A record with no failures is a warning sign
A perfect record usually means the targets were set low. The reasoning works for a company placing many bets and fails for one placing a single bet.
If everything you do works, then you’re not taking many risks and probably aren’t innovating either.
Paul Buchheit wrote this in December 2010, in a blog post arguing that Google was still a good place to build things. At the time the company had just ended Google Wave, a heavily promoted collaboration product that few people used. Critics read the shutdown as a sign of decline. Buchheit, who had built Gmail inside Google before leaving to co-found FriendFeed, read it the other way: a company that never ships anything that fails is not trying anything hard. He added that failing at everything is no better, and that the right place is somewhere in between.
The logic is about expected value across many attempts. A new product has a small chance of becoming very large and a good chance of going nowhere. A team that only attempts safe projects will post a clean record and very little upside. A team that attempts ambitious ones will post visible failures, and a few of its successes will pay for all of them. Gmail itself was an odd project when it started, a free email service offering far more storage than anyone thought reasonable.
The rule is useful as a diagnostic for managers. A group that has hit every target for two years may be good. More often, it has learned to promise only what it knows it can deliver, and the targets have drifted down to meet the team.
It becomes misleading for a company that has only one bet. A start-up with eight months of cash cannot run a portfolio. If its single product fails, there is no Gmail on the other side of the ledger. For that company, reducing risk where possible is the job: testing demand before building, keeping costs low, choosing the least risky route to the ambitious goal. Buchheit’s line assumes the kind of balance sheet that can absorb Wave.
Failure is also a poor measure in its own right. Many projects fail because they were badly run or aimed at a problem nobody had, and those failures say nothing about ambition. The rate of failure is a signal worth watching only alongside what was attempted. A team that fails often at bold things is doing something right; a team that fails often at modest things is not.