Hackonomics 101
The economics of who pays for hackathons, and why.
This past weekend I helped organize HackCon, the first student Hackathon organizer’s conference, in NY. I gave this talk there, and I’ll give it again here. Ready?
So, right, I started the PennApps hackathon back in 2010. Here’s me trying to get more catering at the last minute. I’ve been there.
This was back in 2010-2011, but I’m still around so I have some perspective here. Nicholas Meyer and I made this chart the night before, which is (as you can see) attendance and cost per attendee for the event, across the 7 PennApps hackathons so far.
The cost per attendee was really high in the beginning. I didn’t know what I was doing and over-estimated turnout by about 2x and so over-ordered food (it used to be much harder to get people to come to hackathons because student hackathons weren’t a thing, or I was bad at convincing people). Then, as the event got bigger I got better at estimating this stuff and so the third event we ran on a reasonably slim <$100/person.
Then we (at around this point, it stops being me) figured out how to get other schools to come to PennApps: free travel. Originally, we got Megabus to sponsor tickets, though this ended up shifting to chartering busses for schools and direct travel reimbursements — attendance shot up.
So did cost, though — if you assume the base cost per attendee at a weekend-long event is $100 (HackMIT’s Ishaan explains why), travel reimbursements push that up to — well, however high you want it to go. So we went from economies of scale (cheaper t-shirts printed at bulk, poster printing) to actual diseconomies of scale, because each time the event gets bigger the average attendee costs more, since they’re coming from further.
Student hackathons get bigger every year. Can event size and quantity keep rising? Are sponsors going to continue being OK with rising costs?
Who knows. I took like two Econ courses in college before CS and PennApps consumed my life. But, hey I’m here, so I’ll try to answer the question.
To figure out how much money we can get away with, we have to figure out who is paying us and why on earth they would give us so much money. Not all sponsors are alike, though, and so we have to categorize their motivations before we can quantify the value we’re providing to them.
Twilio is at PennApps because they want students/future developers/ hacker, to know about their telephony API, and to use it at their next gig and tell their friends about it.
Facebook wants to hire you for your next internship or full-time.
Venmo is happy to be associated with the idea of student hackathons, they want to be seen as connected to all of this wonderful innovation — Kortina, the CEO, is like myself a Penn alum that feels “why didn’t this exist when I was a freshman” and wants to help however he can.
Now, in fairness, the categories are more muddled than that. Everybody cares about branding (18-24 college-age demographic, etc), and certainly Facebook and Venmo want you to use their API, and everybody would be happy if they met some future hires at the event. But the point is, everybody has a primary reason for sponsoring, where if it isn’t available to them they probably couldn’t justify sponsoring.
And, honestly, I can’t really quantify branding or evangelism. The value of a customer (or whatever metric evangelism uses) differs wildly from company to company, and to a large extent is an investment into the future. Branding, likewise, is a weird beast, dependent on a particular marketing budget of a particular company, and it’s hard to quantify. If you find a sponsor interested in paying for branding, great. If not, you’ll still be fine.
In fairness, Venmo is probably just my favorite sponsor that I consider to be branding. More common ‘branding’ sponsors are companies like Comcast, or Hearst, or GE — non-tech companies that want to be associated with tech.
But I can talk about recruiting, because I ran engineering recruiting for my last start-up. And because I help students figure out where to work / fix up their resumes pretty often. Recruiting is actually an industry and there’s a quantifiable substitute for going to hackathons. If I had to guess, recruiting accounts for more than half — probably more than two thirds — of your budget.
What’s the value of a hackathon to a company hoping to recruit?
Here’s a back-of-the-envelope calculation.
$20k — If, instead of sponsoring the hackathon, you used third party recruiters, you could reasonably expect to pay 20-25% of an annual salary. Assuming 20% and $100k starting salary, that’s a $20k contingency fee per intro that leads to a hire.
