Launch small, listen closely, then go again

In May 2009 Google showed Wave to a cheering developer conference, with a demonstration that ran well over an hour and a waiting list whose invitations people begged each other for. In August this year Google said it would stop developing Wave as a standalone product. The audience had been enormous from the first day, and very few of the people in it could say what the thing was for.

Jon Yongfook Cockle, a founder who has been building small web products, describes the opposite path. Put a working version in front of a small number of people who enjoy trying new things, watch what they do and listen to what they say, fix what matters, and release again. The quiet part is easy to skim past. It carries much of the value.

Quiet matters because a loud launch spends attention that cannot be spent twice. Press coverage, a launch post, a mention on a popular blog: each brings a crowd once. If the product is still rough, most of that crowd leaves with a poor impression and does not return when it improves. A quiet release holds that option in reserve for a version that deserves it.

Early adopters are the right first audience because they tolerate bugs and describe problems in detail. They also skew the picture. They are unlike the mainstream customers a company eventually needs, a gap Geoffrey Moore described in Crossing the Chasm back in 1991. A product tuned only to their requests can end up too complicated or too narrow for anyone else. At some point the cycles have to include people who are not enthusiasts.

The cycles are the part most teams get wrong. Feedback is useful only if the next release follows quickly enough to test what was learned. A team that gathers comments for six months and then ships once has not run a cycle; it has run a survey. Eric Ries has been writing about this rhythm on his blog since 2008, and Steve Blank’s customer development method describes much the same loop from the sales side.

There are markets where quiet does not work. A marketplace or a social product with no one in it is useless, so it needs a crowd on the first day. The usual answer is to be loud in one small place rather than quiet everywhere: Facebook opened only at Harvard in 2004, and Yelp concentrated on San Francisco before other cities. Hardware and regulated products are another exception, because each cycle costs months and real money, so the early rounds happen with prototypes behind closed doors.

Apple is the famous counterexample. The iPad was unveiled on a stage in January this year with no public beta at all, and it sold millions within a few months. The cycles still happened; they happened inside the company, with a secrecy few startups could afford and a brand none of them have.

For a small team the advice reduces to a schedule. Choose the group of early users by name. Set a release interval short enough to keep them interested, every two weeks or once a month. Before each release, write down the one question it is meant to answer, and after it, write down the answer. The launch party can wait until the answers stop surprising anyone.