Learning and Product Development
By now, we should all understand the need for speed when it comes to product development, but not just any speed, it is the need for speed of learning. This is one of the benefits of an agile approach, but this benefit is not restricted to a development methodology, unless of course, the development methodology precludes or runs contrary to fast learning cycles.
You might think that this learning is only associated with complicated or complex systems, perhaps categories of projects that we have no familiarity. It is true that learning is especially required when we are working on things we have little or no experience. However, in product development, we can ill afford to believe we know things that are simply not true or follow unvetted assumptions. For example, we can model and simulate, but neither of these are true parts. If we do this methodically, we can certainly reduce the need for some prototype parts, but that does not mean exclude them. This is especially true when we have many parts, interfaces and subsystems coming together.
Wire harness and Models
Consider for example, a collection of physical parts such as a wire harness for a vehicle. This may seem like a simple product that consists of a variety of connectors, and wire lengths that are used to connect sensors, solenoids, and the myriad of electronic control units on the vehicle together. It is in fact more of a subsystem for the vehicle. Let us consider the lead time for prototype wire harnesses are 8 – 12 weeks.
The mechanical attributes of this product can be modeled in CAD to learn things about fitment. Assuming the entire vehicle components are also modeled, then fitment issues can largely be relegated to the CAD exploration. Right? There is nothing wrong with making that assumption. Right? We likely all know the saying about AssUMe. I have personally witnessed cases where key elements of the system’s components were not modeled or modeled but all models were not imported so an entire model of the physical system is available. There are a number of ways learning from models can be tainted:
- Some of the system is modeled
- Some parts of the individual components are modeled
- The models are not up to date
- The configuration management of updated models is errant (can’t tell which model is the valid model)
Learning and Physical Elements
Even when we do have every key physical element of the system modeled, and all those models are accurate and up to date and imported to see how every part of the system fits together, we still have a need for prototype parts. There is much to learn through having the physical parts in hand for exploration of the design and interfaces that can not be explored entirely digitally. For example, models may have unknown (missing) elements that will only rear their ugly head when the parts come together. Additionally, the physical parts will be required to perform testing. For modeling that includes environmental simulation, these parts and tests will be the feedback to the environmental components of the model. This feedback provides information on the quality of the simulation. That is, what we learn in the testing should uncover the range of stimuli to which the product will be subjected (such as vibration profile) that may not be congruent with what we suppose the stimuli to which the product will be subjected. What happens if the time for this learning is 8 – 12 weeks? What does this do for your ability to learn and speed the speed of that learning?
It is possible to move toward fewer prototype parts, but not zero. The prototype parts are used to understand the real product in ways that the models alone cannot. The prototype parts are also used to learn more about the impact of the environment upon the product. If our simulations are sophisticated enough to allow for the environmental stimulus to which the product will be subjected, our prototype parts can help us learn about the variety and ranges of these stimulus and if we have made some mistakes in the model. A long gap between the design and the availability of physical part does not facilitate learning about the product. We are not able to quickly adapt the product based upon what is learned, since nothing is effectively learned for months on end, including the implications of the environment on the product. In this case, the learning about the product is stalled for weeks to months on end.