Scheduling when the durations are uncertain
When durations and arrivals are uncertain, the useful schedule is the one that stays good across scenarios, not the one that looks tidy on the plan that will not happen.
Most schedules are built as if the future were known: fixed durations, fixed arrivals, a tidy sequence. Then reality supplies variance, and the plan is re-drawn weekly. The re-drawing is not a failure of discipline. It is a signal that the plan was the wrong object. When durations and arrivals are uncertain, the useful schedule is the one that stays good across the scenarios you might face.
Planning against variance
We treat the schedule as a decision under uncertainty. Instead of optimising a single assumed future, we optimise expected performance across a set of scenarios, jobs arriving early or late, tasks running long, a resource dropping out, subject to the hard constraints that must never break (capacity, skills, precedence, time windows). The objective is risk-adjusted: not just expected throughput or margin, but the downside when things go wrong, because a plan that is excellent on average and catastrophic in the tail is not a plan you can run.
Showing the trade-off
A schedule that maximises utilisation is not the same as one that protects delivery, or one that preserves margin. Rather than collapse these into a single weighted score whose weights nobody agreed to, we return a small set of Pareto-relevant plans: this one protects the deadline at the cost of some idle capacity. That one runs hotter but risks a late penalty. The operator still chooses. The method makes the choice legible.
What this is
This is a description of how we approach the problem, labelled as an idea rather than a measured result. Whether a stochastic plan beats a good deterministic heuristic is an empirical question for a specific operation: one a proof of value answers against that operation’s own baseline, not a claim we make in advance.
Limitations & where this does not apply
This note describes an approach, not a benchmarked result on a specific dataset. Whether a stochastic schedule beats a good deterministic heuristic depends on how much uncertainty there is and how costly disruption is, both of which must be measured for a given operation, not assumed.
Commercial relevance
This is the method behind our capacity, scheduling and acceptance work: build a plan that survives the variance you face, and expose the trade-off rather than hiding it in one score.
Cite this
Flexiana Decision Lab (2026). Scheduling when the durations are uncertain. Flexiana Research. https://flexiana.com/research/scheduling-under-uncertainty