Hacker News new | comments | show | ask | jobs | submit login
Aaron Swartz: How To Launch Software (aaronsw.com)
148 points by mqt 3170 days ago | hide | past | web | 48 comments | favorite

I find this to be excellent advice. This is exactly the approach we took at GitHub almost down to the letter. It took about 2 months until the site was good enough to use to host the GitHub source, another month until we started private beta with invites, and three more months until public launch.

Artificial scarcity is a great technique to generate excitement for a product while also limiting growth to a rate that won't melt your servers. We worked through a huge number of problems and early users gave us some of the ideas that have defined GitHub. By doing a Hollywood launch, things would have been very different and I am convinced, very much worse.

Do not, I repeat, DO NOT underestimate how much your users will help you to define your product. If you launch without having significant user feedback time, you've essentially thrown away a massive (and free) focus group study.

Let me also say that when we finally did our public launch, there was plenty of buzz, and all of it was the RIGHT kind of buzz. The buzz that attracts real, lasting customers (and no, we weren't on TechCrunch, that traffic is garbage).

Completely agree. There is just so much value to be gained from early users suggestions, and going through a few iterations.

Those early users are amazingly important. The overly passionate verging on stalker types that IM you asking if you just changed the color of a link and then tell you exactly what they think about it.

It was a similar story with Mibbit, although I didn't bother with beta, just threw the site up there and started iterating. The growth isn't quite a hockey stick, but it's extremely consistent which means I can grow the product along with the userbase.

Also I agree about techcrunch type traffic. Early on with Mibbit it got to the top on proggit. It was sort of fun, got a reasonable number of people in that night who looked around, said this is cool, then went... never to be seen again. Next day it was pretty dead. I think that's pretty typical.

37Signals has launched four (4) commercial products, and takes in (by my guess, from headcount and subscription numbers) mid-high 7 figures from them.

Aaron Swartz has launched, by my count, no successful products (unless I'm wrong and Reddit wasn't already a success story when he got there).

From Aaron's own web page, here's what I get when I click on his "my day job" link:


Why am I being such a dick about this? Because the argument he's making is disingenuous. Aaron advocates for an extended friends-and-family beta before launch. Aaron has no idea what 37signals does to beta and dogfood their product before the "Hollywood Launch". From what I can tell, everything 37signals does happens just before "step 5" in his master plan.

The difference between 37signals and Aaron Swartz isn't methodology. Methodology is a band-aid. The difference between Aaron Swartz and 37signals is ability to execute.

Here's a much better writeup from a much more credible source on the same topic:


The 'Hollywood Launch' can work if you already have notoriety to leverage, and already have the experience/userbase to do iterative improvement/testing internally. Steve Jobs is master of the Hollywood Launch.

If you're a small, young team making your first product, the dream of a 'Hollywood Launch' is destructive. It causes you to delay getting bulk feedback from disinterested strangers. It slows the organic learning/testing loop. It delays launch as you try to ensure you'll make a good first impression. And as Aaron notes, there's always something that goes wrong on a contrived 'big day'.

Steve Jobs and Apple on their umpteenth launch can usually keep those glitches to a minimum. (But not always: see MobileMe.) 37signals on their Nth launch of another webapp in the same mold as their other N-1 apps can pull this off, too.

But a young team launching a brand-new product will gain more from a measured ramp-up than an engineered hype-storm. They should listen to Aaron.

When you're finished with the ad hominem part of your attack, I'd just like to weigh in and say that Aaron got this exactly right.

Aaron advocates avoiding the "Hollywood Launch", marketing your product only after you've verified that "random people" are satisfied with the offering.

I'm asking you: are you saying that after Fog Creek spends 9 months building and testing a product, dogfooding it with internal users, and running closed beta testing, that you would avoid publicity for it until random paying customers started liking it?

Because I feel like I remember you making an actual movie about the last product you launched.

again, with the ad hominems!

Looked more like a question to me.

Best comeback ever (tptacek, I mean :))

I think you definitely need some balance between the two. Do a glitzy hollywood launch only after you've minimized the things that could go wrong.

This is the definition of ad hominem. For those like me who didn't know.

An ad hominem argument, also known as argumentum ad hominem (Latin: "argument to the man", "argument against the man") consists of replying to an argument or factual claim by attacking or appealing to a characteristic or belief of the person making the argument or claim, rather than by addressing the substance of the argument or producing evidence against the claim.

More on wikipedia http://en.wikipedia.org/wiki/Ad_hominem

It could be that Aaron is talking from bitter experience here.

Infogami was very much a Hollywood launch. He did almost exactly what 37signals suggested. He posted a teaser as soon as he got into the YCombinator program. He had mavens and insiders try out the product (I think; I can't imagine PG not trying out a YC startup's product before launch and showing it to his friends). And then he launched with a big splash.

