Knowing the Unknown....Say What?

Companies live or die by their brand reputation and their ability to innovate and iterate on their existing product. In today’s competitive market every product must be developed with urgency and concern over what that product could do to the company reputation. That means there is only a limited amount of time to mitigate risk. Having a baseline for the engineer and the company to understand roughly how much risk is being introduced to the product is critical to measuring and ultimately managing the risk. Time and experience with a system like the one below allows a company to gauge how much risk they can mitigate within a typical development cycle.

When we think about risk we really need to think about where that risk originates. What makes something risky, and how do we mitigate that particular risk? Risk has many sources from which it originates. By thinking about the source of risk we begin to stratify the risk and break it down into components. Then for each risk source we can approach and mitigate the risk in an appropriate way.

One of the first hurdles one must overcome when approaching risk is the need to measure or quantify just how much risk is present. Naturally each source of risk might have a different method of quantifying the risk. What is important is that the method is appropriate to both the level of risk and the source. The ability to relate one source of risk to another source is also important because when one source of risk is mitigated it is often partially transferred or translated into another source or type of risk. Few people and organizations formally recognize this. To show risk is truly mitigated and not transferred it must be tracked and that means there must be a system designed for this purpose. In this article I will review one method of stratifying and quantifying some of that risk which we call New Content.

The Notion of New Content:

New Content is considered something that is new to a product or process that engineering doesn't have experience with. It is the notion that what is new is unknown, and since it is unknown there is some level of risk. This notion of new includes the application of the product, design, and manufacturing change. Remember we also need to quantify risk so we need to quantify just how much “new” we have. This is equivalent to saying we have to quantify what we don’t know, which appears to be a ridiculous statement because we don’t know what we don’t know. Wait! What?... Yea that’s right, it's confusing. However it is possible to get an idea of just how much we don't know about something, if we have a baseline of what we already know. We find this baseline by using what is known as apportionment of failures.

Apportionment of failures is the allocation of expected failures of a product into the product’s systems, sub systems and components. The best way to explain this is through an example.

The example we will use will be a very simple go-kart. I have chosen a free design from www.kartbuilding.net, as a base machine. I have chosen this for several reasons. It is simple with simple systems, the drawings and technical information is already mostly present, it is complex enough to demonstrate multi-function systems, and it has enough similarity to a car or tractor that we can project this risk mitigation system to more complex machines like off road buggies, cars, trucks and tractor trailers. In addition to these technical reasons, this simple design allows us to easily imagine something safer or with more desirable features. A new product built from this simple go-kart as a baseline would contain something outside of the demonstrated experience we are assuming, and that new something is the new content.

Disclaimer.

The failure data below is not representative of this design. It is completely imaginary and is only used to represent an engineering method of failure apportionment and calculating new content. I will be using this design as an example throughout several articles to show the identification, stratification and mitigation of product development risk.

Our imaginary company and imaginary product history.

Let’s say we own a company that has been building go-karts and selling them for the last 20 years. We only have one model of go-kart but we have made 10 different versions of it in the history of the company. We have divided all of the components in the go-kart into some very basic systems that we have defined as, Chassis, Steering, Drive train & Braking, Power and Cab. You can view the article on Boundary Diagrams for a better description of what drawings are included in each of these systems.

In order to convey apportionment quickly let’s assume the product version developed and released in year 8 was considered our most reliable product. We also sold 1000 go-karts that year so there is a large mature sample size. The failure data from those go-karts in the next 3 years had only random looking failures and there did not appear to be any systemic issues with the product. This mature stable period is when we would want to take a baseline of the systems expected failure rates. If there were 75 failures in this data set, and the count of failures in each of the systems was as follows; Chassis 18, Cab 1, Steering 10, Power train & Braking 21, Power 25, we would give each of these systems the appropriate percentage of expected failures as their apportionment. So the apportionment of failures would be Frame 18/75 = .24, Cab 1/75 = .013, Steering 10/75 = .13, Drive train & Braking 21/75 = .28 and Power 25/75 = .33. So summing these values together and accounting for rounding errors we get 1 as expected.

The idea here is that no system will be completely free of failures and we want apportionment of failures to represent the number of expected failures in each system when the system is considered reliable. We do not want to take a baseline of a product in an abnormally short period of time when no failures occur, nor do we want to take a baseline of failures when the product has unusually high failures due to systemic design or manufacturing problems. The baseline should reflect the failures that happen during the expected life of the product on what is thought or proven to be a good product. It is from this baseline we can measure in relative terms what is new to a design.

When we consider something to be new then we really need to consider at least the three categories of Design, Application and Manufacturing. (See the table below). We could change the design to the point where we only have a degree of experience with a technology, or we could change the design where we have experience with the technology but the complexity of the design has increased which will decrease the engineering certainty that we have it all figured out. We might have the same design but want to use it in a different application. In this case we really have to look at what requirements have changed for the design and evaluate what’s new in light of those requirements. We might decide to change the way the design is manufactured using new materials, processes, or suppliers. Based on the manufacturing changes we would have to look at what’s new in light of those changes. All of these changes to the baseline are assigned a percentage change from the baseline for each system and summed together for a total percent change to the system. This total percent change would then be multiplied by the apportionment value of that system. This would give you a percent new content for that particular system, subsystem or component.

Let’s take a look at what this calculation might look like for the go-kart apportionment example.

So in this case we have a change in the Chassis System, Cab System and Steering System. In this case all of these changes are considered Design changes and are not related to Application or Manufacturing elements of New Content.

Continuing on with our example we add up the fractions of change in each system to get the Change Factor and then multiply that change factor by the system apportionment to get the calculated system New Content. We then sum all of the System New Content values to get the total new content of the Go-Kart.

You may have noticed Cab system had a .15 change factor but 0 system new content. This is because the apportionment for that system when multiplied by the .15 is so low it is rounded to 0. Here the engineer has the option of rounding it up to .01, expanding the estimate to 3 or 4 decimal places or accepting it as 0 system new content. Typically engineers will take a conservative approach and make sure it is included in more refined risk identification activities like DFMEA. For this example we will round it up to .01 to ensure it is included in further analysis.

If we include changing the Cab system new content to .01 we have added a measure of about 7% New Content to the Go-Kart. This 7% is concentrated in three systems the Chassis, Cab System and Steering System. This approach may not be right for your particular product, and in the case of this simple go-kart might not be applicable, but having a systematic approach like this facilitates systematic analysis in very complicated products.

Your brand is your reputation, and your reputation can make or break your company. What gets measured gets managed, so if you want to manage your risk you must have a consistent system for measuring it. Measuring the risk allows you to limit it, and limiting the risk introduced in any given iteration of a product allows the product manager to bring the product to market with greater velocity and higher quality. Managing the risk is key to profitable products, happy customers and a great brand reputation.

Like it? Leave a comment & feel free to share it!