Every Monday, we'd spend an hour estimating the week's work.
Story points, t-shirt sizes, hours — we tried them all. We'd discuss each task, debate the complexity, assign a number. Then we'd start working, and by Wednesday, the estimates were already wrong.
This wasn't because we were bad at estimating. We were fine at estimating. The problem was that estimation assumes a level of predictability that doesn't exist in most software work.
You think a task will take two days, and then you discover a dependency you didn't know about. You think something is simple, and then you find a bug in a library. You think you understand the requirements, and then you talk to a user and realise you don't.
The estimates weren't wrong because we were careless. They were wrong because the future is unknowable, and pretending otherwise is a kind of theatre.
So we stopped.
Not dramatically. We didn't announce a new methodology. We just stopped asking "How long will this take?" and started asking "What's the smallest thing we can ship?"
The shift was subtle but significant. Instead of estimating a large feature, we'd break it into pieces small enough that estimation felt pointless. If something could be done in a day or two, we'd just do it. If it couldn't, we'd make it smaller until it could.
This changed how we thought about work. Large projects became sequences of small experiments. Each one taught us something. Each one could be stopped or redirected without much waste.
The people who wanted estimates were usually trying to answer a different question. They wanted to know when something would be done, or whether it was worth doing at all. These are reasonable questions, but estimates are a bad way to answer them.
When will it be done? We don't know, but we'll show you progress every few days, and you can decide if we're moving fast enough.
Is it worth doing? Let's build the smallest version and find out.
This required trust. It required stakeholders who were comfortable with uncertainty, and a team that delivered consistently enough to earn that comfort. Not every environment can support it.
But in the environments that can, estimation starts to look like what it often is: a ritual that makes everyone feel better without actually reducing uncertainty.
We stopped estimating, and nothing bad happened. The work still got done. The important things still shipped. We just stopped pretending we knew more than we did.