composition.al

Scaling !!Con

As regular readers of this blog are probably tired of hearing by now, !!Con (“bang bang con”) is a weekend-long conference of ten-minute talks about experiencing computing viscerally, held annually in New York since 2014. I’ve been helping organize !!Con since the beginning, and the rest of the organizing team and I are now preparing to put on our 2017 event.

!!Con has been a success. In fact, year after year we find that the demand for what we’re doing is greater than what we’re able to meet: we get a lot more strong talk proposals than we have room to accept, and we have way more potential attendees than we have room for. It feels good to be wanted, but it also means that we’re constantly disappointing people, and that sucks.

So, how can we scale !!Con to meet demand?

The !!Con organizers have been asking ourselves this question and trying to answer it for years. There are different dimensions along which we can scale: in our setting, I think of “vertical scaling” as meaning growing the !!Con conference itself (more talks, more attendees, more days, and so on), while “horizontal scaling” means having more conferences. This post is a personal history of what’s been done to scale !!Con so far, and some thoughts on what I think can be done in the future. Although I think many of the !!Con organizers would agree with what I’m saying here, my opinions shouldn’t be mistaken for the official position of the organizing team.

Probably not very vertically scalable: number of talks

!!Con gets many more talk proposals than we have room to accept.

In 2014, our first year, we accepted twenty-four talk proposals out of 168 submissions. The following year, 2015, we accepted twenty-three proposals out of 220; one speaker had to cancel at the last minute due to illness, so we had twenty-two actual talks at the conference that year.

Those first two years were already painful enough in terms of how many talks we had to reject. Then, in 2016, we had a record 226 talk submissions. That incited us to rearrange our schedule in order to pack the talks in more tightly, making it possible to have thirty talks: twenty-nine new acceptances, plus the cancellation from 2015. Having thirty talks seemed to work well, so this year we’re again planning on thirty talks, selected from the 215 submissions we received.

On average, then, we reject about 87% of talk proposal submissions — not because 87% of submissions are bad (although, yes, some are bad), but because our single-weekend, single-track format means that we can only fit in so many talks. Every year, we end up rejecting a lot of strong submissions.

How many strong submissions does “a lot” mean? We use a version of the “Make Champions Explicit” approach to talk selection, in which each person reviewing a talk proposal individually gives it a grade from A to D, with A meaning “I really want this talk; I will argue for accepting this talk” and D meaning “I really don’t want this talk; I will argue against accepting this talk”. (B and C mean something like “I like this talk fine, but I won’t argue for it” and “I’m not into this talk, but I won’t argue against it”, respectively.) Then we meet as a group to compare notes and decide on the program. We consider for inclusion every talk that got at least one A grade; accepted talks have to come from that set. This year, out of the 215 submissions, there were eighty-six talk proposals that received at least one A grade from a reviewer. In other words, choosing to accept thirty talks means choosing to reject fifty-six talks that at least one reviewer really wanted to accept.

Fifty-six talks that at least one reviewer liked a lot! That’s enough talks for two more entire conferences! The situation was similar in 2014, 2015, and 2016, so, that’s about eight conferences’ worth of strong talk submissions that didn’t get into !!Con, despite at least one reviewer really wanting them.

Why not have the conference be more than two days long, then? Why not have multiple parallel tracks, so that we can fit in more speakers? In addition to the significant logistical and financial hurdles that we would have to overcome if we added more days or more tracks to the conference, there’s also the concern that making the conference longer might make it less fun and accessible. Two days of rapid-fire ten-minute talks is fun, but exhausting; if we did it for more than two days in a row, exhaustion might overtake fun. Adding more days would also mean we’d have to hold at least part of the conference on weekdays, which would make it harder for many people to attend. And the single-track format has numerous benefits: attendees have a shared context for conversations; nobody has to miss a talk they want to see because it’s double-booked with another one that they want to see; and speakers never have to compete with each other for an audience. I think we might be able to fit in a few more than thirty talks by cutting out some of the breaks we have now, starting the days earlier, or ending them later — but, well, see “exhaustion”, above.

Given all this, my suspicion is that we’re pretty close to the limit on the number of talk proposals we can accept at a single conference, and that the limit will have to remain in place no matter how many stellar proposals we get.

For the most part, the academic conferences and workshops that I’ve participated in do not operate this way. They don’t decide in advance on a target number of papers or talks to accept; instead, they accept everything that meets their standard of quality, and then it’s the organizers’ job to fit that many talks into the schedule by tweaking the length of talk slots and Q&A periods as needed, and by running parallel tracks if necessary.1 For !!Con, talk slot length tweaks aren’t an option, because our talks are ten minutes long by definition, and we don’t have Q&A. So we don’t have much wiggle room here.

