There is an ethical issue to using an ad blocker. People who want to make a living creating things on the internet should be paid for their work, and advertising is the most common way to get paid. And I’m not offended by the presence of advertising on other media. Newspapers, magazines, radio, and TV have always had ads, and I don’t go out of my way to block them. But when I’m looking at a newspaper while eating breakfast, the newspaper doesn’t look back at me, taking note of the cereal I’m eating, the clothes I’m wearing, and whether it looks like I’ve put on a few pounds recently. Internet ads do, and it’s that—along with the invisible trackers on so many sites—that offends me. So I’ve used ad blockers despite my qualms.

Pi-hole differs from browser-based ad blockers in that it works at the network level. It acts as a local DNS server that simply doesn’t serve content from sites on its block list. So the voluminous cruft associated with so many modern websites doesn’t even get to your browser to be sifted through. And a single configuration works for all the devices on your network.

I followed these instructions, which start at ground zero and take you through the entire configuration of both the Raspbian OS and Pi-hole itself. There are a few places in process where you’ll have to do a small bit of thinking on your own:

After configuring Pi-hole, reboot the Pi; you should have a functioning ad blocker.

To test it, I changed the DNS settings on my iMac to point to the IP number of the Pi and tried out several pages. It worked for a while and then I did something I’m still not sure of that killed the WiFi connection. It wasn’t a problem with the Eero, as all my other devices were still connected. I rebooted the iMac, its WiFi connection came back, and I haven’t had any trouble with it since.

With a successful test under my belt, I took the big plunge: changing the DNS settings in my router to point to the IP address of the Pi. That required a reboot of the Eero. When it came back, ads were blocked from all of my devices. And my wife’s, too, which led to some difficulties.

First, she couldn’t follow links from Twitter on either her iPhone or iPad. I assumed this had something to do with the t.co domain that Twitter uses for links in tweets. That turned out to be correct, but what was weird was that I could follow those links from all of my devices. I looked in her settings and mine but couldn’t find any reason for this. Going to the Pi-hole admin page and adding t.co to its whitelist solved the problem.

But that got me thinking that other services she uses might be blocked. The people who set up blocklists for Pi-hole, I figured, are probably very militant and will block more than a normal person will want blocked. I was particularly concerned with Instagram and Facebook; Pi-hole could make them unusable.

So we started with Instagram. My wife went to her timeline and started clicking links to see what worked and what didn’t. Most links worked, but a couple didn’t. I looked at the ones that didn’t and they were definitely the kinds of quasi-advertisements that Pi-hole adherents would want to block. “Yeah,” said my wife, “but my friends post these and I like to see them.”

I realized right then that Pi-hole wasn’t going to work network-wide. We didn’t bother checking links in Facebook or, god forbid, Pinterest. I reset the Eero to go back to using my ISP’s DNS servers, and then set the DNS on all of my devices to point to the Pi. This isn’t what the Pi-hole folks intended, but it works perfectly for us. I get the blocking I want, and my wife doesn’t get the blocking she doesn’t want.

In the week since setting up Pi-hole, I’ve had to add only one other domain to the whitelist: bit.ly. I guess I understand why a service that hides what you’re clicking on could seem nefarious, but that’s too severe even for me.

Here’s what the Pi-hole admin dashboard looks like:

Pi-hole admin page

Early on, I added a recommended set of blocklists, thinking it would overwrite the defaults. Instead, it added to them, and I think there are a bunch of duplicates in that 1.6 million figure in the red box. Now that I understand the system a little better, I’ll probably delete all the blocklists and start over with a fresh set.

Overall, I’m pleased with Pi-hole. It was easy to get started and has given me no trouble since I finalized its setup. I’m not sure I’d buy a Pi specifically for it, but it’s a nice service to add if you already have one.


New Apple graphs

It’s a month after the end of a fiscal quarter and time for another round of graphs showing Apple’s performance. As always, you can get a full meal of graphs from Six Colors and MacStories. Here, you’ll be served an amuse-bouche.

The big difference this quarter is that Apple’s report no longer contains unit sales for its devices. As it’s unit sales that I’ve been graphing for the past few years, I had to either give up or figure out a new set of graphs to produce. I chose the latter, but I’m not sure if I’ll keep doing this every quarter. Generally speaking, I don’t find the money Apple makes as interesting as the devices it has out in circulation. It’s the devices that change the world, not Apple’s bottom line. On the other hand, revenue is certainly related to device sales, and moving from unit sales to revenue allows me to talk about services, which are becoming a larger part of the Apple story.

Whatever I decide to do in the future, here’s how I looked at the data for Apple’s holiday quarter. The first thing to know is that Apple breaks down its revenue into five product categories:

The first three are our old friends, the fourth is a lumping together of all the other hardware Apple sells—Watches, AirPods, HomePods, and so on—and the fourth is the category Tim Cook flogs every chance he gets.

