How ONE Bad Business Process Doomed GM

A friend of mine was on a flight a few years back with a Business Process Management (BPM) expert, who had done a lot of consulting for various car companies... the fellow told an interesting tale about how one single bad business metric made him swear off GM cars forever... and it just might be a major reason for their downfall.

The business process seemed innocuous enough at first glance... GM wanted to control costs on their auto parts. So, the process stated to test every auto part, and look for the most expensive parts that were the most reliable. Next, ask suppliers for a cheaper version of that part. Sure, the cheaper version of that part might fail more often... but so what? These are the parts that were severely over-engineered. You don't need space shuttle quality parts in a minivan... so what's the harm?

Notice the problem? If not, don't worry... neither did the millionaires who ran GM.

Now, consider how a rival car company dealt with the same problem. They would also run tests on every car part. They would also keep metrics on which car part failed the least, and which failed the most. However, they did very different things with the data. The rival company took the parts that failed the most often, and either demanded higher quality versions, or searched for a new vendor. Sometimes these changes would increase the price of the part, sometimes it would decrease.

Now do you see the problem? Each business process had an overall side-effect to the quality of the cars produced. As rival companies made their cars more and more reliable, GM was making theirs less and less reliable! Instead of focusing cost cutting on the overall finished product, they decided to tie cost cutting directly to making lower quality cars!

After realizing this, and being completely unable to get anybody at GM to change their metrics, the gentleman decided to swear off GM cars forever...

Peter Senge warned about similar problems 30 years ago in his famous book The Fifth Discipline. Side effects, negative feedback loops, and simple delays in cause-and-effect can wreak havoc on any business process you put together. No matter what your metric, there is always a case where a "good" result is a very bad thing... The key is to try to predict how that could be possible... otherwise, by doing your job better and better, you just might be dooming your company to mediocrity.

Naturally, its probably unfair to blame this all on one single business process. There are also the armies of people asleep at the switch who should have done something to correct it. Unfortunately, the folks who design the business processes are usually unable or unwilling to accept this harsh reality... and if they are politically powerful, bad processes remain. This is why we need "nice" tools that help gather information about these negative consequences that are outside of the model, and make it clear that its time to change...

That's one of the stated goals of Enterprise 2.0 tools... but even they can't help you unless you first try to build up trust and camaraderie in your company. Only then is it easy for people to accept the harsh reality.

Local Maxima

The GM story illustrates a common optimization problem: Local Maxima. By trying to optimize the price and quality of individual car components in isolation, you can trap yourself at a local maximum. If the competition takes a holistic approach to quality -- analyzing each part and how they work together -- your competitor can arrive at a global maximum, finding the very best mix of parts and price and quality.

Big companies fall victim to this problem all the time. By compartmentalizing roles and responsibilities, it's easy for an individual purchaser or line manager to optimize their narrow specialty while ignoring bigger goals.

not only that...

but the individual line manager will receive HEAPS of praise from their manager, for doing such a great job at their role... and since the Peter Principle is alive and well, odds are low that the manager, or the manager's manager has the vision to understand the problem.

"In a Hierarchy Every Employee Tends to Rise to His Level of Incompetence."

sad, but true...

it is true that GM somehow

it is true that GM somehow cut its problem into pieces and somehow wanted to solve them independently as Dean put it the holistic view were missing). But in my view it is not a surprise. In a big corporation the individuals work just part of a process (e.g. to build, service something) and their work have to be measured/rewarded as well. \everybody wants to be the best in her own pond. If you measure on the wrong way then you get the wrong result because "what is measured will prevail" even if the intentions were completely different.
In my view the major issue is that GM used BMP to tackle an optimization problem where it should have used optimization technology! Actually in Dean's replay it pops up too but he mixes up BPM and optimization.

In BPM you know (at least you believe you know) exactly what and how you want to do certain things. In the GM case the condition was "test every auto part, and look for the most expensive parts that were the most reliable" then the action was "ask suppliers for a cheaper version of that part" in a BPM stuff you will always find this condition-action sort of structure (the rule-based system). Formerly this type of systems were called "expert systems" now they have fancy names as business process, or business rules management because of marketing but the core is the same

In optimization problems/models you describe your problem in term of constraints and you have some goal by which you evaluate a solution to compare and naturally you want to optimize (min or max) to get the best solution (by definition a solution has (should) always satisfy all the constraint) with respect to this goal. You apply some search algorithm to find solutions and hopefully the best one (with respect to the goal). When you set up your model you determine what you can get at best (with respect to the goal) then the applied algorithm will determine if you can get it at all (guarantee or no guarantee of global optimum). E.g. if you have a linearly constrained model with linear goal you are set to find the global optimum of that problem (well, it depends on the size and complexity when you find it but it is another matter). If you change your goal then this old optimum might not be optimum! it is not local optimum, it is just not optimum from another goal's viewpoint.
In the GM case you shoul dbuild a cost function for the entire car and have constraints saying that this suppliers' part goes well with this suppliers' part, how a given part is enabling less parts to deal with etc.