Actually somewhat vertically scalable: number of attendees

Besides failing to accommodate all the strong talk submissions, we’ve also failed to accommodate everyone who wants to go to !!Con as an attendee. As one might expect for a conference that’s in New York, it’s usually been an issue of not having enough physical space.

In 2014, our first year, the Recurse Center generously agreed to host our conference. There were 120 people in attendance, about 40 of whom were speakers, organizers, RC staff, or sponsors. The conference was fantastic, but with that many people, the RC space was bursting at the seams. I remember long lines for the bathroom, chairs that were crammed uncomfortably close together, and having to stack up desks in piles so that there would be enough room for people to walk around. Once tickets had been set aside for speakers, organizers, staff, and sponsors, the tickets we had left for regular attendees sold out more or less immediately, and our waitlist grew to 100 names.

The following year, 2015, we decided to set up an attendance lottery, so that people’s ability to attend wouldn’t depend on them having a fast Internet connection and being online at the exact moment tickets went on sale. About 340 people signed up for the lottery. We offered tickets to a randomly selected 80 of those people, and everyone else went onto the waiting list. The RC staff agreed to host us again in 2015, and we all loved RC, but everyone knew it would be an uncomfortably tight fit, and so we made the difficult eleventh-hour decision that !!Con would move out of RC. Finding a conference venue on short notice for free in New York is no small feat, but thanks mostly to RC co-founder Sonali Sridhar, two weeks before the conference we were able to line up a very nice venue at NYU MAGNET. In the MAGNET space, we had elbow room (no more lines for the bathroom! no more precarious stacks of furniture!), but we still didn’t have room to seat more than 120 people, again including speakers, organizers, and sponsors.

I felt terrible that we had to turn away so many people our second year. Shortly before the 2015 conference, I wrote, “Personally, my secret hope is that people realize that we can’t meet all the demand ourselves, and they start their own awesome conferences.” At the time, I didn’t really believe that increasing the number of attendees at !!Con was a viable option. My friend Tom Santero — who knows a few things about scalability, and a few things about running conferences — gently suggested otherwise, and as it turned out, he was right.

2016 was the year we decided to move to a bigger venue. The great venue search of 2016 was among the biggest and most stressful parts of conference organization, particularly for Nabil Hassein and Danielle Sucher, the two organizers who were physically in NYC and therefore responsible for visiting potential venues in person. We ended up paying $10,000 to rent a venue in Brooklyn for the weekend that would accommodate 250 people, allowing us to more than double the number of tickets we could make available. Since more tickets would be available, we decided not to run the lottery, either.

Moving to a bigger venue meant that tickets sold out in hours instead of minutes. Even so, we still accumulated over 100 names on the waitlist after tickets sold out. Moreover, due to the cost of the venue, we had to give up on other things — for instance, we were only able to serve lunch on one of the two days of the conference, and we had to ask the Internet to pitch in and help pay for breakfast — and even then, we still ended up being somewhat in the red for the year. Doubling in size also led to some other logistical problems, like not being prepared to move everyone through registration fast enough and therefore getting off to a late start on the first day.

However, the thing I that had been most worried about — that !!Con wouldn’t feel like !!Con anymore if it doubled in size — ended up being a non-issue. To the contrary, I thought that !!Con 2016 turned out to be the truest expression yet of what we stood for. We were able to make space, literally, for a lot more people, but we didn’t lose the tight-knit feel that the first two conferences had had.

This year, our venue — generously provided to us for free by our sponsor AppNexus — will accommodate 270 people, the most we’ve ever been able to have. (Based on historical trends, though, we still fully expect tickets to sell out.3) The downside of using a sponsor-provided space instead of paying to rent our own space like last year is that we don’t have as much control over the space, but to me, it’s worth it to have that $10,000 to spend on other important things, like food and speaker travel reimbursement.

Overall, I’m happy to say that vertically scaling the number of attendees at !!Con has turned out to be feasible, if not entirely problem-free. We’ve successfully doubled in size since 2015. Even so, the numbers say that we’re still leaving a lot of would-be attendees disappointed. Going from 270 to, say, 500 attendees seems harder than going from 120 to 270. That’s not to say that we can’t do it or that we won’t try, but I think it’ll be a challenge. Venues are expensive, and we were very lucky this year to have a sponsor willing to give us their space for the weekend. And, of course, even if we continue to scale vertically to accommodate lots more attendees, accommodating lots more speakers seems unlikely. So, where does that leave us?