It doesn’t take much analysis to realize that none of the latter four categories comes anywhere close to the iPhone as a revenue generator.1 The iPhone accounts for about 60% of Apple’s income, so I decided the best way to start the presentation of the quarterly results was to aggregate the latter four categories and display that alongside the iPhone.

Apple revenue

There’s a lot in here, so let’s go through it piece by piece.

First, Apple’s fiscal year starts at the end of September of the previous calendar year, and I think it’s easier—especially if you want to compare Apple’s performance with the release of products or the introduction of new services—to deal with dates in the regular calendar we all use. So I plot things by calendar year.

Second, I got into this Apple graphing business because I found it difficult to interpret graphs of raw quarterly data. Most of Apple’s sales are so strongly seasonal that it’s hard to see trends unless the data are smoothed. I chose a simple moving average smoother; the smoothed data are calculated by adding the current quarter’s revenue and those of the previous three quarters and dividing that sum by four. The thick solid lines shown in the graph are these moving averages.

Next, the raw revenue figures are plotted as points. In the graph, these are smaller and somewhat muted in color so they don’t detract from the moving averages. I’ve plotted them as dots at the middle of the quarter to which they apply and have connected the dots of the most recent quarter in previous years with a light dashed line. This gives a sense of the year-over-year change.

Finally, Apple just made a change to the way it assigns revenue to the five categories, and it’s published an explanation that also recalculates the revenue in those categories for the four previous quarters.2 Here’s what Apple says:

Starting in 2019, in connection with the adoption of the new revenue accounting standard, Apple will classify the amortization of the deferred value of Maps, Siri and free iCloud services, which are bundled in the sales price of iPhone, iPad, Mac and certain other products, in Services net sales. Historically, Apple classified the amortization of these amounts in Product net sales consistent with its management reporting framework. As a result, the 2018 net sales information has been reclassified to conform to the 2019 presentation.

In a nutshell, by reclassifying some portion of the sale price of its devices as going toward Maps, Siri, and iCloud, Apple has boosted its services revenue by about $640 million per quarter. As a consequence, it has reduced the quarterly revenues for the iPhone ($450 million), the Mac ($70 million), the iPad ($110 million), and the other devices ($10 million).

Thus, comparing revenue figures before the 2017 holiday quarter to figures after that quarter is not an apt3 comparison. The Six Colors crew have done their best to account for this change by estimating how Apple would have presented its revenue figures using the new system for all of the previous quarters, not just the most recent year’s. I decided to avoid the perils of estimation by using shading to partition the graph into two sections, one for the old accounting method and one for the new. The reader is alerted to the change and can make their own judgments about comparisons across the section boundary.

On the scale of the graph above, the adjustment from the old method to the new would be barely noticeable. $450 million is less than one-quarter of the distance between the minor tick marks on the vertical axis, or about the diameter of one of the raw revenue dots. No big deal.

What is a big deal is the downward kink in iPhone revenue caused by the most recent quarter. It’s not just that Apple missed the estimate it made three months ago; it’s not just that iPhone revenue is down from the year-ago holiday quarter; it’s that iPhone revenue is down from the holiday quarter of two years ago. Unsurprisingly, Apple has been focusing on its overall revenue, which although down, is buoyed by the continuing upward trend of the non-iPhone revenue.

And speaking of non-iPhone revenue, let’s split that out into its four components. This makes for a nice graph, because they are all of comparable scale.

Non-iPhone revenue

On this scale, the $640 million that Apple has robbed from Peter to pay Paul is not insignificant. The Services revenue in the unshaded area would be noticeably below what I’ve plotted if we were still using Apple’s old accounting method. Still, Services is rightly considered Apple’s biggest growth category, even though it is growing at a slower rate than it was a few quarters ago.

Down further on the graph, we see that the iPad is about to be overtaken by Other, an ignominious position for what used to be the top dog of the non-iPhones. In fact, if we were to look at the raw revenue figures instead of the moving averages, we’d see that the iPad has come in below Other for the past two quarters.

But we can’t look at the raw figures because I didn’t plot them. Why not? Because doing so makes a mess.

Non-iPhone revenue mess

It’s not impossible to interpret this graph, but who’d want to? Plotting two aspects of four data sets that intersect one another is just too much at once. If we want to look at the raw data of these, it’s best to plot no more than two on the same graph.

You certainly can’t explain Apple in two graphs, but I think these do a good job of presenting an overview of the company’s progressions and regressions from the product side. One unaddressed problem with them is that they don’t account for inflation—something I didn’t have to worry about when I was plotting unit sales. Apple doesn’t account for inflation either, of course, but that doesn’t mean I shouldn’t. If I’m going to keep doing this, I’ll have to decide on an inflation index and a basis year.


  1. The same was true with unit sales, which is why I typically broke out the Mac and iPad sales into their own graphs. 

  2. Big thanks to Jason Snell for pointing me to this document. I could see in the consolidated financial statement that the previous year’s revenue figures had changed from the values reported last year, but I didn’t understand the changes until I saw the recalculations. 

  3. I will not use the phrase “apples to apples.” Even I have some dignity.