Show Me the Numbers: Serial's Data Transfer Costs

Serial is the most accessed podcast ever from iTunes, according to Apple. By November 18, it was downloaded and streamed 5 million times. The show claims some 1.5 million listeners per episode, of which nine have so far been produced. That would mean nearly 9 million downloads or streaming sessions (assuming people went back to listen to the whole thing) from non-iTunes sources, which seems high, but would also indicate a better distribution of means by which people obtain podcasts, which is good for all podcasters!

David Carr, the lead media reporter at the New York Times, wrote that the episodes were downloaded "at a cost of nothing," which may refer to what it costs to deliver or what listeners pay; hard to tell. But I'd like to guess at the amount. What does it cost to deliver that many episodes?

Let's take the notion for simplicity that roughly 13.5 million downloads or equivalent streams occurred evenly over three months, or 4.5 million downloads a month. Episode 9 is typical and roughly 30 megabytes (MB). That's 135 terabytes (TB) per month. (Yes, some months would be more and others less, but still good for estimating.)

Via Amazon S3, Serial would have paid $12,000 a month or $36,000 so far. Amazon charges on usage, not on a monthly basis. (It charges for storage on a monthly basis, but all the podcast files together aren't even half a gigabyte.)

But, as my information technology friends tell me, that's way too much to pay; instead, Serial is using a content distribution network (CDN), which is designed to take media files and feed them out a bazillion times more cheaply and efficiently. Serial's CDN, Highwinds, doesn't publish its rates and any CDNs only offer private estimates, but MaxCDN has a rate schedule. Serial would pick the 150 TB per month plan, which runs $6,144 per month plus 4¢ a GB over 150 TB ($40 per extra TB). That would be over $18,000 so far. CacheFly has a bandwidth calculator, and reckons a bit over $3,700 per month for 135 TB, or about $11,000 so far.

If Serial has cut an excellent deal, piggybacking as one expects on This American Life's downloads, it's probably paying the least possible, and that sub-$4,000 per month figure seems accurate. But in public radio, that's the same cost as part or all of a full-time entry-level-or-above position. If the show becomes more popular, the costs go up as well, where conventional radio distribution has a very high fixed cost and none of these sorts of high variable costs for extremely popular programs. Some podcast and audio hosting sites, like Libsyn and SoundCloud, absorb some or all of the bandwidth costs — but they're still paying the piper, even if they bill $0 to the podcast producers.

Over time, the price of data transfer has dropped relatively quickly, but it doesn't plummet nearly as fast as hard drives or hosted storage. In 2006, Amazon charged 16¢ per GB for downstream transfer (its servers to the Internet); in 2014, it's 33% lower, or 12¢ per GB. In the same time, hard drive storage dropped from 60¢ per GB to 3¢, or a 95% drop.

Increasingly successful podcasts will need to budget serious sums that, as listenership grows and prices slowly drop, might stay constant for a while, and be a significant line item in the budget.

A T-Shirt Celebrating The Magazine

With our friends at Cotton Bureau, The Magazine is offering a limited-time-availability T-shirt to commemorate our 28-month run, which ends next month. The color is from Issue #1. The back shows our three-diamond "end of story" icon and our run date.

This shirt is an American Apparel Tri-Blend Tri-Black with long-lasting ink — I've got others from Cotton Bureau using this method, and they remain vibrant and stand up to many, many washings.

 

Vital Reads

There's a book I bring up all the time in talking to people about innovation, creativity, and spaces. It's Stewart Brand's How Buildings Learn. I have probably mentioned in it 10 episodes of The New Disruptors because it's so relevant to people getting themselves off the ground, often using odd or temporary spaces.

While Brand looks across a whole swath of issues relating to architecture and the changes that occur to buildings over time (even ones that are claimed to be historically fixed, and are not), the part I like to point to is the importance of space that nobody cares about.

That is, space in which you can experiment, drill holes in the floor, knock out walls, install plumbing, paint walls and repaint them, build and restructure offices. This is usually in old buildings or facilities that people have decided not to update — or are even destined for demolition.

The more you have to care about the space you're in and the less flexible it is, the more it constrains what you do and how you collaborate. (This applies to content-management systems or CMSes, about which I've talked quite a bit, too.)

