A lot of buzz usually surrounds a new Blog post by 37signals. As far as applications on the Internet go these guys are the kings of small, usable applications that do exactly what you want of them. Their Blog often keeps things in the same mindset; breaking old software engineering and management practices in favour of simple approaches to creating and launching applications.

However, the controversial method they have adopted can sometimes get out of hand, and exist solely for the purpose of it working for them. Their word is gospel and many developers hang on their every word, even when they are sometimes off the mark.

The article I would like to bring up today is The Hollywood Launch.

I’d definitely recommend that you read it before you read the rest of this, but for those with a short attention span the idea is to give users a taste of the action, then preview to a select audience, then to launch to fanfares and marching bands. In theory this idea is perfect. It allows you to build up an audience in the same ways Hollywood producers build up suspense for movies, with trailers and promotional posters.

A great concept, but one that has already proven to be flawed in execution.

Aaron Swartz of Raw Thought has beat me to the punch on this one, so instead of continuing with the rant I had prepared for this I’ll drag out what I feel is the most important point he has raised.

The Hollywood Launch only works when you have a quality product.

By quality, I mean in every sense of the word. Anyone remember Cuil? They created buzz, let out some previews and opened to firewords and broad claims, only for the website to fall flat on its face and provide awful experiences to 98% of its customers. The idea behind it was probably a fantastic one, but the quality of scaling obviously wasn’t.

I believe that your launch some depend entirely on what you’re trying to promote, although universally the biggest hurdle is the maintenance side of things. Your product is never complete and when you launch it is only probably 30% complete. Testing is the biggest hurdle of any software project, and the only ones that can perform quality testing is real end-users.

Gmail is probably the best example of a web service launch in history. They started out small, releasing an invite-only system that still allowed for many users to get accounts if they really wanted through invite lists. Throughout this time they spent countless hours improving the experience as users slowly grew. Despite it being in beta until the end of time the period has passed when Gmail has become a quality product with very few flaws. Not only that, but the invite-only scheme allowed users to get the email addresses they desired, instead of spam bots taking over.

The problem with the Hollywood Launch is that it packs all your testing time into the most frustrating week of your life. This simply isn’t healthy, neither is it safe for a business. If you were to launch a product to thousands of users and it were to fall over and die half-way through the day then your business may as well start from scratch.

Despite my criticisms, the Hollywood Launch and the Gmail Launch are mutually exclusive in the sense that hype was created, details were leaked and accounts were given slowly to users in an invite-only system where feedback was not only encouraged, it was demanded. In conclusion the best way to get a product off the ground is to purposely target the power users of similar tools. If you needed to launch a new email service it’s probably best to target businesses that churn their way through thousands of emails a day. It would’ve taken the Gmail team about ten years to find all the bugs in their software, but many of these were discovered within the first week of deployment.