Managing uncertainty in new product development is difficult in a rapidly changing world. Firms need to adopt strategies for transient advantage in turbulent markets as recently observed by Rita McGrath.  Product developers can’t wait though for all the answers and absolute certainty that risks missing market opportunities. Firms need to capture as much value as possible from new products yet product development cycles can be long and potentially beyond the timeframe required for a short wave of transient advantage.

Product developers need to give their firms maximum flexibility to exploit transient advantages so to mitigate and exploit uncertainty product developers need to design-in a higher degree of reliability (in uncertain conditions), robustness, versatility, flexibility, evolvability, and interoperability in their product platform and product lines. To be successful product developers need to understand uncertainty and clarify design strategies available to them during the ‘fuzzy front end’ of design.

What are the varieties of uncertainty and what design strategies can be used to manage a diversity of uncertainties?

Uncertainty Continuum

The simple model of known-knowns, known-unknowns, and unknown-unknowns is a useful starting point to understand the varieties of uncertainty but it is not detailed enough for product development. Schlensinger, Kiefer, & Brown’s uncertainty continuum provides a deeper look at varieties of uncertainty mapping along a scale of predictability from the known to the unknown. Their uncertainty continuum maps from the known along a scale of increasing unpredictability as follows:

• Completely Predictable – You can say with certainty what the outcome of a given situation will be such as with physical laws.
• Predictable Through Probability – The outcome can be defined to a particular confidence level using statistics but extremes may exceed bounds.
• Predictable Through Other Analytic Methods – The outcome might be predicted through chaos theory, computer modelling, which is less precise.
• Predictable Through Pattern Recognition, Experience, and The Like – The outcome might also be predicted based on limited prior experience or from patterns. The emerging world of big data.
• Not Predictable At All But You Can Say What Can’t Happen – The outcome is not predictable but certain cases can be ruled out.
• Completely Unpredictable – The outcome is completely unpredictable.

A linear scale is useful to model the range of predictability to classify variables according to how well the value can be predicted for design but it does not provide insight into the severity of events that is important for risk mitigation / opportunity exploitation in product development.

Uncertainty Framework

Another excellent framework for broadly understanding uncertainty of complex systems was proposed by McManus and Hastings (based largely from experience in the US space program) and one of the best I have seen at capturing a holistic view for managing uncertainty in product development. This framework links categories of uncertainties through risks and mitigations/exploitations to system outcomes to be more useful to engineers.

The framework provides a top-down model to structure uncertainty and risk taxonomies to illustrate cause and effect through the relationship – <uncertainty> causes <risk/opportunity> handled by <mitigation/exploitation> resulting in <outcome>. See the paper for excellent cases to understand the framework. I particularly like this framework because it does not just frame effects of uncertainty as a downside risk but upside opportunity that firms can exploit for transient advantage. The framework is also general in nature allowing it to be applied/tailored to any application.

Varieties of uncertainty used by McManus and Hastings are:

• Lack of Knowledge –  Facts that are not known, or are known only imprecisely, that are needed to complete the system architecture in a rational way. Knowledge in this case may just need to be collected (because it exists somewhere already) or created.
• Lack of Definition – Things about the system in question that have not been decided or specified.
• Statistically Characterized (random) variables/Phenomena – Things that cannot always be known precisely, but which can be statistically characterized, or at least bounded.
• Known Unknowns – Things that it is known are not known. Things are at best bounded, and may have entirely unknown values.
• Unknown Unknowns – Things that are gotchas that we cannot contemplate occurring with our current understanding.

An improvement combines the uncertainty continuum defined by Schlensinger, Kiefer, & Brown with the front end of McManus and Hastings’ uncertainty framework to more clearly understand how uncertainty maps to risks.

Design Strategies For Uncertainty

Both models provide guidance for design strategy to give firms flexibility for transient advantage.  The uncertainty continuum suggests at the extreme of completely predictable proven design heuristics are appropriate. At the higher extreme of unpredictability a short horizon learning experimental approach such as creaction is appropriate.

The McManus and Hastings’ uncertainty framework is more powerful for designers by linking uncertainty to levers of design.  McManus and Hastings provides a useful list of risk mitigation and exploitation strategies for new product developers to consider. These design strategies help to fill in the middle zone of the uncertainty continuum. McManus and Hastings identify nine strategies:

1. Margins – Designing systems to be more capable, to withstand worse environments, and to last longer than ‘necessary’.
2. Redundancy – Including multiple copies of subsystems (or multiple copies of entire systems) to assure at least one works.
3. Design Choices – Choosing design strategies, technologies, and/or subsystems that are not vulnerable to a known risk.
4. Verification and Testing – Testing after production to drive out known variation, bound known unknowns, and surface unknown unknowns.
5. Generality – Using multiple-function (sub)systems and interfaces, rather than specialized ones.
6. Upgradeability – (sub)systems that can be modified to improve or change function.
7. Modularity, Open Architecture, and Standard Interfaces – Functions grouped into modules and connected by standard interfaces in such a way that they can ‘plug and play’.
8. Trade Space Exploration – Analyzing or simulating many possible solutions under many possible conditions.
9. Portfolios and Real Options – Carrying various design options forward and trimming options in a rational way as more information becomes available and/or market conditions change.

Although several of these strategies most engineers would naturally use but the list may provide some suggested approaches that are not often considered. These nine strategies help to realize a new product design with system outcomes for reliability, robustness, versatility, flexibility, evolvability, and interoperability.

Most of these design strategies add cost to the new product development project and the product itself but the benefit is flexibility. Firms need to weigh the cost benefit of product flexibility in an uncertain world to support a strategy of transient advantage.