I used to estimate too precisely.

Like anyone, I do a lot of estimation: how long a work-project will take, how expensive a trip will be, how long it will take until I feel confident about my tennis-serve, and so forth.

For work-projects, I used to say something like: "Based on estimates for the sub-tasks, it will take exactly 26 days."

There were two things wrong with that.

First, people are really bad at estimating, especially when the variables get complex. How can I predict that we'll discover a rare-but-urgent edge case that adds a week to the project? Or that someone will get sick?

So maybe I need to be less precise. Maybe instead of "26 days," I need to give an estimate of "around a month."

But what about those ~four days that we've glossed over?

Most of the time, they don't matter. My audience (manager, product manager counterpart, etc.) doesn't need or care about that level of precision.

So that's the second thing wrong with my example estimate. In addition to being impossibly precise (emphasis on "impossible"), I didn't consider how precise I actually needed to be.

These days, I like to start by asking: "how imprecise can I be for this estimate?" In many cases, that means to the nearest order of magnitude, or identifying if it will take days, weeks, or months.

Following that approach, I can be far more accurate. And what's more: I can avoid the many-flavored stresses of false-precision.