3 Point Estimation Calculator for Software Testing

3 point estimation, also called three point estimating, is a technique that estimates a task from three numbers instead of one: an optimistic value (O), a most likely value (M) and a pessimistic value (P). The most common formula is the PERT formula, (O + 4M + P) / 6. The triangular version simply averages the three: (O + M + P) / 3. Use the calculator below to work out the expected value, standard deviation and confidence range for one task or a whole test plan.

What is 3 point estimation?

3 point estimation replaces a single guess with a range. Instead of saying a task takes 12 hours, you say it could take as little as 8 hours if everything goes well, most likely 12, and as much as 24 if things go wrong. The technique then combines the three numbers into one expected value and tells you how uncertain that value is.

It works because single point estimates hide risk. A team that says 12 hours has no way to show that the downside is twice the upside. A three point estimate keeps that information and turns it into a number you can plan with. In software testing, where environments fail, builds arrive late and defects cause retests, the gap between best case and worst case is usually large.

3 point estimation formula

There are two common ways to combine the three values.

Method	Expected value	Standard deviation
PERT (beta approximation)	(O + 4M + P) / 6	(P – O) / 6
Triangular	(O + M + P) / 3	square root of (O² + M² + P² – OM – OP – MP) / 18

• O is the optimistic estimate: the best case, assuming nothing blocks you.
• M is the most likely estimate: what usually happens.
• P is the pessimistic estimate: the worst realistic case, not a disaster.

The PERT formula gives the most likely value four times the weight of the extremes, so it stays close to M. The triangular formula treats all three values equally, so a long pessimistic tail pulls the answer up further.

How to calculate a 3 point estimate, step by step

Take a task where O is 4 hours, M is 6 hours and P is 14 hours.

1. PERT expected value: (4 + 4 x 6 + 14) / 6 = 42 / 6 = 7 hours.
2. PERT standard deviation: (14 – 4) / 6 = 1.67 hours.
3. Range at about 68% confidence: 7 plus or minus 1.67 gives 5.33 to 8.67 hours.
4. Range at about 95% confidence: 7 plus or minus 2 x 1.67 gives 3.67 to 10.33 hours.

The triangular method on the same numbers gives (4 + 6 + 14) / 3 = 8 hours. The one hour difference comes from the long pessimistic tail.

3 point estimation example for software testing

Here is a small test plan estimated with the PERT method, in hours. This is the example loaded in the calculator above.

Task	O	M	P	Expected	Std dev
Write test cases	8	12	24	13.33	2.67
Execute test cycle	16	24	48	26.67	5.33
Regression run	10	16	40	19.00	5.00
Defect retest	6	10	26	12.00	3.33
Total				71.00	8.47

The expected total is 71 hours. Task expected values simply add up. Standard deviations do not: you add the variances (the squares) and take the square root, which gives 8.47 hours. That is smaller than adding the four standard deviations, because it is unlikely that every task goes badly at once.

From there, the 95% range is 54.07 to 87.93 hours. If you want a figure you can commit to with 95% confidence of not overrunning, plan for about 85 hours, roughly 20% above the expected total. That gap is your contingency, and it comes from the data instead of a rule of thumb.

PERT vs triangular: which should you use?

• Use PERT when you trust your most likely value and want a result that stays near it. It is the standard in project management and the method most people mean by 3 point estimation.
• Use triangular when the pessimistic case worries you and you want the risk to show up in the expected value. It gives a higher answer whenever P is far from M.

If you are unsure, calculate both. When they differ a lot, that is a sign the pessimistic estimate is driving the risk and deserves a closer look.

Standard deviation and confidence levels

The standard deviation tells you how spread out the outcome could be. For a roughly bell-shaped result, about 68% of outcomes fall within one standard deviation of the expected value, about 95% within two, and about 99.7% within three. The calculator shows the 68% and 95% ranges, and also a planning estimate at the confidence level you choose, such as 80% or 95%. A higher confidence level gives a larger number and more contingency.

Choose the confidence level by the cost of being late. A release that cannot slip deserves 90% or 95%. An internal task with slack can use 50% or 70%.

How to choose good O, M and P values for testing

• Estimate the three values separately. Do not set M first and add and subtract a fixed percentage. The point of the method is that the range is not symmetrical.
• Optimistic means no blockers, not zero effort. Assume the build is stable and the environment works.
• Pessimistic should include the usual troubles: an unstable environment, late builds, blocked test data and defect retesting. Leave out genuine disasters.
• Use history. Look at past sprints or test cycles to see how long similar work really took.
• Estimate small tasks. Breaking a plan into tasks of a day or less makes each estimate more reliable and the total range tighter.

Common mistakes

• Padding P until the estimate is safe. This inflates the total and hides real risk.
• Setting O equal to M, which means you have no real best case. Check that the optimistic number is achievable.
• Adding standard deviations instead of variances. Combine task uncertainties with the square root of the summed variances.
• Treating tasks as independent when they are not. If one late build delays everything, the true spread is wider than the calculator shows.
• Estimating once and never revisiting. Update the numbers as you learn more.

When 3 point estimation works, and when it does not

It works well for work you can break down into tasks, such as test design, test execution, regression testing and user acceptance testing. It is weaker for work with no clear tasks yet or no history to base the three values on. For relative sizing of backlog items, teams often use planning poker instead, and some teams feed story point ranges into a three point estimate for a release.

3 point estimation FAQ

How do you calculate a 3 point estimate?

Estimate three values for the task: optimistic (O), most likely (M) and pessimistic (P). Then apply (O + 4M + P) / 6 for the PERT estimate, or (O + M + P) / 3 for the triangular estimate. The calculator above does this for every task and totals the result.

What is the PERT formula?

The PERT formula is (O + 4M + P) / 6. It gives the most likely estimate four times the weight of the best and worst cases. The matching standard deviation is (P – O) / 6.

What is the difference between PERT and three point estimating?

Three point estimating is the general idea of using three values. PERT is one way to combine them, using a weighted average. The triangular method is the other common way, using a simple average.

How do you calculate the standard deviation in 3 point estimation?

With PERT, the standard deviation is (P – O) / 6. With the triangular method it is the square root of (O² + M² + P² – OM – OP – MP) / 18. For a total across several tasks, add the variances and take the square root.

What does the planning estimate at 95% mean?

It is the expected total plus enough extra time that the work should finish within that figure about 95% of the time, if your three values are sound. It is a one-sided figure, so it adds contingency on the upside only.

What is the most accurate estimating technique?

No single technique is always the most accurate. Estimates improve most when you use historical data, break work into small tasks and give a range instead of one number. Three point estimation does the last of those.

Can I use 3 point estimation for story points?

Yes. Pick Story points as the unit and enter your optimistic, most likely and pessimistic values. Remember that story points are relative, so only compare ranges from the same team.

Is the calculator free, and does it save my data?

It is free. All calculations run in your browser, and nothing you enter is sent to or stored on a server. Use Download CSV to keep a copy of your estimate.

Need help estimating a testing project?

If you want a second opinion on a test estimate, you can book a free software test consulting session.