I have to add to this book one I've just read by Ed Catmull, one of the heads of Pixar, and its driving force from the beginning. He, Steve Jobs, and John Lasseter developed it into the powerhouse combination of research lab and movie studio that has churned out a series of successes. Catmull and Lasseter took what they learned to Disney's animation arm after Pixar was acquired by Disney and turned it around as well.

Creativity, Inc., isn't about success. It's about failure and managing failure. Catmull unsparingly describes the many hundreds of times he and his colleagues made errors in judgment often based on continuing something that had proven successful, even when they thought they were already taking into account the bias of success. It's revealing from a man running a company that, to outward appearances, has done nearly everything right.

It's a good read about the history of computer animation and the studio, but it's packed with lessons that are applicable to an army of one (like yours truly) or a large corporation.

We had a very nice, long conversation on The Incomparable podcast as a book-club episode about what this book has to teach.


Approaching Halfway with Kickstarter

The Magazine: The Book is nearly at 50% of the goal we need to make it happen.

Kickstarter campaigns can follow a few arcs.

They can flatline, which is about 20% of them, last time I was able to get statistics. 20% of all projects approved by the company get no bids. Another 20% get less than one-fifth of the way to their goal amount. 16% of all projects fail between about 20% and 50% of the total amount they plan to raise.

But at the halfway mark, when you raised 50% of your total, the odds are pretty dramatic: 97% of Kickstarter projects that fund halfway proceed to fund fully by the end of the campaign.

We're about 48% of the way to our total, and I'm confident that, as we hit our last 12 days, we'll start to see some steam as people both see that it's coming to a close and it's not fully funded. It's an exciting thing, and daunting, and nail-biting. But we'll get there.

The campaign is to make a hardcover book of some of the wonderful stories that reporters and essayists wrote between October 2012 and October 2013 (and an ebook as well). The reward for pledging is the book! (And more.)

The campaign covers all the costs of paying for design, printing, shipping, and contributors, and will leave us with some print books left to sell and the ebook to offer online. It's a great way to run a project like this: to scale production with actual demand.

I'll admit that it's scary to sit here with about 33% of the time and 50% of the money left to go — but I also feel strongly about the stories in the book, the design that's being created for it, and the interest in making something cool and new that people will enjoy. Thanks for those of you who have backed the project already, and I hope those who haven't will consider jumping in to get a copy as soon as it's hot off the presses!

Multi-Modal Economics on Recent Trip

For my trip to Portland for the XOXO event, I opted to go by train and use car2go here in Seattle and down in Portland. The train was $100 with tax round-trip including business class each way. The train has Wi-Fi, which is halfway reliable in the more populated areas of the route. (Business class on the 3 1/2-hour route gets you a nicer seat, including a single when you're traveling alone, and a power outlet, as well as a $2 discount on the club car, which has fairly decent food.) 

car2go is a one-way car-sharing service, unlike ZipCar, which requires that you return the car to its allotted parking space at the point of origin. car2go has negotiated with cities to pay them in lieu of lost parking fees. The car has to be returned within the home zone, which can be as large as an entire city or metro area, but not to any particular place within it.

Depending on the city, you can park nearly anywhere there is no time-of-day "no parking" restriction. In Portland, everywhere; in Seattle, it has to be a two-hour or longer or unrestricted spot. You don't feed a meter, either; that's included. The cost in Seattle and Portland is 38¢ per minute up to a cap of $14 per hour or $72 a day. Seattle adds 20%  tax! Portland adds none.

While it's only 180 miles to Portland, and it should take about three hours, my trips invariably take four or longer. There's always an accident or slowdown. I need to stop and stretch my legs. The commuter train is about 3 1/2 hours; it's a bit longer for the long-haul Amtrak runs that span the whole Pacific coast. 

Not only do I have to deal with the tedium of being at the wheel, it carries a cost beyond gas. The IRS currently allows 60¢ per mile reimbursement. For the 360-mile roundtrip plus probably 50 miles of driving within Portland, I would have paid or incurred a total of $260 in gas and wear and tear.

