In some industries it is completely alien to think about shipping a product before it is completely ready, but in others it is regrettably common. The semiconductor equipment industry lies somewhere in the middle. There are many industrial businesses in this middle category, hardware and software, and this essay is for you.
There are lots of reasons to want to ship a product before the beta is complete, the most common being competitive pressures and learning. These reasons are real. Let’s look at them more closely.
From a competitive standpoint it is not uncommon for a project to be late. The customer needs the product. You committed it. And, in the equipment business, if you don’t take up that floor space, your competitor will. As one sales guy once put it starkly to me, “Do you want them yelling at you that your product doesn’t work or yelling at the competitor that their product doesn’t work?” It feels like if you don’t ship the tool now you will lose the order and we already discussed the high importance of market share.
The main other reason for shipping early is “learning.” It is true that for really new products, there is only so much you can learn with the product still sitting in the factory. The real learning doesn’t really start until the product is in the customer’s facility. For products where one already has some market share and experience, the beta test, by which I mean controlled testing in two or three customer facilities, is for this learning, which for an established product, should be somewhat incremental. You can and should take some risk with a limited few customers. You should already know the use cases and should only be fine-tuning your understanding. But for a really new product, sometimes you have to just get out there to learn. History supports this. For instance, you may have noticed that many really new product families only get it right on the second try, e.g., the Apple Lisa before the Apple Macintosh. Indeed, one of the things that distinguish successful startups from unsuccessful startups is whether or not they have enough money left for a second try after their first idea does not take succeed. Pilots talk about altitude in this way—they want to be at least two mistakes up. People who study education talk about they key of learning through a series of non-fatal mistakes. T.J. Watson said that most people don’t make enough mistakes in research.
Despite all this, shipping a product early is generally a mistake. Let’s take a look in detail at these two reasons for shipping before it is complete.
When you ship a product early in response to competitive pressures, what happens is that you have to finish the product in the field in front of the customer. There are several issues here. The first and perhaps the most important is that it is very inefficient. Your engineering team has built up an infrastructure in your facility that is designed to make them maximally productive. Sometimes it doesn’t seem so, but if the environment is not productive you should change it (e.g., call fewer meetings). In any case, the environment is in your control. Send the product out to the field early and you will need to send scarce system engineers with it. In the field they won’t have the infrastructure to work efficiently. Analysis programs, interferometers, logic analyzers, and colleagues will all be far away and you will be thrashing around in front of the customer. The customer will be impressed by the number of bodies that show up and how many hours per day they work. Not how efficient they are at getting to root cause.
In some cases I hear that only in the field can one get a good supply of test product. In the semiconductor equipment industry that usually means advanced test wafers, but the point is more general. In my experience, this is just an excuse. I am sure it is true somewhere, but I have yet to see the case where a strong effort to get test wafers or reticles or whatever back to the factory for testing does not succeed enough to get through alpha test (by alpha test I mean comprehensive testing in the factory done NOT by the design engineers). Sometimes it needs to be escalated and sometimes the customer needs to hand-carry the wafers, but a priority must be put on getting the wafers here. If you are contemplating sending a tool to the field just to get access to wafers, please calculate how much this is going to cost per wafer. For that much money, are you sure that you cannot find another way?
In any case, when you send all your key engineers out to the field to finish designing and debugging the tool, progress slows. This means that while the first few tools might have shipped early, the ramp will definitely be delayed. Manufacturing and field service documentation will be late. (Some engineers think that their work product is the working prototype. This is not true. The proper product of engineering is the documentation that enables the rest of the company and supply chain to build and service the product.)
On top of that, and perhaps most perniciously, you will tie down your engineering team on this project and they won’t be available to work on the next one. Thus starts a vicious cycle where when one product is late and introduced in a sloppy manner it causes the next product to be even later. It is said that winning teams learn to win and losing teams learn to lose. Both are stable states. The most difficult management challenge is breaking a losing cycle. It is always traumatic and expensive because generally people cannot break the cycle and do the right thing until there is a catastrophe. Often you need new people.