10% — How many students actually meet their future employers at a hackathon? Dave Fontenot from MHacks tells me the result of his informal survey is about 10%. We did a show of hands at HackCon and that figure seemed about right.
50% — Let’s cut that figure in half again. Internships don’t pay. I mean, they do a little bit, but that’s not where the money is. Nobody is paying third-party recruiters $20k for internship candidates. Maybe $5k. So we can only count seniors and intern-to-drop-outs, which (coupled with the $5k for interns) I’m going to say is ~50% of the pool. Yeah, it’s a rough estimate, but go with me here.
So, if your hackathon has (say) 100 attendees and 10 meet their future employer through the event and 5 of them become full-time and the value of a full-time hire is $20k, your event generates about $100k in value for recruiters, or $1k/value per attendee. There, we have an estimate.
Now, on the cost side, PennApps raises a total of about $200 per attendee right now, which means things seem pretty good — after all, it looks like hackathons are still 5x of a better deal relative to external recruiters. Woot.
Except, when a company sponsors your hackathon, they’re going to send representatives (hopefully), and the company has to pay them and put them in hotels, and pay for their flights too, and those flights are being booked way later and so they cost more, and so a rough estimate is the total overhead cost for sponsorship is roughly equal to the sponsorship amount. So you raise $200 per attendee, but your sponsors incur $400 per attendee costs, and those are the trade-offs they’re considering when they’re trying to figure out if they should use you or pay a recruiter.
Still, $1000 value > $400 cost, so we should still be good, right?
Well, maybe, but for how long?
Why I worry
There’s a lot of new hackathons out there now. Dave Fontenot has this great slide (which wasn’t in my presentation, but here it is)
Hackathons are getting bigger, and we’re getting more big hackathons this semester. I would guess not 1 of the ~14 big hackathons that are supposed to happen this semester (and haven’t happened already) are done with their fundraising yet, and who knows how many of them will hit their goals? Luckily, when you find out you’ll only raise 60% of your target, the thing that gets cut is travel reimbursement, so the hackathons can still happen, they’ll just have half the busses they were planning to.
Why are these hackathons bigger? In part, because we’ve figured out outreach models and they have been scaling. In part, because Dave Fontenot dropped out and has been going to each one of these and advising, and Mike Swift decided he’s going to start Major League Hacking and pursue this thing full time, in part because of inertia, but either way — here we are.
The thing is, we’re all eating from the same recruiting budget pie. Sure, it’s a big budget, but there’s certainly an upper limit, though not one that I can easily quantify.
This is kind of like peak oil — in the early 20th century, we figured out we love oil, and even though it kept getting more expensive to produce it (drill deeper, in remote locations, etc) we were OK with that until a combination of an oil exporting consortium (OPEC) and just getting closer to running out of oil lead us, in the 70s, to freak out about Peak Oil. Ask your parents.
We made a concerted effort then as Western Civilization to back off and prevent peak oil, which worked (though oil is still very expensive now, hmm). What insights can we gleam as we try to prevent peak hackathon?
Now, a caveat. Here is where we jump from fact to my conjectures.
Well, first, in the 70s we started investing a lot more in alternate energy sources — President Carter put solar panels on the White House (and Reagan tore them down, long story).
On the hackathon side, Nick from SendGrid brought up a similar problem — as hackathons get pricier to sponsor, evangelism-focused companies start saying no. That really sucks, because the evangelists are among the best mentors at a hackathon and add so much character. Replacing them with recruiters with deep pockets hurts our community and results in things like 3-hour-long intro talks from companies that are not experts at API talks.
So how do we harness the evangelism funding and awesome mentorship of the Twilios of the world without offending recruiting-focused companies?
What’s the one thing a company wouldn’t come if they couldn’t get? For recruiting-based companies, it’s probably the resume pile.
Aside: this probably isn’t as true for Google/Facebook, because most everybody is applying to Google/Facebook anyway and so they’re just reminding you to apply rather than depending on a resume pile.
For Twilio, it’s the ability to go talk to all the hackers in 120 seconds and amaze them and get them excited about Twilio.
