The consensus on WWDC’s sellout time this year seemed to be about 2 minutes. As one of the many people who hammered the servers trying to get a ticket, only seeing error messages until being shocked by a “Sold Out” banner far too quickly, it felt like a lot less time — the unofficial “71 seconds” time sounds right to me.
It’s a great story. Last year, WWDC tickets became available suddenly, without advance notice, and they took almost 2 hours to sell out. Even this year’s Google I/O tickets, notorious for selling out very quickly, took 49 minutes. Surely this year, in the height of Apple pessimism, WWDC selling out so much faster than Google’s conference would be a great story.
You’d think Apple might even mention it afterward — if that’s really what happened. Maybe it is. I have no evidence and haven’t heard any tips to the contrary. But I do have another hypothesis that I think is more likely.
For the last few years (since WWDC started selling out), Apple has limited tickets to approximately the same total: 5,000 to 5,500, depending who you ask. Presumably, at least a few hundred of those are allocated to press, VIPs, big partners, student scholarships, etc. that Apple can give away as they see fit, so let’s assume about 4,500 are made available for public sale.
After this year’s unexpectedly rapid, effectively random, and error-filled sellout, we started hearing about people getting calls from Apple offering them tickets. Most of these were people who successfully added the ticket to their cart during those 71 seconds, but weren’t able to complete the checkout due to server errors.
These phone calls seemed to go to a lot of people. I’ve seen at least a hundred people report that they got a call after having the ticket added to their cart but having errors with checkout. It’s likely that I didn’t see all of them, of course — I’d estimate that Apple may have called at least 500 people after the sellout who had tickets in their carts. I don’t think I’ve heard from anyone who had a ticket added to their failed cart and didn’t get a phone call from Apple within a week of the sellout. That’s a lot of extra tickets to be selling.
And that’s not all. Apple has since called many people who reached out to developer-relations contacts, offering them tickets as well. I know a lot more people who have gotten tickets this way, even as recently as this week. Apple’s still selling tickets — there are probably at least a few hundred more that have been sold this way.
So I see two possible explanations.
Apple could be selling far more than the usual number of tickets, probably by at least 1,000, because they want to please the large number of people who tried and failed to get tickets during those 71 seconds. If this is the case, it should be easy to tell: WWDC was already crowded, and there aren’t many places to hide an extra thousand people. They’ll be noticed.
Or maybe all of those server errors meant something.
My best guess: some part of the infrastructure handling the purchases mistook 4,500 connections, transactions, or sessions for 4,500 sales. And when the front-end servers collapsed under the load of everyone hitting them at once — a first this year, since the availability time was preannounced — we all started refreshing, those connections started stacking up, and something on the back-end triggered the “Sold Out” state early because it was mistakenly counting all of those failed sessions.1
Maybe the actual number of tickets successfully sold during those 71 seconds was much lower than 4,500. It sure didn’t look like a system that was successfully processing transactions and behaving as expected.2
If so, it probably took more than a few minutes to realize what happened, by which point it was too late — the press already knew, the public already knew, and the stories were already being written: “WWDC sells out in two minutes!”
Apple would be left with a choice: admit a major screw-up and go through the hassle and PR cost of scheduling another time to sell the rest of the tickets, let the conference happen with far fewer people than usual (which we’d definitely notice, and which would be a tremendous waste of opportunity), or quietly try to sell the rest of the tickets behind the scenes.3 Obviously, Apple would choose the quiet, face-saving option.
If that’s what happened, it should also be easy to tell: we shouldn’t notice an extra thousand developers in Moscone this year.
It’s an effective way to paper over an inconvenient failure. Unfortunately, this strategy could backfire: what will happen next year? Obviously, what happened this year was undesirable for just about everyone involved, so Apple will probably try to fix the system — but if they fix it, tickets will probably take longer to sell out.4 How will it look if WWDC’s sellout time increases in 2014?
Many developers even reported seeing other random developers’ information in the checkout screen when refreshing. Obviously, the errors weren’t simple load throttling — things were breaking in unexpected ways. ↩︎
This hypothesis also may explain why the added-to-cart people got the calls first. Apple can’t realistically call all 250,000+ registered developers. Maybe they were able to get a list of added-to-cart people from logs, so it was an obvious list of which members of the Developer Programs wanted tickets. ↩︎
If everything worked properly this year and the first 4,500 people to attempt to buy tickets succeeded, I imagine the sellout time wouldn’t have been a lot longer — possibly 5–15 minutes. I don’t see any signs of weakened demand. But it will still look bad if tickets sell out in 71 seconds this year and 15 minutes next year. ↩︎