# Estimates *Know which number you're asking for.* --- Your team says a feature will take two months. You build it into the plan. You tell the board it ships mid-quarter. Three and a half months later it's done. The team thinks they did well - they hit a snag with the integration but recovered quickly. You're frustrated - you're a quarter late. The board is annoyed because they committed your forecast to investor documents. --- Your team gave you their honest middle estimate. Half the time a project like this finishes faster, half the time slower. Two months is the point where they'd bet even odds. That's a reasonable number to give you. But think about the shape of what could happen either side of it. Could they finish in a month? Almost impossible - there's a floor on how fast the work can go. Could they take five months? Easily - a dependency slips, a requirement changes mid-build, a key person is pulled onto something urgent. There's no ceiling on delay. That asymmetry means the average (mean) is always higher than the middle (median) - and the gap is usually bigger than you expect. The early finishes save you a few weeks. The late ones cost you months. 1.5x isn't a bad rule-of-thumb. >*For those so inclined: task durations tend to follow a lognormal distribution - bounded on the left, unbounded on the right. The ratio of mean to median depends on how uncertain the work is. For routine tasks, multiply the team's middle estimate by around 1.2. For uncertain or novel work, multiply by 1.5. The less you know about the work, the wider the gap between the number your team gives you and the number you should plan against.* --- The project that started this piece wasn't a failure. Two months was the right estimate for the team. Three months was the right number for the board. The only problem was that one number did both jobs. Your team needs the middle estimate - two months. That's the number everyone coordinates around. QA blocks their calendar, marketing pencils in a launch date, downstream work gets scheduled. If you told them three months "to be safe," they'd plan for three and take three. The working plan needs to be the honest middle. But the board needs the average - three months. Across a year of projects, this is the number that makes the budget land roughly on plan. Some projects will finish in six weeks. Some will take four months. The average absorbs both, and that's what the commitment should reflect. The difference between "end of May" and "end of June" might be the difference between hitting a quarterly number and missing it. --- "The team's working estimate is two months. I'm budgeting three." That's not sandbagging. That's two different numbers doing two different jobs.