Mysteries of the Comic-Con ‘Hotelpocalypse’

Marriott-Marquis-SDCC-2015-photo-by-Kendall-Whitehouse-1020x492

Due to the enormous popularity of San Diego Comic-Con, acquiring a ticket to the annual pop culture event is difficult. If you are lucky enough to snag a ticket, however, getting a hotel room may be even more challenging. In both cases, the demand far outstrips the supply. In the case of hotel rooms, confusion over how the sale process works exacerbates the situation. This past week, the quest to secure rooms for San Diego Comic-Con 2016 — affectionately (or not so affectionately) known as “Hotelpocalypse” — left many attendees confused and frustrated.

When Demand Exceeds Supply: Randomize!

As with most large conventions, Comic Con International, the organization that runs San Diego Comic-Con, arranges for blocks of hotel rooms to be available for attendees throughout San Diego’s Gaslamp Quarter and as far away as the airport, Mission Valley, and Coronado Island. This year travel planning agency onPeak managed the annual hotel sale for Comic Con International.

Hilton-Bayfront-SDCC-20150-photo-by-Kendall-Whitehouse-480x480While rooms are available outside this system, the high demand drives prices to astronomical levels. When booked through Comic-Con’s travel agency, onPeak, a room at the popular Hilton Bayfront costs $308 per night. (For comparison, on weekends other than that of Comic-Con, rooms at the Bayfront go for as low as $169.) If you book the same room outside of the onPeak sale during the weekend of Comic-Con, the price is $1095 per night (plus the usual taxes and fees). Understandably, many want to acquire rooms through the Comic-Con sale.

To apply for a room, attendees submit a form with their top six hotels choices listed in order of preference, enter their personal data, and specify what should happen if none of their six hotel selections is available: book a room in any hotel on the shuttle route, book a room at any hotel in the area, or discard the request.

In previous years, the form went live at a specified time and requests were processed in the order submitted. It was a mad dash to complete the form as quickly as possible. Getting a room in one of the popular downtown hotels required racing through the multi-section form in less than a minute or two. The high peak demand on the system led to numerous technical problems such as pages not loading or submissions inexplicably lost.

Tickets to San Diego Comic-Con were once similarly allotted on a first-come, first-served basis, with equally problematic results. Two years ago, Comic Con International switched to a system of randomized ticket sales. During a recap of the previous year’s event, CCI President John Rogers noted that the earlier approach had little advantage over a random lottery, stating, “There are so many people hitting the system at the same time that, in fact, it is random.” (“See Comic-Con Ticket Sales: Systematizing Randomness“)

This year, Comic-Con International decided to randomize the submission of hotel forms as well. Users were given a three-hour window to access an online “waiting room” and then, at the appointed time, were randomly placed in a queue to gain access to the hotel request form.

While this seemed to go well on the day of the hotel sale, there was a great deal of uncertainly about the details, particularly when attendees started getting the results a few days later.

People reported having relatively low numbers in the queue, yet failing to receive any of their hotel selections. Reports filtered in of friends with later spots in the queue getting rooms at popular hotels while people earlier in line got none.

Confusion over Sequence, Timestamp, and Duplication

While those who failed to get the room they wanted were obviously disappointed, much of the frustration arose from confusion about how the process worked.

The early word was that, even with the randomized queue as a gating factor, requests would still be processed in the order of the timestamp when they were submitted. Thus, as in past years, once the form appeared, many raced to complete it as quickly as possible, risking the possibility of making critical errors that would invalidate their submission. After the sale closed, a tweet from onPeak stated that forms would be handled in the order of the user’s assigned place in queue rather than the submission timestamp. Some anecdotal reports seem to contradict this, however, leaving it unclear how the forms were sequenced for processing.

Users were also told that duplicate submissions would not increase the odds of getting a specific hotel and, in fact, it would likely decrease one’s chances since “only the most recent submission received will be the one processed.” What qualified as a “duplicate submission” remains unclear, however, with reports (or speculation) of requests being disregarded for myriad reasons, including forms with the same mailing address, or with the same phone number, or for no apparent reason whatsoever.

A parody Twitter account, “Fake onPeak,” appeared shortly after the sale lampooning the organization behind the process, issuing tweets like: “Reminder: As you call us today, you’ll be placed in a random order. Call twice and we will have to ignore your request.

Hotel Allocation Schemes: Sequence and Preference

Another aspect of the system is unclear: What is the exact method by which hotels are allocated based on people’s sequence in the randomized list and their priority selections?

A common view is that the Comic-Con hotel sale works something like this: People’s requests are first randomly ordered and then the system goes through this list and examines each person’s six selections in the order given to look for an available room.

In other words, the system starts with the first person in the queue. If a room at their first choice hotel is available, they’re allocated that room. If not, the system looks at their second choice. If a room in that hotel is available, they’re granted that room. And so on through the set of their six hotel options. If none of these is available, the system follows the option specified for this situation such assigning the person to another arbitrary (randomly selected?) room on the shuttle route or in any available room. (The details of how this subsequent selection process occurs are also mysterious.)

While this system is simple to understand and relatively easy to implement, it can lead to potentially sub-optimal outcomes. For example, let’s assume that when my slot in the sequence arrives, my first five selections are unavailable and there is one room left at my sixth choice hotel, say, the Omni Hotel. I would get the room at the Omni. If the next person in the randomly-ordered sequence selected the Omni as their first choice, they would not get the room.

You can argue that this is fair, since I was randomly placed ahead of the other person wanting to stay at the Omni. But it’s unfortunate that someone who wants the Omni as their first choice loses out to someone who cares only slightly for that hotel (particularly since the selection was made randomly).

An alternative approach would seek to maximize the overall satisfaction of all participants by following a different algorithm. Rather than looking through all six hotel options for each individual before moving on to the next person, the system could fill each hotel by looking through the ordered list of all the requests by priority first. In other words, start with a hotel (in order of size or popularity — or simply alphabetically), then go through the ordered sequence of users looking for those who made this hotel their first choice before looking at the second-choice selections for this hotel

For example, when looking to fill the rooms at the Hilton Bayfront, look through the ordered list of people to identify those who listed the Bayfront as their first choice. Proceed through the requests in the ordered sequence – looking only at each person’s top choice — until either the Bayfront is filled with people who picked it as their first choice or the end of the list of people is reached. If the system gets to the end of the list of first-choice selections and rooms are still available, it then starts over at the beginning of the list looking for people who selected the Bayfront as their second choice. Continue in this way for each hotel until all the rooms are filled.

This approach would assure that each hotel is filled with those who are most desirous of that location. The randomized sequence would still be important, but it would carry less weight relative to the individual’s prioritization of the hotels.

Until Next Year

Because the demand exceeds the supply — particularly for the popular downtown hotels — there will inevitably be complaints from those who failed to get the room they wanted. But the lack of transparency about how the process works exacerbates the problem. Hopefully, for Comic-Con 2017 attendees will know more about how requests are sequenced, when they might be rejected, and how people’s hotel priorities are allocated.

This year’s Comic-Con International: San Diego takes place July 21 through 24, 2016, with a Preview Night on July 20.

Update 2016 April 13:

The SDCC Unofficial Blog has an update that clarifies some of the issues surrounding this year’s Comic-Con hotel sale, with a few technical details that will no doubt be helpful for next year’s Hotelpocalypse: “CCI, onPeak Offer Insight to San Diego Comic-Con General Hotel Sale 2016.”

Update 2016 April 26:

An analysis of the results of this year’s hotel sale based on data supplied to the Friends of Comic-Con forum provides additional insight into how the process worked: “SDCC 2016 Post Hotel Lottery, a Statistical Analysis.”

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s