Texas State University
McCoy College of Business.Research Insights
← Back to articles
Supply Chain Article3 min read

When Data Are Thin, a Simple Capacity Planning Rule Holds Up

Firms rarely know exactly how much of every resource their products need. A study finds that the common shortcut of planning a few resources carefully and scaling the rest by ratios performs well, and shows where better data pay off most.

Illustration of a planner at a desk filling in a capacity table, with a miniature factory floor laid out above it: two production machines drawn in full color at the center, linked by arrows to fainter sketches of tool benches, parts bins, packing stations, and work crews.

To decide how much capacity to build, a firm needs to know how much of each resource its products consume. For machine time on a main production line, that information is often good. For many other resources, such as indirect labor, setups, or support services, it is incomplete or wrong. Bills of materials have gaps. Records contain errors.

Planners still have to commit to capacity, so they rely on rules of thumb. A study by Vic Anand, assistant professor of accounting, with co-authors Ramji Balakrishnan of the University of Iowa and Srinagesh Gavirneni of Cornell University, tests how well those rules work. It was published in Production and Operations Management. Anand conducted the research while on the faculty of the University of Illinois at Urbana-Champaign.

The study

The researchers compared five heuristics for setting resource capacity when information is limited. Two come from practice. Two are modified versions of approaches seen in practice. One is motivated by theory.

Each heuristic was scored the same way. The researchers calculated the expected cost of having too much or too little capacity under the heuristic, then compared it with the cost a firm would face if it had full information. The closer the two, the better the heuristic.

What the researchers found

A simple approach that many firms already use performed well. Under it, the firm plans rigorously for a few “driver” resources for which it has high-quality information. It then sets capacity for the remaining “non-driver” resources using ratios, for example 0.25 indirect labor hours for each machine hour. The numerical analysis shows this heuristic is robust and efficient.

Three further results show how to get more from it.

First, using more than one driver resource to plan for the same non-driver resource produces significant gains.

Second, not all data improvements are equal. Reducing measurement error in how driver resources are consumed is worth more than reducing error anywhere else. When that measurement error is high, gathering information that reduces other sources of error can lower overall performance.

Third, the choice of drivers has a clear answer. A greedy rule, which selects the most expensive resources as drivers, is optimal.

What it means for managers

Firms that plan with drivers and ratios do not need to treat the practice as a compromise. The study supports it.

The results also point to where to spend effort. Measure the consumption of driver resources accurately before investing in other data. Tie each non-driver resource to more than one driver where possible. Choose the costliest resources as drivers.

For accountants and operations managers who design cost and planning systems, the message is that more data do not always produce a better plan. The quality of the few numbers the plan rests on matters most.

This summary is based on the paper’s abstract. The full article reports the model, the five heuristics, and detailed results.

What it means for managers

  • The common shortcut works. Planning carefully for a few well-measured driver resources and using ratios for the rest is robust and efficient.
  • Two drivers beat one. Using more than one driver resource to plan the same non-driver resource delivers significant gains.
  • Fix driver measurement first. Reducing error in how driver resources are measured matters more than other data improvements, and collecting other data while that error is high can make plans worse.
  • Choosing drivers is simple. Picking the most expensive resources as drivers is optimal.

Anand, V. V., Balakrishnan, R., & Gavirneni, S. (2023). Capacity planning with limited information. Production and Operations Management, 32(9), 2740–2757. 10.1111/poms.14004