Service parts for this products can be predicted based upon installed base and usage. The forecasting is a subset of causal forecasting as can be seen in the graphic below.
What This Article Covers
- What is the purpose of gaining mean time between failure (MTBF) documentation?
- Why is all service parts demand dependent demand?
- How prevalent is causal forecasting with MTBF?
What Type of Demand?
All service part demand is dependent demand. That is the demand for service parts is based upon purchases that have already been made. Service parts can be forecasted using simple demand history, as with finished goods, or they can take advantage of the installed based and usage of the equipment that is in the field. For some things just the population information is available (population information is obviously much easier to attain, generally only large and expensive equipment like airplanes or construction and heavy industrial equipment has the usage tracked.)
How Does the Mean Time Between Failure Fit In?
MTBF is one particular modality of causal forecasting. Most causal forecasting simply uses one or many independent variables to predict the future dependent variable. However, causal forecasting with MTBF in service parts uses a developed failure rate for the in the field item.
This is a simple example, but it captures how MTBF forecasting generally works.
Combining MTBF and Other Methods
Often the different forecasting categories are thought of as only being used independently. That is if you one for a product or group of products you cannot also use another. MCA Solutions actually allows you to use both a time series and an MTBF forecast. They call this the composite forecast and this forecast has the ability to give different weights to each forecast type. For instance you could weigh the MTBF at 70% and the time series forecast at 30% or any other set of percentages that you wanted.
Prevalence of MTBF Data and the Usage of This Type of Forecasting
Many companies talk about forecasting using MTBF data, but few of them are interested in doing the work to maintain the data. What is unfortunate here is that the data is not that difficult to maintain. There is not one level of granularity that companies have to drive to in order to use causal methods. They can get benefits from using just a basic high level value of their installed base. This should be available for even consumer items by taking previous sales data and applying degeneration percentages (for items that fall out of service) in order to develop a very basic installed base number. Once this number is attained it can be used for MTBF forecasting.
Some basic mathematical estimation can get companies close to the real values. Once these basic installed base numbers are generated, it opens a new opportunity to begin managing the service forecasting process differently.
It would be nice to report that causal methods on only underused in service organizations. However, this is not the case. It also extends to most supply chain organizations. See this post for details.
Obviously, there can be no causal forecasting without causals. This type of data should be elementary to maintain, but it is often not maintained. John Snow, in his Uptime Blog, which is associated with Engima, provides some good insight as to why below. It seems that the natural inclination of many service departments is to focus on quickly getting equipment back in service, with less concern for proper equipment maintenance and calibration. During a break-fix event (unscheduled maintenance) this is a rational response: the equipment is down, revenue generation has stopped, so get the machines working again. However, even during scheduled service events mechanics can become overly focused on speed. This is an example of reacting to the urgent rather than resolving the important. The problem is that service departments are often measured more on productivity than on quality.
See the full article here.