Three-Point Estimation (PERT)
The estimation technique that makes uncertainty visible. Optimistic, most-likely, and pessimistic estimates, weighted into a single number — and a standard deviation that tells you how much you should trust it.
What Is Three-Point Estimation?
Three-point estimation — sometimes called PERT estimation, after the Program Evaluation and Review Technique it grew out of — is a structured way to estimate work that explicitly accounts for uncertainty. Instead of producing a single number, the team produces three: optimistic (O), most likely (M), and pessimistic (P). Those three numbers get combined with a weighted formula that gives more weight to the most-likely value.
The formula is simple: E = (O + 4M + P) / 6. The 4× weight on M reflects the fact that, in normal development, the most-likely outcome is much more probable than either extreme. The optimistic and pessimistic numbers serve as risk anchors, not as candidate answers.
Three-point estimation pairs well with planning poker rather than replacing it. Planning poker gives you a fast consensus story point estimate. Three-point estimation gives you that estimate plus a measure of how confident the team should be in it. Use planning poker for the bulk of your backlog and three-point estimation for the high-risk stories where uncertainty deserves an explicit number.
The Three Estimates
Optimistic (O)
The best-case effort, assuming everything goes right. No interruptions, no surprises, no missing context. The "if-the-stars-align" estimate.
Most Likely (M)
The realistic effort, accounting for normal friction. Some context-switching, a few minor surprises, the usual rate of small unknowns being discovered during implementation.
Pessimistic (P)
The worst-case effort, assuming things go wrong. Hidden complexity, cross-team dependencies, environment issues, and the bug that takes a day to reproduce.
The PERT Formula
Two formulas matter. The first gives you the expected estimate. The second gives you the standard deviation — how much the estimate could swing in either direction.
EXPECTED VALUE
E = (O + 4M + P) / 6
The weighted average. The most-likely value carries most of the weight, the extremes pull the result up or down based on risk asymmetry.
STANDARD DEVIATION
σ = (P − O) / 6
A wider gap between optimistic and pessimistic = more uncertainty. Stories with high σ deserve more refinement before sprint commitment.
Worked example
The team estimates a story: O = 3 points, M = 5 points, P = 13 points. E = (3 + 4×5 + 13) / 6 = 6.0. Round up to the nearest Fibonacci: 8 points. The standard deviation: σ = (13 − 3) / 6 ≈ 1.67 — moderate uncertainty. A story where O = 3, M = 5, P = 5 would have σ = 0.33, much tighter. Same expected value (4.83 → 5), very different risk profile.
How to Run a Three-Point Estimation Session
1Have the team understand the story
Before any estimate, the team reads the story, asks clarifying questions, and aligns on scope. This is the same prework as planning poker — without shared understanding, three-point estimation just produces three flavors of guess.
2Each person writes O, M, P privately
Anchoring bias hits PERT exactly like it hits planning poker. Each team member writes their three numbers without showing anyone. Use a digital tool, sticky notes, or just the chat with messages staged in drafts.
3Reveal and compare the spreads
Reveal everyone's O / M / P at once. Look at the spread. If one person has a P of 8 and another has a P of 40, that gap is the most valuable signal in the room — it points at risk somebody is seeing that nobody else is.
4Discuss the gaps and re-estimate
Like planning poker, the value lives in the discussion. The pessimist explains the worst case. The optimist explains why they are not worried. The team converges on a shared O / M / P after one or two rounds.
5Apply the PERT formula
Compute the weighted average: E = (O + 4M + P) / 6. The "4M" weight reflects that the most likely estimate is far more probable than the extremes. Round the result to a Fibonacci value if your team tracks story points.
6Capture the standard deviation (optional)
σ = (P − O) / 6 gives you a rough standard deviation. Stories with a high σ are high-uncertainty — flag them for risk discussion or refinement, not just sprint commitment.
When to Use Three-Point Estimation Instead of Planning Poker
Three-point estimation is heavier than planning poker. Six numbers per story (three per voter, then a weighted average) versus a single Fibonacci card. So use it deliberately, not for every story:
- High-risk or high-uncertainty stories where leadership wants a confidence band, not just a point estimate
- Cross-team work with external dependencies that could swing the timeline
- Architectural changes where the worst case is materially different from the typical case
- Major releases and milestones where you need to communicate uncertainty to stakeholders honestly
- Spike-adjacent work where the team explicitly does not yet know the scope
For routine sprint stories, stick with planning poker. The throughput advantage matters more than the extra precision when the stories are well-understood.
Common Mistakes to Avoid
Using P as a sandbagging value
The pessimistic estimate is the worst credible case, not the worst imaginable case. If your team is regularly seeing P values 5–10x larger than M, the team is using P to hedge against deadline pressure rather than to surface real risk.
Skipping the discussion
The numbers are not the deliverable. The discussion that explains the spread is. A team that just averages O/M/P without unpacking why one person had a 20-point P loses 80% of the value.
Treating E as a commitment
E is the expected value, not a deadline. The σ tells you how much variance is built in. Reporting "this story will take 8 points" without the uncertainty band misrepresents what the team actually said.
Doing PERT on every story
It is too heavyweight for routine work. Reserve three-point estimation for stories where uncertainty is material. Default to planning poker for the bulk of the backlog.
Forgetting to round to Fibonacci
If your team tracks velocity in Fibonacci points, an E of 6.0 is awkward. Pick the nearest Fibonacci value (5 or 8) so the estimate slots cleanly into your existing planning workflow.
Estimate the work. Estimate the uncertainty.
Three-point estimation makes the uncertainty in your story estimates explicit. Pair it with planning poker for the most honest agile estimation workflow your team can run.