How to Estimate Task Durations (Without Guessing)
Every schedule stands on its duration estimates, and most of them are a single hopeful number typed in under pressure. That is why so many plans slip: not because the work took longer than it should have, but because nobody estimated it honestly in the first place. Estimating is a skill with real methods behind it. Here are the three that matter, when to reach for each, and how to turn a wild guess into a defensible number.
Why a single number is the problem
Ask someone how long a task will take and you get one figure: "three days." It hides everything useful. Is that three days if all goes well, or three days on a normal run, or the worst case? A single estimate quietly assumes the best case and then gets treated as a promise. When reality lands in the middle, the schedule is already late.
Good estimating does two things a single number cannot: it makes the uncertainty visible, and it uses more than one person's gut. The methods below are just structured ways to do that.
Analogous estimation: start from the past
The fastest method: find a similar task you have done before and adjust. "The last onboarding flow took eight days; this one is a bit bigger, call it ten." It is quick, needs little detail, and is surprisingly accurate early on, when you have nothing else.
Its weakness is the analogy itself. If the past project differed in a way that matters and you did not notice, the estimate inherits the error. Use analogous estimation for a first pass and for the top level of a plan, then refine the important tasks with something sharper.
Parametric estimation: estimate by the unit
When the work scales with something countable, estimate the rate and multiply. If writing takes about half a day per page and the manual is 20 pages, that is 10 days. If laying tile runs a fixed time per square metre, measure the floor and multiply.
Parametric estimates are as good as the rate behind them, so they reward keeping records of how long a unit actually takes. Where a reliable rate exists, this is the most objective method you have, because two people estimating the same task will land on the same number.
Three-point estimation and PERT
The method that makes uncertainty explicit. Instead of one number you give three: optimistic (O), most likely (M) and pessimistic (P). The expected duration is a weighted average that leans on the most likely case:
Expected = (O + 4M + P) / 6
This is the PERT formula. The extra weight on M keeps the estimate realistic, while O and P pull it toward honesty when the range is wide. A task with O=2, M=4, P=9 has an expected duration of 4.5 days, not 4, because the long tail of the pessimistic case deserves to count.
A worked example
- Optimistic O = 2, most likely M = 4, pessimistic P = 9.
- Expected = (2 + 4x4 + 9) / 6 = (2 + 16 + 9) / 6 = 27 / 6 = 4.5 days.
Which method to use when
| Situation | Best method |
|---|---|
| Early, little detail, similar past work | Analogous |
| Work scales with a countable unit | Parametric |
| High uncertainty, task matters | Three-point / PERT |
| A routine task you know cold | A single estimate is fine |
Most real plans mix all four. Estimate the whole thing analogously, then spend your three-point effort on the handful of tasks that are both uncertain and on the critical path, where an underestimate does the most damage.
Turning estimates into a schedule
Once you have durations, they become the length of each bar on your Gantt chart. Add dependencies and the tool computes the finish date and the critical path for you. Keep a note of which estimates were confident and which were rough, so when something slips you know whether the plan was wrong or the work was.
- Estimate each work package with the lightest method that fits.
- Use three-point for anything uncertain and important.
- Enter the expected durations as your bar lengths in the gantts.app editor.
- After the project, compare estimate to actual and keep the numbers, that record is what makes next time faster.
The short version
Do not type one hopeful number and call it a plan. Use analogous estimates to get moving, parametric where a rate exists, and three-point estimation to make uncertainty honest on the tasks that matter. The PERT average, (O + 4M + P) / 6, costs a minute and buys you a schedule that survives contact with reality.
Templates that use this
Keep reading
- Critical Path Method (CPM): Steps, Formula & Example
- What Is Float (Slack) in a Project Schedule?
- How to Make a Gantt Chart: Step-by-Step Guide
Frequently asked questions
What is three-point estimation?
Three-point estimation replaces a single guess with three values: optimistic (O), most likely (M) and pessimistic (P). The expected duration is the weighted average (O + 4M + P) / 6, which keeps the estimate realistic while accounting for uncertainty.
What is the PERT formula for duration?
PERT expected duration = (O + 4M + P) / 6, where O is the optimistic time, M the most likely, and P the pessimistic. The most likely value carries four times the weight, so the estimate stays grounded but still reflects the range.
What is the difference between analogous and parametric estimation?
Analogous estimation adjusts from a similar past task ("like last time, but bigger"). Parametric estimation multiplies a known rate by a quantity (half a day per page x 20 pages). Parametric is more objective where a reliable rate exists; analogous is faster when it does not.
Which estimation method is most accurate?
It depends on the task. Parametric is most objective when you have a solid rate. Three-point is best for uncertain, important tasks because it makes the range explicit. Analogous is best early on. Most plans use a mix.
How do duration estimates fit into a Gantt chart?
Each estimate becomes the length of a task bar. Once bars have durations and dependencies, a Gantt tool calculates the finish date and the critical path automatically, so better estimates directly mean a more reliable schedule.