When Purchase Price Variance Turns Deadly

There has been a lot of press lately covering GM’s CEO testimony on Capitol Hill regarding the fallout from a faulty ignition switch on Chevy Cobalts due to an initial poor design of the purchased assembly from Delphi. Details are still emerging about the sequence of events, but let me abbreviate this story and take a little creative re-license of the Direct TV-style commercials:

  • When you have a ruthless focus on purchase costs, and don’t focus on total costs (including quality/warranty costs), design trade-offs get made
  • When design trade-offs get made, especially during a product launch, quality can suffer
  • When quality suffers, systems fail
  • When systems fail, people can die, including your customers
  • When your customers die (e.g., due to a sub-optimal design of a purchased assembly) you ask your supplier to fix it
  • When your supplier quotes you a price to fix it (like 90 cents per part and $400K in tooling), you realize that you will have unfavorable Purchase Price Variance (PPV) and that you won't get your bonus
  • When you are a slave to PPV and facing unfavorable PPV, then you delay implementation "until the piece cost can be eliminated or significantly reduced"
  • When you delay fixing the problem, more of your customers die (this is not just ‘external failure’ as the worst form of quality cost, but epic failure)
  • And when an investigator finds that you finally made the change, but buried it by not changing the part number, then your CEO gets called to testify before Congress…and you get fired (maybe)
  • DON’T have your CEO testify before Congress. Stop being a slave to PPV!

I know, it’s not clear exactly how everything went down at GM. Was it ‘just’ poor communication between legal and engineering?  Was there a Monte Carlo (pardon the pun) simulation to do the cost/benefit trade-off of making the engineering change vs. the potential ‘warranty costs?’ Did engineers really decide not to change the part number to help minimize detection (especially since revision control and lot control capabilities suck at so many firms)?

And even more questions:

Was there a PPV metric in place that would have been unfavorably impacted if the price went up? Was this a factor in the decision? Should Delphi have just eaten the cost of the design that was not up to performance standards? Does the current GM-Delphi relationship foster this and an environment geared towards eradicating such quality problems? Did unrealistically low-target costs drive this initial poor design quality?

It is hard to know the answers to these questions, especially since GM has gone into lockdown mode, but it does highlight how things can go terribly wrong when a narrow purchased cost focus is optimized at the expense of total costs, and in this case, the total costs were all too high. Our thoughts and prayers go out to the affected families in this situation, and we hope that other manufacturing firms learn from it.

The challenge of measuring total costs, including quality costs, is a big one, and the ability to define a balanced scorecard of supply is an even bigger one. But, the best supply chain organizations focus on advocating and implementing a ‘balanced scorecard of supply’ that aligns to corporate/customer objectives, rather than an old-school purchasing-led focus on just one of those metrics (i.e., purchase cost reduction). When metrics get siloed like the departments, then the total supply picture is lost, and alignment becomes vertical/stove-piped rather than horizontally aligned across the value chain – to often-disastrous results.

What are your thoughts on this situation? Do you see it all too often?

For more information on Supply Performance Management and Procurement Performance Management (and measurement via scorecards as a subset), please see our research coverage in the Related Articles below.

Share on Procurious

Discuss this:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.