In BPM (if you apply it for some business optimization) you make practically the search (if this-that then do this-that). Then (except rare cases) you are set to get a solution which is not optimum at all and you don't even know it . Who's put together the "if this-that then do this-that"? the experts, right. Why? because most experts are (1) operational people responding to some conditions daily (2) they have no idea that something else exists (3) most people who decides what IT systems should be built understand better what is going on with BPM appli then what an optimization appli can or can't do.

Re: it is true that GM somehow

In my mind, they failed at differentiating the tactical business metrics, from the strategic business metrics. From a tactical viewpoint, cost cutting is always good. But the strategic quality metrics were completely ignored... for whatever reason, nobody saw or nobody cared that the quality metrics were going down, down, down...

Based on what I have been

Based on what I have been reading this is really due to the unions and the incredible wages retirees make!

Re: Based on what I have been

Yes and no... Unions are a convenient punching bag, and an excuse for why GM management failed miserably. Unions can slow down adoption of new technology, but that's not always a bad thing for the business...

GM spent billions on difficult-to-program robots trying to make workers irrelevant. By doing so, they angered the unions, and shot themselves in the foot. If they spent that money training their workers instead, they would have been a lot more agile, and able to bring new cars to market faster.

I worked as an ISO-9001

I worked as an ISO-9001 certification consultant analyst for the Big Three for nearly six years; after every "successful" project I watched each one lose a little more market share and bleed a little bit more from some self-inflicted wound or another. I spent four years looking for a way out of the industry and am incredibly grateful I no longer have to work directly for the OEMs or their suppliers (though as a Michigan resident I am still their prisoner to some extent).

Something I rarely see mentioned anywhere in the business or industry media, including the blogosphere, is the damage that GM, Ford, and Chrysler did to themselves by outsourcing their entire middle management tier.

Not only did this effectively destroy "corporate memory," it politicized the day-to-day operations of the business and invited fraud and malfeasance from the very white collar contractors who had brought in to run things. I am not implying that the contractors themselves stole from the company; I am saying from experience that they were under tremendous pressure from their home offices to bill as much time as possible and to expand each contract house's footprint in Dearborn, Detroit, and Auburn Hills.

Too many projects "required' overtime from consultants. Too many projects "required" additional staff from consulting houses. And when the exorbitant consulting and OT fees were paid, there was no one internally to pass knowledge of TGR/TGW to. As soon as something went wrong, the process would repeat itself.

I'm speaking in generalities, of course. There are excellent contracting firms that provide invaluable services for the money they're paid. It's not the fault of even the least ethical contracting outfits that the Big Three thought it would be a good idea not to have anyone representing the home team involved in, or even available to take notes on, multimillion dollar projects. At the end of the day, however, a contractor's job is to bill "as much as the market (client) will bear" and to perform his or her job in such a way that the client is motivated to reengage the contract house, if not that specific contractor, at the next opportunity.

The Big Three's contracting decisions may not have been intended as an invitation to gouge them, but it certainly was taken that way by far too many consultancies from what I saw over six years; they should have known better.

GM Business Process

I'm going to defend GM for attempting to identify over-engineered parts. Obviously, you have to assume that this isn't the sum total of their quality initiatives. But if it isn't - and I don't see now it could be - then turning up resources being wasted in order to put funnel them into other areas that need beefing up ... that's quite sensible. It's one way to improve quality without increasing costs, which is a tricky thing to do. Even if the information is used simply to cut costs, which is a less appealing use of the data, this is still much better than cutting the costs of any other component of the car. They’re not cutting costs because some components of the car are over-engineered, they’re attempting to cut costs with a minimum of impact to the quality of the car.

Evolution does this sort of thing to extremes. People are engineered to last around 80 years, and at that point everything starts failing more or less at once. This is because an animal with one component - its eyes, say - that's engineered to last twice as long as the rest of its body would be at a competitive disadvantage. The organism would be investing unneeded resources in these eyes, and over time would lose out to another animal that invests these resources in something with real benefit.

Too bad GM couldn’t evolve more quickly, but bringing the quality of components in line with each other doesn't sound to me like where the problem was.

Interesting but this is about decisions not processes

Liked the post and the story but it seems to me that it is really about decisions and not processes as I said in my post on the topic

James Taylor

good point... but...

I did say the process itself wasn't alone in the blame... as you say, people need to make smart decisions based on the output of the process.

However, if GM -- like many big companies -- use process as a substitute for competence, then those kinds of decisions will simply never get made.

So... does that mean they need better processes, or smarter people? Probably both... but if folks are obsessed with processes, I'd argue fixing the process would be easier.

Recent comments