Date Defense Kit · an interactive worked example

How uncertain task durations change the finish

A small software team must finish discovery, integration and testing before a pilot. Enter three duration estimates for each task, then compare possible finishes under those assumptions.

What does a finish percentile mean?

P80 is a duration at or below which about 80% of the simulated runs finish. It describes this model and these inputs; it is not an assurance about a real delivery date.

How to use this example · assumptions and limits
  1. Inspect the sample and set a start date if you want calendar dates. Dependencies are finish-to-start IDs, separated by commas or semicolons.
  2. Enter optimistic (O), most likely (M) and pessimistic (P) durations, with 0 ≤ O ≤ M ≤ P. Run a baseline.
  3. Read the distribution and criticality table. A task is critical in a run when it has zero total float in that run.
  4. Try reducing the pessimistic estimates of the top three positive scores. The comparison lists exactly what changes and keeps your original inputs.

Every task duration is sampled independently. Beta-PERT uses a beta distribution rescaled to [O,P], with shape parameters 1 + 4(M−O)/(P−O) and 1 + 4(P−M)/(P−O). Triangular sampling uses the same bounds and mode. Equal estimates give a fixed duration. The seed makes a run repeatable; it does not make its assumptions more reliable.

Dates round duration up to whole elapsed days, excluding the start instant. Working days skip Saturdays and Sundays; holidays, shifts, resources, costs, lags, correlations and unlisted risks are absent. The salience score is criticality × (P−O), a discussion prompt rather than a computed marginal benefit. P90−P50 is a percentile gap, not a recommended project or feeding buffer.

The GAO Schedule Assessment Guide explains schedule-risk analysis and its input requirements. For a reference to the chosen distribution, see RiskAMP’s Beta-PERT explanation. Explore the omitted dependence assumption in task delays that happen together.

1. Define work and uncertainty

Task durations and finish-to-start dependencies. Horizontal scrolling keeps all fields available on a narrow screen.
IDNameGroupOMPDepends onAction
Bulk format and saved data

Use ID, Name, Group, O, M, P, DependsOn. Separate dependency IDs with semicolons. A header row is optional; quoted CSV fields and tabs are supported.

ID,Name,Group,O,M,P,DependsOn
T1,Scope spike,Discovery,1,2,4,
T2,Data contract,Partner,2,3,8,T1

Save stores valid inputs, settings and your narrative under the earlier app's browser key. Export includes those fields too; older task-only JSON still imports. Results are recalculated. Share links contain your inputs and settings in their fragment: anyone receiving the link can read them. Nothing is uploaded by this app.

2. Simulate the finish

No simulation yet.

3. Compare a what-if

Reduce P for up to three positive salience scores, keeping P at least M. Each comparison starts from the latest baseline and uses the same seed and sampling method.

Run a baseline first.

An editable account of the result

A new completed run replaces this paragraph. Saved edits remain in this browser.

Paste tasks

The entire table is replaced only after every row and dependency validates.

Share this plan