Skill level: Intermediate
Predictive modeling uses current relevant knowledge to estimate the future operational parameters of a system. In the service industry the most obvious examples are for cost estimation and service performance of new products.
The process involves using data currently known about similar or predecessor products or processes to predict future performance. The information can be in the form of engineering estimates, failure rates from individual components, component costs, failure rates from similar current products, return rates, and more. A working knowledge of the process or product is essential so that input data can be screened for relevance.
- Increases service personnel effectiveness in debugging obscure problems
- Provides an organized depository for “embedded” specific knowledge
- Information can be quickly and efficiently retrieved through the use of queries
- Adaptable to many types of information
- Can be portable through the use of online interfaces
- Can be linked with other databases, increasing the value
How to Use
- Step 1. Determine the parameter to be modeled.
- Step 2. Brainstorm possible data sources and variables.
- Step 3. Screen variables for relevance.
- Step 4. Develop the predictive model.
- Step 5. Test the model for consistency.
- Step 6. Use the output and monitor against actual performance.
- Step 7. Modify the model as required.
Screening: To eliminate extraneous variables.
Generic Motors is developing a 2012 model of its popular four-door sedan. The model has had five previous generations, each of which has had some reliability problems with the automatic windows.
On the 2012 version, a new motor and gearbox combination has been introduced. Analysis has shown that the motor/gearbox was the primary failure component of the automatic window system.
The engineer in charge uses predictive modeling to assess future failure rates. The car’s onboard computer records duration, frequency, and actuations of each automatic window. The performance of the current motor and gearbox combination is modeled mathematically. Screening the variables shows that duration of actuation and torque are the most significant parameters. The company determines that it should introduce a new gearbox with automatic disengagement based on a maximum torque value. The new model also includes a higher ratio gearbox that would further reduce required motor torque.
After introduction, the new model is found to have 50 percent fewer failures in the first six months of production.
The company then conducts a financial analysis to determine if retrofitting older models would be cost beneficial. The cost of retrofit is $150 per car. The failure rate was high enough that over 50 percent of the cars would require replacement prior to warranty expiration. Based on these parameters, the company decides to initiate a voluntary recall on all prior models to upgrade the window motor/gearbox.