And people hated it.

I suspect that product quality has far more bearing on startup success than how you launch. And as you say, the difference between 37signals and Aaron Schwartz is ability to execute. But Aaron's advice is a good way to find out if you haven't yet executed well and improve the product before everyone else finds out.

I think people hated it because there was nothing there.

He promised that Infogami would revolutionize making webpages the way that the Macintosh revolutionized the computer industry.

What he delivered was another wiki. I can't believe PG thought this was worth investing in.

Geez, that's a lot of ad hominem. Did I accidentally murder your dog or something?

I wasn't saying anything about what 37signals does, I was criticizing what they recommended. You know, in their book. That I link to in my first paragraph.

So, while 37signals doesn't break it out as clearly as you did, they do in fact recommend the same closed testing that you're recommending. For instance, in their book "Getting Real", they talk about how they do closed beta testing of individual features inside of their applications to selected groups of users, just like Google Mail does. They also publicly did closed beta testing of Highrise, their most recent application. I'd be surprised to hear that they didn't quietly test Basecamp before it shipped.

On the other point, I think I felt like in a discussion of whether we should be doing things this-way-or-that just because somebody said so, your resume was fair game. I think differently of that now. I'm sorry I offended you.

I believe that when 37signals created Basecamp it was for internal use. They built it for themselves. As time went on other small consulting firms (possibly that 37s knew) saw it and thought it was useful. From that point, Basecamp just continued to grow in popularity.

When nobody knows you or your startup do you really need a closed beta??

No, he is testing Jason Calacanis's approach to PR. By attacking Aaron Swartz ad hominen, he raises his fame/notoriety in hacker community.

Oh. I just forget Linus Torvalds called security people m________ting monkeys. It seems very hard to discuss issues rationally lately without calling names that based on some a priori assumptions of the opposite side.

He called OpenBSD people masturbating monkeys. He called security people whores. He actually had a sympathetic audience among many security pros for that statement. The whole thing is indeed a bit shady.

That doesn't make me any happier with my original comment, though.

Does this mean that the cause of hate comes from security people charges high price by hour to give people false sense of satisfactions? It seems that all people charges high price service by hours are hated in general such as attorneys and so on.

How life is so unfair and I feel your pain. But Jesus Christ loves everyone who charged clients by hour and then turns over a new leaf to wash his corpse in funeral.

Dude, I think you just wrote a Tool song!

Geez, that's a lot of ad hominem. Did I accidentally murder your dog or something?

Nah, that's pretty much tptacek's commenting style, I wouldn't take it personally.

While I agree 37signals has a much more impressive product track record than Aaron does and know quite a bit about marketing their products, the source of a statement has nothing to do with its validity. Either Aaron's argument is right, or it's wrong, independent of Aaron's status as a developer or entrepreneur.

That being said, I think Aaron is right about the proper method to launch a website, and also that if you look how 37signals actually acts, they launch products exactly that way.

I think there's a difference between the 37signals plan and the Aaron Swartz plan, which is that the Aaron Swartz plan says "don't market until you have happy customers", and then uses the spectre of product quality failures to back the argument up. Happy customers and crashing servers are orthogonal issues.

I actually agree with Aaron, but I do think his experience is relevant. Yes, he's either right or wrong regardless of his track record, but the intended reader doesn't know whether he's right or wrong, and won't know it without trying it. If you don't have reasonable way to verify whether what somebody says is true, looking at their credentials is probably the best you can do.

In practice, we do that all the time. For example, you probably have an opinion on global warming, or whether smoking is bad for you, but did you read the relevant papers, or just take the word of the researchers you thought were most credible?

In this instance, we all happen to be people who spend a lot of time thinking about startups and websites. So it's like someone posting something on global warming to a climate modeling forum - I would expect people to judge on merits not authority.

I think that's a harsh analysis and I don't think it's necessary to get personal about it. I'd say the launch of gmail was expertly done, and one that a lot of people copy.

Look at what happens when companies decide to "release" to everyone at the same time with a big PR frenzy - cuil. The companies that slowly but constantly grow seem to do better than the ones that go for a big style "launch" IMHO.

Releasing with a big style "Version 1" approach just doesn't fit with the dynamism of the web. Throw something up, get a few users, update it every day, and grow slowly but surely.

You're right: if you do a glamour launch of your product without making the effort on quality, you end up like Cuil (and I concede that Cuil is a good counterexample).

I object to the "either-or" argument the original post makes, and I object to the snide "we're only doing it this way because 37signals says to", as if 37signals doesn't have a store of credibility here that Aaron lacks.

Apart from 37signals though (Which isn't a good example IMHO - they have a rails 'fanbase'), what successful "Big release hollywood film" examples are there?