So, great. Let’s make them pick. Deep-pocketed recruiters get the resume pile but no API demo. Giving an API demo is cheaper, but you get no resume pile. If you really want both, you can — but that should be way expensive.
Aside: A big side-benefit here is that API demos at the kickoff aren’t longer than an hour, which is a far superior experience.
You’ll probably want two separate PDFs for recruiting and evangelist sponsor, too. Mr Gorbachev, put up this wall!
Next, I want to talk about growing the pie.
Online advertising is an interesting parallel to look at here.
The way we price recruiting at Hackathons is basically CPM — companies pay a fixed fee for a certain number of website impressions (in this case, a certain number of attendees and maybe their resumes) but that’s it. It’s not a complicated structure.
Online advertising, though, has (roughly) three types of models:
- CPM (cost per impression), where Coke pays to have its banner somewhere and then you keep associating Coke with american-ness and bears and whatever else they want you to associate Coke with that isn’t diabetes,
- CPC (cost per click), where Fandango pays Google per click, and they pay a lot because they understand that when some googles “showtimes palo alto batman” (or better yet, “hotels in vienna”) there’s a pretty good chance the clicker is going to spend a bunch of money,
- CPA (cost per action), where you write review on your blog, and then include a link to Amazon to buy the book and add your affiliate link, and get 5-10% of the cash from purchases you generated.
And it turns out, there’s a boatload of money in CPC and CPA — otherwise we’d still all be doing CPM for online advertising. Advertisers love those models, because it de-risks the hell out of their business model — no more $4M all-or-nothing Superb Owl ads, advertisers can just pay money whenever they make money. Then the budget for ad spend doesn’t need to account for risk. If the only time you have to pay for advertising expenses is when you actually sell something, you’re printing money. It’s magical.
Aside: this is roughly how start-ups work. If the cost of acquiring a customer is less than the revenue generated by that customer, you win.
Back to hackathons. We have CPM down — we’re all doing it.
CPC and CPA, we haven’t even tried and it’s not clear what they look like. Maybe CPC is the cost sponsors pay per resume that specifically applies to their company. Maybe CPA can be everybody that gets a job.
CPC and CPA are both hard, by the way. Hard to measure, both internally and openly, so sponsors trust that you’re not inflating the numbers (look up Click Fraud to see this happening in online advertising with CPC).
They’re also hard from a cash flow standpoint — you need the money now and the hiring / paying you isn’t going to close until months after the event. So you definitely still need some CPM, or a financial backer of some sort that is willing to front some money in exchange for a cut of profits.
The point is, there’s a boatload of money in CPA and if we’re going to avoid Peak Hackathon we’re going to need to figure out how to get at it. That lets us grow the pie and put on more high-quality events.
It only takes 1% of your attendees to indirectly find a job through you to completely fund your next event. How magical would that be?
And while you’re figure it out, email me if you’re playing with ideas on this theme and let me know what you’re thinking. I’ve been thinking about this problem for a little while and I’m going to try to find a way to help.
PS. Raise more or spend less.
Wait! One more thing. The third way you can help prevent peak oil is to conserve energy.
- Travel reimbursements were a great way to kickstart Hackathon culture but may no longer be necessary — ask Maryland about how they got their school to sponsor them, for example.
- Smaller events are cheaper to run (you usually don’t have to pay for venue).
- Shorter events (HackDays) are cheaper to run — overnight means you have to pay for guards, have a harder time with venues, etc.
There’s room for innovation here, too — I’d like to see somebody run (and talk about how they have run) a quality event for <$50 per attendee.
PS.
- Not all companies pay for 3rd party recruiters.
- The numbers I am bringing are very back-of-the-envelope and may be off by 2-3x in either direction (plus, trends are shifting).
- I’m probably wrong about a whole bunch of things here, and I welcome your criticism and feedback. This talk (and post) are a way to express an idea that I think is worthy of considering and discussing.