Skip to main content

Resource Leveling

Resource Optimization · Schedule Smoothing

Peak-to-trough labor swings of 40–60% between weeks cost contractors €800–€2,500 per worker in hiring/layoff overhead, with productivity losses of 15–20% from constant crew turnover. Every month, a project superintendent must smooth the resource histogram — shifting non-critical activities within their float to minimize peak demand. This is the resource leveling problem — computationally hard when combined with precedence and deadline constraints.

Where This Decision Fits

Construction operations chain — the highlighted step is what this page optimizes

Site SelectionLocation & permits
Portfolio SelectionProject prioritization
Project SchedulingCPM / RCPSP
Resource LevelingSmooth labor histogram
Site OperationsExecution & control

The Problem

Why resource leveling matters in construction

Given a scheduled project with 12 activities, precedence relationships, and float windows, the superintendent must shift activities within their slack to minimize peak resource usage (workers per day). Activities cannot violate precedence constraints or extend the project deadline.

Unleveled schedules create sharp peaks and valleys in the resource histogram. Peaks force expensive overtime hiring or subcontracting, while valleys leave crews idle. The resource leveling problem seeks a schedule that keeps daily resource demand as uniform as possible — reducing peak-to-average ratios and smoothing the workforce over the project duration.

This problem is NP-hard in the general case: even with a fixed project deadline, finding the optimal shift for each activity within its float is computationally intractable for large networks. The classic objective, proposed by Burgess & Killebrew (1962), minimizes the sum of squared daily resource usage — a convex proxy that penalizes peaks quadratically.

Resource Leveling Formulation minimize   Σt r(t)²   // sum of squared daily resource usage
subject to
  ESi ≤ si ≤ LSi    // activity i starts within its float window
  si + di ≤ sj    // precedence: i finishes before j starts
  r(t) = Σi active at t wi    // daily resource usage

Where si is the start time of activity i, di is its duration, wi is its daily worker requirement, ESi / LSi are the earliest/latest start from CPM, and r(t) is total resource demand on day t.

See RCPSP theory and scheduling algorithms

Try It Yourself

Level the resource histogram for a construction project

Resource Leveling Optimizer

12 Activities · Target Peak 18
ActivityDurationWorkersESLSFloatPredecessors

Ready. Click “Solve & Compare All Algorithms” to run.

AlgorithmPeakΣr²Time
Schedule details will appear here after leveling.

The Algorithms

Approaches to resource leveling

Optimization

Burgess-Killebrew

O(n · T · F) per iteration

The classic resource leveling heuristic (1962). For each activity with float, evaluate all possible start times within its float window and choose the one that minimizes the sum of squared daily resource usage. Iterate until no further improvement is found. The squared objective penalizes peaks quadratically, naturally smoothing the histogram toward uniformity.

Heuristic

Greedy Shift

O(n · T)

A fast single-pass heuristic that identifies the highest-peak day, then shifts the activity contributing most to that peak as late as possible within its float. Repeat until no further peak reduction is possible. Simple and fast, but may get stuck in local optima since it only considers one activity at a time.

Metaheuristic

Simulated Annealing

Iterative · O(iterations · n · T)

A probabilistic search that randomly shifts activities within their float windows. Improvements are always accepted; worse solutions are accepted with decreasing probability as the “temperature” cools. This allows escape from local optima and often finds solutions superior to greedy approaches, especially on larger networks with many interdependent float windows.

Key References

Foundational works in resource leveling

  • Burgess, A. R. & Killebrew, J. B. (1962). “Variation in activity level on a cyclical arrow diagram.” Journal of Industrial Engineering, 13(2), 76–83.
  • Hegazy, T. (1999). “Optimization of resource allocation and leveling using genetic algorithms.” Journal of Construction Engineering and Management, 125(3), 167–175.

Need to optimize your construction scheduling?

From resource leveling to project scheduling and workforce planning, mathematical modeling can transform your construction operations. Let’s discuss how Operations Research can work for you.

Disclaimer
Data shown is illustrative. Activity names, durations, and resource requirements are representative scenarios for educational purposes and do not reflect any specific construction project.
Back
ESC