Estimating: why every guess is wrong
How long will it take?
A publisher asks Kai: "When will Shard Garden launch?" Kai answers: "12 months."
Kai is not lying, and Kai is not bad at estimating. At this point nobody could know. An estimate is a best guess at time or cost, made with what is known today.
The cone of uncertainty
Construx, Steve McConnell's company, draws how far off skilled estimators can be at each point in a project. It is called the cone of uncertainty.
At the first idea, the real result can be 4 times the estimate, or a quarter of it (0.25x). That is a 16 times range: "12 months" could really be 3 to 48 months. The range shrinks as the project is defined: 0.5x to 2x, then 0.67x to 1.5x, then 0.8x to 1.25x.
Quick check
At the first idea, a skilled estimate says 10 months. What range does the cone give?
- 2.5 to 40 months
- 9 to 11 months
- Exactly 10 months, because the team is skilled
Show the answer
2.5 to 40 months
At the first idea the cone runs from 0.25 times to 4 times the estimate: 0.25 times 10 is 2.5, and 4 times 10 is 40.
It narrows only when you decide
Construx says: "The Cone narrows only as you make decisions that eliminate variability." Another week staring at the same unclear plan does not help. Deciding what the game is, including what it will not do, does.
And the cone is the best case. Construx again: "It isn't possible to be more accurate; it's only possible to be more lucky."
Three habits of good estimators
- Give a range"9 to 15 months" is honest. "12 months" sounds more certain than it is.
- Use your own historyCompare guesses with how long similar work really took. If tasks took 1.5 times the guess before, plan for that.
- Commit laterConstrux says firm commitments become possible about 30% of the way into a project, once work has narrowed the cone.
Quick check
Giving the estimators an extra week to think makes an early estimate much more accurate.
- True
- False
Show the answer
False
The estimate is uncertain because the project is. Only decisions that remove unknowns, like defining the features, narrow the range.
Read the numbers
1Task Guess Actual2Shop screen 4 63Tutorial 2 34Cloud saves 6 95Daily quest 2 3
Remy guesses 8 days for the friends list. Using this history, what should the plan say?
- About 12 days: each task took 1.5 times the guess
- 8 days: Remy is getting better
- 4 days: a friends list is simpler
Show the answer
About 12 days: each task took 1.5 times the guess
6 is 1.5 times 4, 3 is 1.5 times 2, and so on. Remy's own history says to multiply by 1.5, and 8 times 1.5 is 12.
Try it on Shard Garden
A range only becomes a promise when enough has been decided. The next lesson puts that to the test with Kai's 12 months.
Bolt says: Next: watch the cone narrow on Shard Garden's own estimate.
Fact-checked 12 October 2026
- Construx: The Cone of Uncertainty construx.com
- Construx: Cone of Uncertainty poster construx.com