## Haystack Syndrome

Monday, September 3rd, 2007

When I recently read Kevin Fox‘s Blue Light anecdote, it spurred me to go back and read the stack of old Goldratt books I’d inherited. I loved The Goal, but, frankly, I wasn’t impressed with his Theory of Constraints (the book, not the idea), or The Race (which is little more than a proto-PowerPoint), or The Haystack Syndrome.

On the other hand, The Haystack Syndrome does have one good demonstration of Goldratt’s key point. Suppose we have a company that produces two products, P and Q, from three different raw materials and one purchased part:

How much money can our hypothetical company make in a week? Eyeballing the numbers, we see that it can sell 100 of Product P and 50 of Product Q.

Each P brings in \$90 but costs \$45 in materials, so it contributes \$45 to covering overhead. Each Q contributes \$60.

If we sell 100 of P, that’s \$4,500. If we sell 50 of Q, that’s \$3,000. If we then subtract out our \$6,000 in overhead from salaries and the like, that leaves us \$1,500 in profit.

Simple, but wrong — because we’ve ignored all our internal constraints. We can’t actually produce 100 of P and 50 of Q per week.

OK, so we produce some fraction of that. Naturally, we want to maximize the number of Product Q we produce; those are more profitable — \$60 versus \$45, 60-percent margins versus 50-percent margins.

Simple, but wrong again — because we’re still ignoring our constraints.

Well, how expensive are P and Q in terms of resources used? If we add it all up, each P requires 60 minutes of effort from various resources. Each Q requires just 50 minutes.

So Q requires less labor and less allocated overhead, too?

Sure, but none of that matters.

Our bottleneck is at Resource B. If we produce 100 of P, that uses 1,500 minutes of B’s time. If we produce 50 of Q, that uses another 1,500 minutes of B’s time. That adds up to 3,000 minutes per week, but we only have 2,400 minutes.

Our goal is to maximize the dollar throughput of our system, which means we need to maximize the dollar throughput of our bottleneck. In this case, our bottleneck is Resource B, so we need to look at the dollar throughput of producing Product P versus Product Q at that bottleneck.

Product P contributes \$45 for 15 minutes at Resource B. Product Q contributes \$60 for 30 minutes at Resource B. That’s \$3/minute versus \$2/minute.

There’s no comparison. Produce as much Product P as you can!

This may all be clear when you walk through a nice diagram, but imagine how a large organization tackles a problem like this. How is the sales team rewarded? Based on sales (revenue) over some threshold? Or revenue minus product cost (which includes labor and allocated overhead)? Or contribution margins?