Scaling horizontally: more !!Con-style events, run by more people

I think that the long-term answer to our inability to meet demand will be to scale !!Con horizontally, by having a lot more conferences like it. I don’t mean that the existing !!Con team should run a lot more conferences, though! We all have jobs, not to mention other stuff going on in our lives; we probably don’t have the bandwidth to run more than one conference per year as volunteers. Instead, what I’d love to see is for people who like the !!Con approach to go out and start their own similar events around the world.

To some extent, this has already happened! The first !!Con-inspired event that I’m aware of was EnthusiastiCon, held in Berlin, Germany in July 2015:

EnthusiastiCon 2015 was a conference for the programming community. Over two days of short presentations, we talked about what excites us about programming – the strange, the wonderful, and the clever solutions to unusual problems. Wonderful people joined us to share their enthusiasm with us or listened to what makes fellow programmers beam with joy about our craft.

Like !!Con, EnthusiastiCon was an eclectic two-day conference of short talks. (A playlist of videos is available.) The way they structured their schedule reminds me a bit of !!Con. In fact, one of the talks at EnthusiastiCon was Igor Wiedler’s reprise of “MissingNo., my favourite Pokémon!”, which he originally presented at !!Con 2015.

Another !!Con-inspired event was Curly Braces, held in Atlanta, Georgia, USA in November 2015:

The Curly Braces conference is a storytelling event about computing, and the intersection of computing with the arts and sciences.

This is a single-track, single-day conference, where all the talks are lightning talks. […]

Curly Braces is a local, community, for-the-love-of-it, event. There are no sponsors, and there is no cost to attend. No one will be selling you anything, and you are not the product. It is a friendly, welcoming space, so all speakers and attendees must abide by the code of conduct.

On their website, the Curly Braces organizers specifically acknowledged !!Con for giving them “the idea of a conference of short talks about exciting things.”

And most recently, there was Hello, Con! in Toronto, Canada in February 2017. Organizer (and !!Con 2015 speaker) Josh Matthews described Hello, Con! as a “tech conference in the style of @bangbangcon”, and he asked for advice from us a couple of times while putting the conference together (which we were, of course, delighted to provide!). After Hello, Con!, I exchanged emails with one of the attendees, who wrote:

At the end of HelloCon, I was genuinely excited to start delving into projects that the speakers had talked about (such as rr by Mozilla and Phillip Offerman’s blog series on writing an OS in Rust). A lot of tech events that I go to are focused on recruiting interns or handing out swag, but HelloCon was simply about getting exposure to the field of technology and getting to meet a very diverse and skilled group of technologists :)

Hearing this from a Hello, Con! attendee made me really happy. It suggested to me that some aspect of what I think of as the !!Con aesthetic was also present at Hello, Con!2 Even more excitingly, the attendee that I heard from is now planning on organizing a similar event herself, which was her reason for contacting me in the first place. As a result, I’ve been working on putting together a list of answers to questions that I hope will be generally useful to people who want to run their own events in the style of !!Con, and I’ll publish that list later this year.

The most important ingredient, though, is just enthusiasm for the cause. If you’re a fan of !!Con and want to help; if you’re frustrated that our tickets sell out so fast, or that we have to reject so many good talks, or that we’re too far away from you, or that the conference doesn’t happen often enough; if you want there to be more !!Con goodness to go around — then I’m urging you to help us scale horizontally by getting a team of organizers together and starting your own !!Con-inspired event. Other people have done it successfully, and you can, too.

  1. Another difference between how !!Con makes acceptance decisions and how academic conferences do it is that at !!Con we aim to make the program as a whole interesting, with an eclectic set of talks that complement each other. So, “we already have a talk on this topic” can sometimes be sufficient reason to turn down a submission, and we’ve never permitted more than one talk from the same person at a single instance of the conference. At academic conferences, on the other hand – at least, all the ones I’ve been involved with – it doesn’t work that way. The reviewers consider each submission independently and decide to reject or accept it based only on its own merit or lack thereof, so “we already have one on this topic” isn’t grounds for rejection. I think this is as it should be, but it sometimes leads to situations where there are several acceptances to the same conference that are quite similar in topic. There was a funny case of this at PLDI 2011, where both “Taming the Wildcards: Combining Definition and Use-Site Variance” and “Taming Wildcards in Java’s Type System” were accepted, and at the conference they appeared together in a session called, naturally, “Taming Wildcards”!

  2. One of the Hello, Con! speakers, Max Veytsman, was yet another person who also spoke at !!Con 2015. !!Con alumni are everywhere!

  3. Update (April 24, 2017): As predicted, we sold out. I imagined that since we had a few more tickets available than in 2016, it would at least take hours to sell out like it did in 2016, but it only took 22 minutes, and as of this moment, there are 240 names on the waitlist. So it seems we could fill a 500-person venue, if we could manage to get one.

Comments