by R.T. "Chris" Christensen

One of the most useful maintenance inventory management tools we have at our fingertips is forecasting. Unfortunately, it’s also one of our most unreliable tools. And because of the unreliability, we need to be extra careful. In forecasting, we try to predict the future in order to minimize the risk of an unexpected, unpleasant event (i.e. a breakdown). 

Reasons for forecasting
Forecasting helps us determine when a machine will need repair. We forecast the outage so we can do one of two things: 

1) Be in a position to repair the machine or equipment just before it breaks and wrecks everything associated with it. On my car, I’d rather replace the fan belt than have the belt break while I’m on the road. Waiting until failure could cause the car to overheat, kill the battery or, worse yet, ruin the engine. At the same time, I don’t want to replace the belt before it’s necessary. It’s a waste of money (parts and labor) and time.

Transferring that thought to manufacturing, if we take the machine out of service unnecessarily, it can’t produce and the company can’t generate revenue. 

2) Have parts on hand in advance of a breakdown. That way, when the machine is down — notice I said "when," not "if" — we have parts available to get it back in service with minimal downtime. 

This way, we plan for outages and repairs. And by saying "when" rather than "if," you begin to understand the need. By understanding the need to determine when you must have the parts, you now have the basis for the forecast.

An imperfect science
Since we have the two reasons behind forecasting, let’s look at why it is our most unreliable tool. What we are trying to do is plan for an unforeseen future event. There is a main element of risk and inaccuracy here because we can’t accurately plan for the unknown.

If we could plan for such an event, our cars wouldn’t need spare tires. We carry that "inventory" spare because we can’t forecast exactly when or if we’ll have a flat tire. 

Manufacturing is the same way. We must understand the future needs of our process and try to forecast the unknown so we can have spare parts ready should an unforeseen event occur. The unpredictability of the event leads to the unpredictability of the forecast. 

Another reason forecasts are inaccurate is the data used is based on a past event. Our forecast tools tend to look at what happened and then extrapolate that data and apply it to a future need based on machine loads, which is based on the past history of plant productivity. The error is using the past to forecast the future.

A piece of software capable of forecasting cannot perfect this imperfect process. I had a long discussion recently with a corporate staff maintenance manager who was looking for THE ONE software that would forecast the rebuild requirement for 10 heat-treat furnaces that the corporation had in several locations. 

He wanted to shut down each furnace just before it failed. He said that a furnace rebricking took two weeks if the furnace is taken out of service before a failure. If he waited for the failure, the time out of service jumped to four to six weeks or longer. And, the plant would be down even longer because the heat-treat furnace was a bottleneck piece of equipment in each plant. 

The problem he faced was the number of hours of operation between overhauls varied by as much as six months. Most of his furnaces ran around two years between shutdown for rebricking. But if he took the furnace down at two years and it still had another six months of use left, he was wasting maintenance dollars. If he waited an additional six months, the risk of furnace failure was high and the cost even higher. He had a problem. 

What he wanted to do was maximize the length between repairs and overhauls and minimize the risk of equipment failure. He thought that forecasting software could solve the problem. All he needed to do was find that software.

I told him the software he was looking for didn’t exist. Why?

You’ll find the answers by clicking here and reading Part 2 of this forecasting tale.

R.T. "Chris" Christensen is the director of the University of Wisconsin School of Business' operations management program. If you have an inventory management question, contact Coach Christensen by phone at 608-441-7326 or e-mail cchristensen@execed.bus.wisc.edu.

Back to top

Back to MRO Coach archives