That's a good question. I'm having a slow day today; let me go look. If you don't see a response, I'll have conceded that there is no other good example of a successful "Hollywood" launch by a small startup. That would definitely hurt my argument.

[edit: my slow day got fast, but I also did have trouble coming up with a clear-cut example of a small company glamour launch success.]


We actually did the same thing for Kongregate. Except we gave out an invite code to everyone who asked - usually same day. So the artificial scarcity was... artificial. But if we'd had a big problem we could have stopped the invites.

We did that for about two months and then took the invite barrier down. Worked for us.

edit: I forgot the second half of this story. A few months after we took the invite barrier down, we decided to do a "Hollywood" launch - really all we were doing was changing the "alpha" to "beta" and releasing some new features (earning cards by playing games). It worked, we got lots of press. Fox News even called and asked if I could go on their afternoon cable show. I did (here's the video - http://tinyurl.com/2gpz7x). Our site immediately cratered - it was a much, much bigger surge than Digg. We got it back up in about 20 minutes and the follow on traffic was good as well.

So I guess we did a hybrid Gmail/Hollywood.

Can people please stop using "software" and "webapp" interchangeably?

I realize most members here are founders of startups that deal with webapps, but there are some startups that deal with traditional desktop applications - true "software," who share similar interests.

That said - I wonder which of these two approaches is better for desktop software development. I'd imagine the fact that a user has installed your software is a "hook" of sorts - they're easier to reel in than website visitors which can escape rather more easily.

-3? Looks like I'm the only desktop developer here then. My bad.

Looks like I'm the only desktop developer here then.

I'm sure it's because you imply that the only installable software is on the desktop. You insensitive clod. (I'm kidding, of course, but I also build non-web-apps and also found this advice only marginally useful, though I do find that having users from very early in the process is vital.)

You're not alone, I develop for the desktop too.

Isn't that a subset relation? webapps are a subset of software. So interchange only works in a single direction.

Maybe more like an intersection: web apps contain software, but the software isn't usually what's being sold; the user community, collected data and marketing/branding of the original site are a bigger deal than in shrink-wrap or contract software.

The big elephant in the room that nobody here seems to be taking into consideration: 37Signals charges their users to use their product.

The Hollywood launch allows for such a case, whereas this example makes that much more difficult. Do you force beta users to pay? Do they pay the same amount as others? Etc etc.

Hollywood launches allow you to broadcast a product as "finished". The issues Aaron cites seem to arise from broadcasting the "finished" announcement when the product is clearly not in that state.

I think GitHub did a free beta launch, followed by a few iterations with feedback, and then they started charging users.

That's reasonable for beta users, especially if they get a discount for their feedback.

Can you guess which method cuil chose. If you remember what cuil is of course.

Perhaps they should have read this first.

Cuil, the company that announced it was beta 10 months prior, refined its crawler multiple times when people start complaining, and the one that did the traditional massive PR push a couple weeks back but failed because their marketing team are just inept?

That one?

Somehow Cuil vanished at the end of Week 1. My biggest concern is those 25 millions that came from Teachers' pension funds.

That's exactly what I was thinking when I read it. But Cuil was trying to pull off a bigger sting than most.

But does the Hollywood launch hurt? It seems like the reasonable way to go is to get a big bump when you open doors, iterate using the feedback of people that stay, and get more coverage whenever you release a big new feature set.

Certain sites don't really make sense with invite codes too.

You're right. I think web apps get a huge benefit from soft launches, beta testing and user feedback. Its just invaluable information to have before you really present yourself.

On the other hand, if you're launching the next ValleyWag site, the Hollywood launch might be the way to go.

I think it boils down to how comfortable you are with your product.

If your stuff feels iffy and has less-than-ideal engineering practices, the slow and steady launch might work great. Expose enough of your idea to keep people idling around and begin free beta testing.

This isn't bad, per se. Everything is a trade off, and if engineering skill or time is not something you have an excess of, the slow and gradual launch can work.

On the other hand, if you have experience building and scaling and are confident that: a) the core of your idea is apparent in your product, and b) you did enough testing that things won't break on launch, by all means, build the hype and open the flood gates on launch day.

One of Fleaflicker's biggest advantages was having a smaller userbase than other sites.

This meant I could mess up in spectacular fashion and still fix things before anybody noticed. "Fixing things" involved computationally intensive tasks and the bigger sites just couldn't do that quickly.


I think that both can work. It depends on the goals of the company and type of software they are launching. I do think that gmail is a terrible example though. Google can launch anything, slowly or Hollywood style and it will get attention.Granted not everything they do succeeds, but Google has an advantage with any new product launches. A better example might be the original launch of Google.

I think both of them are important.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact