This is Peter Adkison's official response to Martin Ralya's open letter titled, "Peter Adkison Hates us: An Open Letter About GenCon Registration"
Dear Martin Ralya, and everyone else who is curious about this issue,
First, my apologies for the slow response. I don’t have a great excuse for taking so long to write. It’s just that the whole Gen Con registration situation is one of those heart-wrenching topics that’s been difficult for me to know how to deal with. For the last couple of weeks, whenever I looked at my to-do list I’d see the item, “Respond to Martin Ralya’s post,” and just not feel up to it and move on to something more fun, like having a root canal, or calling my IRS agent.
But, alas, I can’t put this off any longer, or I’m just compounding “bad registration system performance” with an image that not only do I hate you, I don’t care about you either. Those of you who know me know that’s not true, but a slow response on my part certainly hasn’t helped the situation.
So here I am, fumbling with words, trying to figure out how to begin.
The main reason that this problem has been difficult to fix is that it’s a problem we only have once or twice a year. What this means is that when you have a theory on how to fix the problem, it takes a whole year to figure out if your theory was correct or not. If you're a cook and you cook a bad dinner, you can try again the next day, and within a week you've had 7 points along the learning curve. But with this problem, we have to wait a whole year between each instance on the learning curve. This drags out the improvement process considerably.
The problem is that we’re running along day-to-day, no problems. Then Indy event reg goes live and WHAM we’re hit with something like 10-100 times as much traffic as normal. And, of course, something goes wrong. We try and fix it best we can “real time” but, of course, that’s nearly impossible, but by day 2 or 3 we’re over the hump and things are moving along. Then we look at the whole situation in more detail and the tech guys go, “Here’s a list of things we can do that will improve the situation.” Then we talk about how much these fixes will cost, we weigh against other priorities, and decide upon a course of action.
Then we implement these changes, or at least some or most of them.
Then the next year when Indy event reg rolls around we all bite our nails and hope for the best.
And this year it looked at first as if we’d finally licked the problem. Whereas last year we had 632 people successfully register for an event during the first day of Indy event reg, this year we had 2,673 people successfully do so. And the average processing time per attendee was way less. In fact, most of these 2,673 registrants concluded their registration during the first four hours of the system going live, whereas last year it took almost to the end of day 3 to hit that level. We were quite excited around the office, getting ready to break out the champaign. Four times as many registrants on opening day! Woo hoo!
But then I started hearing about the problems. Even though wait times were significantly reduced, it still wasn’t at satisfactory levels for many of you.
So, what do I do?
Honestly, I’m not sure.
The tech guys have told me that they’ve identified a bunch more things they could do to improve the system, without a huge additional outlay of cash. Things like moving some of the processing to a different server, optimizing the code, etc. So, we’re going to do those things and see how much that improves things for next year. Maybe it’ll be enough that performance is “acceptable”, or maybe we’ll still have a problem. But as long as we can find reasonable things to do to improve things, and those things actually do that, I feel we’re on the right track.
Another thing I’ve been looking at is the question, “How many people are suffering, and how badly?” If I’m reading the consensus correctly the main concerns are as follows: double charging due to the shopping cart not clearing correctly, events selling out, and over all processing time.
I’d think the biggest of these issues is the processing time. The shopping cart problem affected about 150 of you and we think we’ve fixed this problem for next year and, of course, we’ve refunded all the double-charges. Embarassing, but I think we can move on without additional therapy. Regarding the problem of events selling out, that’s disappointing, but I think everyone understands that if Monte Cook runs a D&D game, there’s no way we can “ramp up” that event to accommodate 200 people. So I’m assuming the main issue is processing time.
So, let’s look at the data…
I had our tech guys pull the data on the first six hours of event registration. During these six hours, 433 people purchased event tickets. Of these 433 people, 112 were able to purchase their event tickets in under 10 minutes. For 146 people it was 10-20 minutes, for 89 people 20-30 minutes, for 48 people 30-40 minutes, for 29 people 40-50 minutes, for 8 people it took 50-60 minutes, and for one poor soul, over an hour.
So this begs the question, what’s acceptable? On most days, I’d say you should be able to get through this process in under 10 minutes. But on the opening day of event registration for the largest consumer games convention in North America? And given that this includes all the shopping time, looking at event listings to decide what to pick, picking something else if what you wanted isn’t available, and eventually making the final decision and going through the shopping cart routine? Sure, 5 minutes would be nice, but for this situation I think 30 minutes is probably acceptable, especially since most of you are probably doing this from the convenience of your home or office computer, possibly playing computer games or eating lunch in the meantime (way better than standing in line onsite).
If that’s a reasonable goal, then how many people did we disappoint? Well, according to this data, that would be 86 people.
I certainly don’t want to disappointed even one person.
But how much should I spend to fix a problem for 86 people? A serious upgrade would cost about $100,000 or more.
So there’s the situation. As far as I can tell, all this hub bub is over just 86 people, all but one of which still finished in under an hour.
I gotta say, this doesn’t seem that bad. It certainly doesn’t seem worth spending $100,000 to $200,000 to fix.
Could I just take the 86 of you to dinner instead? It would be a lot more fun and cost a lot less money!
Anyway, now you see why its been hard for me to respond. I sound like a jerk if I trivialize the problem. So, instead, I’m doing exactly what I did last year, which is to listen to the tech guys, conclude that the list of things they want to do to improve performance is sufficient (they do have what looks like a reasonable list), and go forward with that.
I apologize to those of you who feel this isn't sufficient, but I think this is our most prodent course of action.
Sincerely yours,
Peter D Adkison