So smart companies have processes that do not include ship tools before they are done because it causes divisions to fall into a death spiral where each product is later and sloppier than the one before. It may seem like shipping is an imperative at the time, I have been there, but it is very bad over the long term. A “law” that I heard from Steve Jacobsen of U. of Utah seems to apply here. It is called “The first law of holes.” When you are in one, stop digging! In an earlier essay I talked about what it takes to be a general manager and I put a lot of emphasis on experience. Situations like this are one of the reasons. When you are in a corner and it is “impossible” not to ship, an experienced general manager will have been there before and know that 80% of what people say is impossible is bullshit. Yes, it is horrible now. Yes, it seems impossible. But it is not. Doing the wrong thing will only make a worse mess for your successor to clean up. As one of my good friends says, the electrons and photons don’t care about your deadlines or bonus objectives.
OK, so once in a great while one does need to ship early. Then what? The key point to realize is that all of your processes for field support don’t work. This means Applications Engineers are not hired and trained, spares are not built and stocked, Field Service is not hired and trained, etc., etc., etc. If you ship early then you have to make sure all of this is dealt with by exception. Very inefficient and mistake prone, but it can, and must, be done.
What about shipping for “learning”? It depends. I remember having a discussion with a new general manager who wanted to ship a product early for learning. I told him, “I can tell you right now what you will learn.” “What?” “To never, never do that again.” Seriously, there are cases where it does make sense to ship early for learning. If you are entering a new market where you have done an exhaustive due diligence on the requirements but still don’t know the requirements well enough, it can be critical to get the product into the hands of the customer. But in this case you need to think about it as a design of experiments. What are the questions you need answered? What do you plan to learn? How are you going to learn it? A common problem is that “learning” is sometimes just an alibi for doing the wrong thing and shipping the product before it is done. Think of the design of experiments. Do you need a customer to learn that the product is unreliable? That it doesn’t have factory automation? That a DOS prompt is an inadequate user interface? That 45 minutes is a poor mean time between interrupts? That blurry images that jiggle on the screen are not competitive? There are legitimate questions that may need to be answered on the customer site, but don’t hide behind them. Write them down and decide whether the plan to get these answers is optimal. When someone says that they want to do something out of process for “learning,” make sure that you have inspected the written design of experiments plan and know that the state of the tool is in adequate shape to perform those experiments. It is easy for poor reliability to get in the way of any experiment. In certain cases, putting a tool and support at a customer can be exactly the right thing to do. Just think it through clearly.
The final topic I want to cover is reputation. The company’s reputation and your reputation. Both need to be trusted. It is good to have a policy never to put your customers at risk. I know, I put customers at risk on an important project and it cost my reputation terribly; it took a very long time to recover that trust. Let’s talk about what happens when you put a customer at risk. In most businesses you are in long term relationships with your customers. In the semiconductor equipment business, each of these customers is worth hundreds of millions of dollars, and some are worth billions. The second thing to remember is that companies are made up of people. Semiconductor equipment is generally expensive enough that people put their careers on the line when the make a tool selection. When you burn a customer by ruining their career, you create an enemy for life. And in my experience these people often do not leave the industry; they simply migrate from company to company, taking with them their poisonous enmity toward the organization that burned them.
I was once talking with a colleague about doing joint R&D with a customer. I am not talking about the usual JDP where we think JDP stands for Joint Development Program and the customer thinks it stands for Jumbo Discount Program. I am talking about getting very close to a customer and working with them on a revolutionary product. If you think about it from the customer’s viewpoint, someone there is taking a big personal risk by betting on you. My colleague looked at me and said that the JDP documentation clearly said that either party could back out at any time and that there were no binding commitments, just “intent.” Everything was above board. With a wolfish grin he added, “We are both consenting adults.” OK, let’s take that suggestive analogy and run with it. You can both be consenting adults, but if you get the customer pregnant and then want to abruptly break off the relationship, you should expect the customer to get highly emotional with you, and with your company. Every time one fails with a customer one creates a lot of long term damage. It takes years to rebuild that trust with the customer company and it may never be possible to rebuild the trust with the person who was your sponsor, who believed in you.