Principle Two: Get the Manufacturing Right
You can’t sell it if you can’t make it.
Most of the companies that I have worked with over the past 10 years have had a heavy dose of engineering content in their products. The companies made physical stuff (i.e., not software) and the reproducibilty and scalability of the production processes were of paramount concern. Time and again investors would ask, “Is your process scalable?” Time and again we would look at scalability and feel good about how to achieve it, and make strong pronouncements to our stakeholders. While scalability is VERY important, we missed the larger issue that was more important at the time.
The bigger issue is that the manufacturing process needs to come under control first. Unless you have a process that is in control (yields consistent output) at a small scale, it is not scalable. I’ve seen in many companies an inability to control the process–even at the prototype stage. In the semiconductor world, people are known for making wafers full of products (sometimes as many as 10,000 chips can be on a wafer). Instead of knowing that each die is good, instead they have to ferret out the good ones from the bad ones. And if you’ll permit me to get crass for a moment, engineers are known for finding the KGD (known good die) among thousands of bad ones, and showcasing the good ones as if to say that the product works. Clearly a KGD proves the concept is feasible (proof of concept) but it does not prove that you will have a successful business.
If there’s a way to affirmatively discern a good one from a bad one, one can probably live with this–for a while. When Intel started legend has it that their yield was on the order of 2%. But imagine if you were a machine shop, and your goal was to manufacture products by finishing them to a certain tolerance. Imagine if, prior to performing the finishing operation, you had no idea if it would work. Forget about scaling it, in this example the process is not under control.Having a random assortment of dice on a wafer representing every possible combination of unstable variables does not a product make.
Now extend this issue to the need to grow sales. Companies rightly want to be selling and shipping product. But unless you can make things under control in small batches first, you’re going to be in a world of hurt. If your process is stable and you understand it, then you have a chance to scale it successfully.
If manufacturing is part of your operation, I cannot stress enough the importance of focusing on this issue that directly hits your bottom line. Otherwise no one will buy your products (that’s obvious), but you also won’t have a chance to sell your business or license your process either.
Please comment if you have something to add for surely I’ve not done the subject justice.