Case Study – Integrated Circuit tradeoff of On Time Delivery, Inventory Cost

Case Study – Integrated Circuit tradeoff of On Time Delivery, Inventory Cost

The planning manager pulled me into his office. “Eric, I’d like your help – as the old Pogo comic strip said, ‘We have met the enemy, and he is us’”.

Dave McLaughlin, the planning manager for the IC manufacturing area, went on to describe how every month he would report on our “On Time Delivery” performance. Every month, it..sucked.

Moreover, his team’s primary tool to improve On Time Delivery (OTD) was prioritizing certain lots of material going through the manufacturing line – the prioritized lots were called “Red Rushes”. Even higher priority lots were named “Blue Blazes”. We joked that the extremely high priority lots could be named “Indigo Out-They-Go”.

As Dave saw it, part of the problem was that, by rushing certain lots, other lots needed to wait – and that was causing those lots to be late, impacting OTD for their customers. From my studies in Industrial Engineering – I was on a long term path to eventually perhaps get my doctorate – I added some insight: that rushing certain lots in manufacturing was increasing cycle time variability, and Kingman’s Equation showed that increasing cycle time variability also increased the average cycle time.

So, I agreed with Dave’s sardonic observation – in trying to fix the problem we were making it worse.

I realized that there were two key aspects of the OTD issue:

1. We were looking at OTD in the rear view mirror – talking about what the OTD was last month, when there was nothing we could do about it…it would be better to deal with OTD proactively, with a predictive model so we could improve OTD in the next month.

1. We were reacting to possible OTD misses by rushing some lots and impacting others…and those possible OTD misses might be due to non-optimal planning – starting too little material and/or starting them too late to meet the commitment, and rushing to make up for our mistaken planning.

The solution was a predictive model for On Time Delivery – that it was the product of two probabilities: the probability that enough good parts could be manufactured, allowing for yield variability during manufacturing, multiplied by the probability that the parts would be completed by the committed date, allowing for cycle time variability.

Quick analysis of yield and cycle time information from Motorola’s manufacturing areas, and from the literature, led to statistical yield and cycle time models that fit historical data remarkably well. Developing an On Time Delivery model from this information allowed Dave’s team to start the right quantity of material at the right time so that prioritization of manufacturing lots virtually disappeared – gone were the Red Rushes and Blue Blazes! On Time Delivery reached high levels and the planning people’s jobs became less stressful – here is a graph they used for choosing a lead time to quote to customers before planning starts:

Shortly thereafter, George Turner, the Vice President of the Logic Division of Motorola Semiconductor Product Sector, asked me to help him. He told me, in confidence at the time, that the Logic Division was in trouble. “When Tommy George, the VP, congratulated me on my promotion to VP of Logic Division, he also told me, ‘George, we have 22 divisions in Motorola SPS. Do you know where the Logic Division ranks financially? It is 22nd! George, I don’t need 22 divisions – I would be perfectly happy with 21 divisions. You have 2 years to turn the Logic Division around”.

Part of the solution for the Logic Division’s financial woes was to get a handle on ridiculous inventory levels. I used the OTD model I’d developed to recommend appropriate planning strategies and inventory levels for each of the 7 Logic families of products to achieve a goal of 95% On Time Delivery at the lowest inventory holding cost. Below is the figure for those families – the proposed and actual results for inventories when the Logic Division quickly achieved 95% On Time Delivery for all 7 families of products:

This helped Logic Division immensely, contributing to its rise from the 22nd most profitable – to the 2nd most profitable division of Motorola SPS over the course of 2 years.

This also had an impact on people – George Turner was promoted from Division to Group Vice President, and as for me…I was asked to present on these results at an internal Motorola meeting in Austin, Texas. One of the external invitees was Dr. John Fowler, a professor of Industrial Engineering at Arizona State University. After my talk, Dr Fowler approached me, and suggested that this would make a great PhD dissertation….and that is how he became my faculty advisor and this led to my doctoral dissertation, “Modeling the On Time Delivery and Inventory for Semiconductor Supply Chains”.