CPM Schedule Risk Modeling — Special Considerations

Why schedule risk needs special care

Cost risk simply adds up — sum the items and you have the total. Schedule risk doesn't, because a schedule is a network: activities run in parallel, paths merge, and the finish date is governed by whichever path happens to be longest. That network structure creates effects with no equivalent in cost modeling — and the most important of them, merge bias, systematically makes naive estimates too optimistic.

Merge bias, illustrated

Imagine two parallel paths that must both finish before a milestone, each with a 50/50 chance of being on time. Intuitively you might hope for an on-time milestone — but the milestone is only on time if both paths are. That's 0.5 × 0.5 = just 25%.

The pitfalls to guard against

• Ignoring correlation — If activities affected by a common cause are modeled as independent, their variations wrongly cancel out — drastically understating total schedule risk. (The risk-driver method of Lesson 16 fixes this at the source.)
• Near-critical paths — A path with little float today can become the critical path once risk lengthens it. Risk-analyze the whole network, not just today's critical path.
• Poor schedule logic — Missing links, dangling activities, and excessive constraints break the network's response to risk — the simulation faithfully amplifies the flaws.
• Hard constraints & lags — "Must finish on" dates and large lags can mask risk by holding activities artificially in place, hiding slips the project would really suffer.
• Over-optimistic duration ranges — Three-point ranges set too narrow (or symmetric) understate risk — the same skew and optimism traps from the distributions lesson.

Doing schedule risk right

Start with a clean, logic-driven CPM schedule — sound links, minimal hard constraints, realistic durations. Model correlation explicitly (the risk-driver approach is the cleanest way), risk-analyze the whole network so near-critical paths can surface, and set duration ranges that respect real-world skew. Then read the finish-date distribution knowing merge bias is properly captured — and fund schedule contingency to a chosen confidence level, just as you do for cost.

Ten things to remember

1. Schedule risk isn't additive — the network structure changes the math.
2. Merge bias pushes real finish dates later than naive estimates.
3. Two 50% paths merging → only 25% on time — optimism compounds.
4. Ignoring correlation understates total schedule risk badly.
5. Near-critical paths can become critical — analyze the whole network.
6. Poor logic and hard constraints distort the simulation.
7. Over-narrow duration ranges understate risk — respect skew.
8. Most pitfalls trace back to schedule quality — fix the schedule first.
9. Monte Carlo captures merge bias — a deterministic CPM never does.
10. Fund schedule contingency to a confidence level, like cost.