How to Generate More Accurate Spare Parts Demand Forecasts
Written By: Ed Wodarski

The short answer: Replace proxies with actual data.

The reason why generating accurate parts demand forecasts is so difficult is because parts planners often use estimates of the following factors:

  • equipment utilization rates
  • the number of installed assets
  • the locations of assets across the installed base

When your think your machines are in operation for 10 hours a day, but are actually in operation for 12, that obviously causes forecasting errors. The problem worsens when you assume you have four assets in Service Region A, but actually have 11.

Fine, but where are you going to get reliable, accurate data on utilization rates and equipment locations? From the assets themselves.

Using data from equipment to generate accurate forecasts

Now, before you close your tab, let me clarify something: Not all of the equipment in your installed base needs to be connected. It probably isn’t – research from analyst Bill Pollock found almost half of parts planners said less than 25% of their equipment is connected.

So, what if 25% of your equipment is connected? Let’s call that installed base “Segment A.” In this case, you’d pull utilization and location data from those assets to generate a parts demand forecast for Segment A.

Another key point I want to make: You don’t need real-time data. If you run demand forecasts on a monthly basis, all you need to do is pull equipment utilization data generated over the past 30 days. Real-time data is useful when you’re running predictive maintenance programs, but it doesn’t do parts planners much service – the reason being that parts planners don’t run forecasts 24/7.

The fact that you don’t need real-time data could actually help you navigate whatever data governance policies your organization has in place. My colleagues, Steven Caldwell and Vinod Arekar, discussed this issue in a guide about connected service parts management. Organizations can store data from connected assets in databases that have no direct connection to the assets themselves, thereby minimizing the chances of hackers exploiting whatever systems planners are using to compromise equipment.

The first steps to using predictive analytics

Parts planners don’t need IoT data to use predictive analytics, but it’s preferable, based on the points I made earlier. Arekar and I are co-presenting on the applications of predictive analytics in service parts management on Aug 7th, during a one-hour webinar. You can register for the event below:


Start Exploring!
Tags: CAD Industrial Internet of Things Service Lifecycle Management (SLM)
About the Author Ed Wodarski

Ed Wodarski is a Service Parts Planning (SPM) expert for Servigistics with a special focus on the commercial aviation ecosystem. Ed has over 36 years of experience in SPM software design, deployment and sales support. Starting his career at Xerox in 1981 as a part of the design team for the first bespoke global parts planning system, Ed is widely acknowledged as an industry founder. He later then designed the first commercial offering for LPA/Xelus which has since been incorporated into the Servigistics platform. Ed has also been a Senior Executive at Accenture consulting globally on parts planning best practices. At PTC, Ed has worked closely with a number of leading aviation enterprises including Boeing, Aviall, JetBlue, and Southwest.