I asked Google to stop shipping

By 2019, Family Link was growing about five times over year on year — and buckling under it. The product had been assembled by more than thirty engineers, most of whom had moved on. Critical journeys broke, complaints piled up, a few outages made the news, and parents — who need this kind of product to simply work — started to doubt the basics.

So I went into leadership reviews and proposed the opposite of everything a growth org is built to do: freeze the feature roadmap for three quarters and spend them making the product trustworthy. The hardest part wasn’t the engineering. It was that no agreed definition of “reliable” even existed — I had to invent the exit criteria, then persuade a working group and the org’s leads to bet on them.

Convincing a growth org to stop shipping is a wager on a number that doesn’t exist yet.

We did it the unglamorous way. SLOs and real monitoring where we’d been running blind and learning about outages too late. An on-duty rotation. A postmortem culture instead of a blame one. Over three quarters, support complaints fell roughly 70% even as the user base grew about 500%, and the overdue-P1 backlog shrank from dozens to a handful. The practices we wrote spread to around ninety engineers; my own shorthand for it was that the org finally moved out of “startup mode.”

What I took from it: reliability is a feature you can’t market, and the willingness to stop is a leadership act, not a failure to ship. For a product whose entire value is dependability, trust is the only moat there is — and you earn it back the same slow way you lost it.