Merge Bias

Merge bias is a statistical phenomenon that occurs wherever two or more parallel activity paths converge on a single successor task or milestone. Because the merged activity cannot start until all of its predecessors are complete, the actual duration to that point is driven by the longest path — not the average path. When paths are roughly equal in length, the probability that at least one of them will be late is much higher than the probability that any single path will be late. Standard deterministic scheduling tools ignore this effect entirely.

Merge bias matters because it means that deterministic schedules consistently underestimate duration at points where paths converge. The more parallel paths converge, the more optimistic the deterministic date becomes. This is one of the main reasons that Monte Carlo-based Schedule Risk Analysis (SRA) will often produce a P50 date that is later than the deterministic critical path end date, even before any specific risks are applied. Understanding merge bias helps explain to clients and stakeholders why a probabilistic schedule analysis gives a later expected completion than the baseline Gantt.

In practice, the best way to account for merge bias is to run a proper Monte Carlo simulation rather than relying solely on the deterministic schedule. However, even without simulation, planners can sense-check a schedule by identifying near-critical paths — those with small amounts of float — and considering whether the convergence of several near-critical paths creates a realistic risk of delay that the headline programme does not reflect. Programmes with many parallel workstreams all converging on a single commissioning milestone are particularly vulnerable to merge bias.