Estimation
What is Estimation?
Estimation is forecasting the time, effort or cost a piece of work will need, from rough roadmap sizing to sprint-level detail. Uncertainty is highest early and people tend to underestimate, so ranges help.
Updated November 2026 · 3 min read
Estimation is the practice of predicting how much time, effort or money a piece of work will take, so that a team can plan, prioritize and make commitments. In product and engineering work it ranges from a quick "is this days or months?" judgment for a roadmap idea to detailed task-level estimates for the next sprint.
Estimates are always uncertain, and the uncertainty is largest at the start of a project, when the least is known. Barry Boehm described this pattern in the early 1980s, and Steve McConnell later popularized it as the "cone of uncertainty." Psychologists Daniel Kahneman and Amos Tversky named a related bias, the planning fallacy: people tend to underestimate how long their own tasks will take even when they know similar tasks ran late before. Good estimation practice is designed around these two facts.
Common estimation techniques
| Technique | How it works | Best for |
|---|---|---|
| Expert judgment | Experienced people estimate based on what they know | Quick, early sizing |
| Analogous (top-down) | Compare with a similar past project and adjust | Roadmap and budget planning |
| Bottom-up | Break work into small tasks, estimate each and add them up | Detailed plans for well-understood work |
| Three-point (PERT) | Combine optimistic, most likely and pessimistic estimates | Tasks with significant uncertainty |
| Relative sizing | Compare items with each other using story points or T-shirt sizes | Agile backlogs |
| Probabilistic forecasting | Use historical throughput and simulation (such as Monte Carlo) to forecast a range of dates | Teams with stable delivery data |
Three-point estimation: A hypothetical example
The PERT formula weights the most likely estimate: expected time = (optimistic + 4 × most likely + pessimistic) ÷ 6. A rough measure of spread is (pessimistic − optimistic) ÷ 6.
Consider a hypothetical team estimating a data migration. Its optimistic estimate is 4 days, the most likely is 6 days and the pessimistic is 14 days because an unfamiliar legacy system might need cleanup. The expected time is (4 + 24 + 14) ÷ 6 = 42 ÷ 6 = 7 days, and the spread is (14 − 4) ÷ 6, about 1.7 days. Rather than promising "six days," the team communicates "about seven days, likely between five and nine," which matches the risk far better.
Estimation in Agile teams
Agile teams usually estimate in relative terms. Instead of hours, they size backlog items against each other with story points, often using Fibonacci agile estimation and a consensus technique such as planning poker. Over a few sprints, the team's velocity converts points into a forecast of how much work fits in a sprint or release. Some teams go further with the #NoEstimates approach, splitting work into similarly small items and forecasting from the count of items completed per week rather than estimating each one.
How to improve estimates
- Estimate as a team. The people doing the work should estimate it, and differing views reveal hidden assumptions.
- Break work down. Smaller items are easier to estimate accurately. Anything estimated at more than a couple of weeks deserves splitting.
- Use ranges. A range communicates uncertainty honestly; a single date implies false precision.
- Compare with actuals. Track estimates against outcomes and look for patterns, such as integration work always running long.
- Re-estimate as you learn. Update forecasts when new information arrives instead of defending the original number.
Common pitfalls
- Treating estimates as commitments. When estimates become deadlines, people pad them or stop giving honest numbers.
- Anchoring. Hearing a manager's hoped-for number first pulls everyone's estimate toward it. Estimate independently before discussing.
- Ignoring non-coding work. Testing, reviews, deployment and coordination often take as long as the build itself.
- Comparing story points across teams. Points are relative to one team's scale and are not a productivity measure.
Estimates feed sprint planning, release forecasts and capacity planning. They are also one half of most prioritization methods, such as the RICE framework, which divides expected value by estimated effort.
Frequently asked questions
How do you calculate a three-point PERT estimate?
Use expected time = (optimistic + 4 × most likely + pessimistic) ÷ 6, with (pessimistic − optimistic) ÷ 6 as a rough spread. For a task estimated at 4, 6 and 14 days, the expected time is 42 ÷ 6, or 7 days, with a spread of about 1.7 days, so the team might say about seven days, likely five to nine.
Why are software estimates usually too low?
Software estimates run low partly because of the planning fallacy, a bias named by Daniel Kahneman and Amos Tversky: people expect their own tasks to go smoothly even when similar tasks ran late before. Uncertainty is also highest at the start of a project, the pattern known as the cone of uncertainty. Giving estimates as ranges and using historical data counter both effects.
How do agile teams approach estimation?
Agile teams usually estimate in relative terms, sizing backlog items against each other with story points or T-shirt sizes instead of hours. Many use a Fibonacci-style scale and a consensus technique such as planning poker. After a few sprints, the team's velocity converts points into a forecast of how much work fits in a sprint or release.
Related terms
Story Points
A relative measure of the size of work, used instead of hours.
Fibonacci Agile Estimation
Story point sizing on a scale whose steps widen as items grow.
Planning Poker
A card-based team game for estimating backlog items without anchoring.
Sprint Planning
The meeting that starts each Sprint by setting a goal and a plan.