Ask most software teams what they track and you'll get a list: velocity, burn-down, story points, hours logged, sprint completion rate. We used to track some of that too. Then we noticed almost none of it predicted whether a project actually went well.
The metric that survived
The one number that's stuck around is embarrassingly simple: how long does it take a client to reply to a message asking them to look at something. Not how fast we build. How fast they respond once there's something real to react to.
When that number is short — same day, sometimes same hour — projects move fast and end well almost every time, regardless of how technically messy the work turned out to be underneath. When it stretches to a week or more, something is usually wrong, and it's rarely a coincidence.
What a slow reply actually means
A slow reply is almost never about being busy. It usually means one of three things: the client doesn't trust that what we built is worth reviewing carefully, the person we're talking to isn't actually the person who needs to sign off, or the scope drifted somewhere along the way and nobody wants to be the one to say it out loud.
All three of those are fixable, but only if you notice them. If you're only watching your own team's output, you'll miss every one of them, because your team can look perfectly productive while the actual project quietly stalls on the other side.
Why we stopped watching internal velocity
Internal metrics measure whether we're busy. They don't measure whether the thing we're busy building is the right thing, and busy-but-wrong is worse than slow-but-right, because it costs more to unwind later. Response time is the closest proxy we've found for "is this project actually on track," because it's a signal that comes from outside our own assumptions about our own work.
What we do differently because of it
We ask for a specific, small reaction early — not "here's the finished thing," but "here's five minutes of it, does this feel right." A short, specific ask gets a short, specific answer, and that answer either confirms we're on track or tells us exactly where we went sideways, days before it would've shown up any other way.
We still ship. We still estimate. But if I had to keep exactly one number on a whiteboard, it wouldn't be anything about our own output. It'd be that one.