Build the rough version the technology allows today
Dodgeball ran on text messages and early Foursquare trusted its users. Both approximations taught more than waiting would have.
If there’s something you want to build, but the tech isn’t there yet, just find the closest possible way to make it happen.
Before phones knew where they were, Dennis Crowley was already building a service about where people were. Dodgeball, which he started with Alex Rainert in 2000, ran on text messages: a user texted the name of the bar they had just walked into, and friends nearby received a text saying so. There was no map, no app and no location sensor. Google bought the service in 2005 and closed it in 2009, and by then Crowley had moved on to the idea again with Naveen Selvadurai.
Foursquare launched in March 2009 and made the same kind of compromise. Phones had GPS by then, but it was slow and often wrong indoors, so the app asked people where they were and mostly took their word for it. That left room for cheating, and in 2010 the company began comparing a check-in with the phone’s reported position before handing out points. The approximation came first, the precision later, and the years in between produced the thing that mattered most: a record of how people actually used a check-in.
That is the case for the advice. An approximation gets a product in front of people while the ideal version is still impossible, and real use answers questions no planning session can. Whether people want to broadcast their location at all, which places they share, what makes them come back: Crowley learned those from Dodgeball users years before the hardware caught up.
The advice fails when the approximation becomes the product’s reputation. Apple’s Newton, released in 1993, relied on handwriting recognition that was not good enough, and the mistakes became a running joke that outlived every later improvement. Three years later the Palm Pilot took a different route to the same problem. It asked users to learn Graffiti, a simplified alphabet the device could read reliably. Both were closest possible ways. Palm’s worked because it was honest about the limit and designed around it, while Newton’s promised the finished thing and delivered a rough one.
The line comes from Crowley’s advice to founders as it circulated in 2011; its original setting could not be pinned down. The test it implies is simple enough. Pick an approximation that is modest about what it does, and let people use it before the technology arrives.