I checked on rental cars to combine train and a local car: $250 was the best I found for renting a car anywhere near the train station. Plus, I would have had to pay the high cost of parking in downtown Portland, which rivals Seattle. Parking is tight, thus the cost. 

I stayed with my brother-in-law and his wife, and their neighborhood always seemed to have a few cars within a half-mile to one mile away. It was good to need to walk, too. A few nights, I'd return late and park in front of their house, and the car would be there in the morning to carry me off.

Between using a car2go from home to the Seattle train station and back on my return (several were within two blocks of the station), and a couple dozen short-distance uses in Portland, the grand total was about $104: no feeding meters, no worrying about finding a long-term spot, and several miles of good walking, too. I think it was the right call. 

 (car2go ain't perfect. A car ostensibly within a few blocks of my house wasn't present, and I had to dash to get to a cluster a half-mile further away, which disappeared from the app display after I found the missing car was — missing. I couldn't sort out the weird trunk opening mechanism in a rush, and had to call to get the firm to deal with it when I parked near the train station. [I had made sure everything was closed, and the car insisted it wasn't.] They took care of it. But I made the train and everything after that was smooth sailing.)

Explaining Apple's Incremental Approach

Many people have asked me since Apple's Tuesday announcement: Where was the sex? Where was the sizzle? Where was Apple's disruption? Where was the "one more thing"?

My reply is that Apple makes its living through punctuated equilibrium[1], not through disruption. Revolutions are hard; small but significant improvements are far easier. The all-in-one iMac, the MacBook Air, the iPod, the iPhone, and iPad all changed the way in which the entire industry created similar products.

Those were released at years-long intervals, not every year. On the Mac side, the ones I think of as having the greatest long-term impact on both Apple and the rest of the computer world:

  • The iMac (1998) had USB and no floppy drive. Clearly, it would fail, because people needed a built-in floppy and there were practically no peripherals that used USB. ADB (Apple Desktop Bus) was the standard.
  • The MacBook Air (2008) with its sealed battery and small, expensive internal SSD, relatively high price, non-upgradable RAM, and ability to be thrown out unintentionally with the trash, was a non-starter.
  • The 2013 (or 2014?) Mac Pro is a black trashcan without internal drive or card slots.
  • (Okay, I'll give you the G4 Cube [2000], even though I owned one, loved it, and used it for years. It had a lot of compromises, and the Mac mini [2005] was the "fixed" version.)

What happened in each of these cases, you may recall.

Floppy drives hung on, but USB quickly became much more widely adopted because peripheral makers created stuff for Macs that could also be used with Windows systems. All-in-one designs became de rigueur. The MacBook Air, after being ridiculed and after necessary improvements in various features, became the model for "ultrabooks," a category into which Intel poured hundreds of millions of dollars to help PC makers produce their own versions with varying levels of success.

On the gadget/mobile side: 

  • The iPod (2001) lacked user-changeable batteries, a low-power FM transmitter, and support for memory cards. Way too expensive.
  • The iPhone (2007) was a toy with its touchscreen, already tried by Nokia, and a lack of a physical keyboard. It only had 2.5G (EDGE) networking instead of the 3G that was available in more advanced phones at the time. Way too expensive.
  • The iPad (2010)  is just a giant iPod touch. At least it wasn't too expensive, but nobody is going to buy one because it's not a netbook.
  • (The iPad mini (2012) has had an impact that's yet to be seen fully, because Apple's competitors already had slightly smaller between-phone-and-tablet devices.  They've sold a lot of them, and it seems more about conserving customers and keeping their money going to Apple rather than creating a new market.)

In the device world, the iPod slaughtered all competitors, and remained ascendent until Apple started killing its own babies. The iPhone destroyed market leaders BlackBerry and Microsoft in America and Nokia worldwide. All smartphones had to become iPhones, ultimately. The iPad killed the low-performance, low-margin, compromised netbook market that everyone had urged Apple to get into, and ate the heart out of new PC sales, while forcing competitors into developing tablets, long seen as toxic after Microsoft's earlier failures.

Apple also managed to keep margins high for all these products. It didn't discount anything. The iPad came to market cheap relative to competitors just before and even long after because of Tim Cook's excellent supply-chain advance purchases that cornered the market on things like 10-inch displays. It had a high margin, too.

