The Cathedral vs. the Bazaar, and the Lean Startup Methodology

by Helen Weng

In computer science, there is a famous essay written by Eric Raymond on two software development models. The Cathedral model is customized, constrained, with source code tightly controlled. Microsoft comes to mind. The Bazaar model is essentially open source software and Raymond argues that when source code is open for public scrutiny and experimentation, more bugs are discovered and fixed in contrast to the Cathedral model. More information about his essay can be found here.

I’ve been thinking about the Cathedral and Bazaar in relation to startup business models. The model of a tightly controlled, end to end solution applies to startups of product and process innovation. Apple, Warby Parker, and Bonobos are Cathedrals. Bazaars on the other hand, are open marketplaces like Etsy and Aardvark, or business process innovations to existing products like Rent the Runway. These companies do not control the entire ecosystem, but rather facilitate that ecosystem in a better or different way. When looking at Cathedrals vs. Bazaar business models, is the lean startup methodology more applicable to one vs. the other?

When building a Cathedral, the true customer experience relies on all the pieces that go into building the business to work perfectly together. Warby Parker manufactures, distributes, and sells fashionable and less expensive eyewear over the Internet. The founders spent 1.5 years working on their business plan and researching the market before launching a completed website with heavy PR. They hit their first year sales target in 3 weeks and acquired a 20,000 person waitlist. However, they did not test small hypotheses prior to launch because their core value proposition depended on a seamless integration of all parts of the product development and selling process. Perhaps the lean startup methodology is less relevant for Cathedrals. Sometimes, truly testing key hypotheses without the completed business is impossible.

Bazaars rely on third parties or customers to fulfill certain parts of their business mission. Rent the Runway wanted to rent designer dresses over the Internet but relied on established designers to supply dresses which they knew would be in high demand. Thus, because they did not fully control their product, they could test small hypotheses around the process, such as whether women would rent dresses, rent without trying on dresses, and return rates. Aardvark relied on users to ask and answer questions, so could conduct “smoke screen” tests to experiment with features or validate new ideas. Similar to open source software, a Bazaar is more easily testable because small pieces of the model do not ultimately depend upon a closed, end to end ecosystem.

To launch a Cathedral right seems to call for a fat startup mentality, otherwise a founder risks failing to fully test out a core hypothesis. Cathedrals are inherently more capital intensive and risky. Perhaps that is why, similar to a growing proliferation of open source development in software, more Bazaar business models are being launched today.