Nobody Throws a 180 on Their First Visit: Give Your Estimates a Handicap
Photo: Sarah Ganderton, CC BY-SA 2.0, via Wikimedia Commons
Pub darts is a game of honest starting points. You do not walk into the Red Lion, pick up three arrows and announce that you are on 501. You start on 301. You might give the other player a handicap. You might play round the board. You might accept that the first visit is going to be mostly warm-up and mild embarrassment. Software estimates could learn a lot from this. Instead, every project begins at zero and then everyone acts surprised when it finishes somewhere else.
The fantasy of the blank scoreboard
A blank scoreboard is satisfying. It suggests a clean start. It makes the slide look decisive. 'Five days,' you say. 'Maybe six.' The room nods. The plan is approved. Then day one arrives with a legacy codebase, three meetings, an integration nobody mentioned and a security review that wants to know why the login page is held together with hope.
Zero is not a starting score. It is a polite fiction. Every project starts with a handicap already attached: the state of the code, the knowledge of the team, the number of stakeholders, the calendar, the dependencies and the general tendency of software to behave like a pub toilet — fine from a distance, alarming up close.
Why zero is a lie
Estimation at zero assumes perfect information and uninterrupted time. Neither exists. We know this. We still do it because a single number feels like confidence. It is not confidence. It is a wish with a deadline.
The planning fallacy makes it worse. We imagine the best case, add a small amount for safety and then spend the safety on the first unexpected thing. The estimate was never honest. It was optimistic with a straight face.
A handicap is not padding. Padding is a number you add because you are afraid. A handicap is a number you add because you have looked at the game and recognised its conditions. There is a difference. One is superstition. The other is method.
Three handicaps worth a beermat
First, the discovery allowance. Give the team time to understand the problem and the existing system before anyone promises a delivery date. Ten to twenty per cent of the effort is a reasonable starting point. If the team has never worked in that part of the codebase, make it more. Discovery is not a luxury. It is the bit where you find out the job is not the job you thought it was.
Second, the integration tax. Auth, payments, legacy data, third-party APIs, the one supplier who communicates by fax and the internal service that nobody owns. These are not edge cases. They are the game. Add days for each one you can name and a few for the ones you cannot. If the estimate does not include integration, it is not an estimate. It is a hope.
Third, the life-admin buffer. Teams do not work at theoretical capacity. They go to meetings, review each other's work, take holidays, get ill, answer support tickets and attend the occasional mandatory training session about fire safety. Use the team's real capacity, not the version in the spreadsheet. If you plan for eight hours of focus per person per day, you are planning for a team that does not exist.
How to write an honest estimate
Start with a baseline from the people doing the work. Not the people who are excited about the work. The people who will actually write the code. Then add the handicaps. Then give a range and a confidence level. 'Eight to twelve days, with seventy per cent confidence' is more useful than 'ten days' because it tells the listener what kind of number they are getting.
Name the assumptions. List the unknowns. Write them on the beermat next to the estimate. If the estimate is a single number with no assumptions, it is not a forecast. It is a dare.
Review it after the work. Compare the estimate with what happened. Look at where the handicap was too small or too large. That is how a team gets better at estimating. Not by being told to be more accurate, but by learning its own starting conditions.
The culture bit
Estimates are forecasts, not contracts. If you punish honesty, you get sandbagging or heroics. Both are expensive. The team that says 'we need a day of discovery before we can say' is not being difficult. They are being professional. The team that says 'it will be fine' is usually about to have a very bad week.
So give estimates a starting score. You are not on 501. You are on 301, with a dodgy set of arrows and a pint you are trying not to knock over. That is the game. Play it honestly and you will miss less often. You will still miss sometimes. That is darts.