But those are all the revolutions. Those are the big sticks in the sand. Apple didn't kill established markets and rework the industry just by coming out with a new model.

If the iPhone had remained mired in 2.5G networking and its chunky (though seemingly elegant at the time), slow original form for long, it would have foundered. An original iPad also seems like molasses, low-res, and horsey. Instead, Apple pounded away year after year with new models, most of which were not astounding improvements over the year before. They were early to bring Retina (quadruple density) displays, but late to LTE (which was earlier a battery burner and not widely available enough to be worth the cost). The fingerprint sensor on the 5s is, of course, not new, but Apple waits until features, cost, and the ability to deliver to tens of millions of people are there; then they pull the trigger.

Every other year, and sometimes only every third or fourth year, Apple drops a bombshell feature that sets the bar higher for competitors, or checks a box that its rivals might already have managed (often at the sacrifice of high cost or poor battery life) after Apple has watched them drain margins or lose sales to make it happen.

Apple makes its money through incremental improvements that appeal to people who bought their last version of whatever it is Apple offered two to four years previously. For a new device, like an iPad, they get early adopters. For a refresh of existing devices, like a MacBook upgrade, they hit people who are finally feeling the speed differential enough to budget for a new machine.

The iPhone 5c is not supposed to convince an iPhone 5 owner to "upgrade." It's the same device. It's designed for iPhone 4 and 4S owners, first-time smartphone owners, and switchers from other smartphones, to be sexier (colors!) than the original iPhone 5, but not feature different.

The 5s likely won't convince many 5 owners to upgrade yet — most iPhone 5 owners in the U.S. can't get the fully subsidized price for another several months or more than a year; those elsewhere who bought a 5 at its full list price aren't likely to sell it used to get a 5s.

But the 5s has just enough to be interesting: the new camera features are intriguing (slow-mo, burst mode, auto-selection of "best" shots, bigger individual sensors and thus better low-light shots, bigger maximum aperture, two-color LED flash mixing), the fingerprint sensor sounds like a way to get security and avoid irritation, and one wants to wait and see what happens with a 64-bit processor and the separate motion processor. (In fact, I qualify for an upgrade and I'm seriously considering getting a 5s.)

From a financial perspective, this is how the phones line up for Apple: 

  • The 4s is Apple's very high-margin phone. It uses two-year old components and a perfected production system that's been streamlined over time. Though it is $200 less than when it originally retailed, its margin has still likely a much higher percentage than when it debuted and somewhat higher than a year ago.
  • The 5c almost certainly has a higher margin, even with a $100 drop in price, than it did at debut as the iPhone 5. The polycarbonate backing drops the raw cost, in addition to the same year-over-year benefits as the 4S's continued production.
  • The 5s will have lower margins at introduction, but it's necessary to keep the price in line to bring aspirational buyers, early adopters, and frequent updaters to the party, as well as grab switchers. In a year, when all the 5s features become standard on the next-cheaper phone, they'll have driven the cost way way down.

There is a very reasonable sense of deflation, though, despite all of this being sensible. A software developer and App.net friend, Isaiah Carew, wrote in a message on App.net, of which I quote part:

I've never cared that they're a company. I've never cared whether they succeeded or failed. I've never cared that they made or lost money.
I care that they do amazing things that make life better.

I understand this sentiment. Apple is a company that makes boatloads of money. It's hard to wait through the boring periods in which they're swimming in cash like Scrooge McDuck, to see how they will challenge the industry again.

In March 2012, I wrote an essay for TidBITS that touched on some of the issues of incrementalism, focusing more on the competitive issues of cost and features in the mobile-phone market.

[1]: Alistair J. Cullum, a biology professor, wrote in with a correction that I very much appreciate. The evolutionary term "punctuated equilibrium" refers to long periods of relative stasis, he writes, alternating with short period in which change occurs rapidly. The incrementalism between releases that I write about is more accurately called "gradualism," in which improvements pile up a bit at a time over long periods. I suppose I'm suggesting a synthesis: Apple has gradualism, not stasis, between its points of huge change. Thank you, Alistair!