In a common definition used by engineers reliability is a probability of success. It is the chance of an item operating as expected over some duration in a given environment. In this case, we have probability as part of the definition of reliability.
Reliability in common use definition includes trustworthiness, dependability and similar definitions. It’s more than how many times your friends help you move to a new apartment, it’s a feeling or sense we have concerning our ability to count on our friends for the help.
The ‘what if’
From my own experience of asking my friends to help me move, I experienced the twinge of uncertainty. The uncertainty that they would have neither time or interest. It was a statistical calculation without the numbers. I was both pretty sure and not entirely sure that if I asked they would provide their support.
In the world of gambling, whether playing poker or a lottery, there is a moment of ‘what if’. That is a tangible sensation that many have concerning the probability and uncertainty concerning the results.
In b, siness we call this risk.
Reliability and risk
In any product development program, there are a lot of unknowns.
- Will the functions actually meet customer expectations?
- Is the color appropriate for this market?
- How and where will this be used most often?
- Is it safe?
- Can we manufacture this design well enough?
The development process is one of uncovering and resolving uncertainty and reducing risk. At some point, the risk is tolerable or acceptable and we release the product to the market.
Some fail, some do very well.
The ability of a product to operate over some duration is its reliability. It is the risk of warranty claims or downtime that adds to the overall business risk. We attempt to quantify the ability of a design to survive and be reliable, and we use a lot of math to do so.
We may report test results such as ‘product x has at least a 95% reliability over 5 years with 90% confidence. This statement has reliability, duration, and uncertainty quantified.
We are expected 95 out of each 100 to work for 5 years without failure – kind of.
Reliability beyond statistics
In some cases, the above claim of 95% reliable is based on a single accelerated test or on engineering judgment. In either case, the uncertainty extends well beyond the statistical confidence figure.
How do we quantify our ability to know what we don’t know?
Beyond the sample size and test result support statistics, how certain are we that the testing and samples involved actually are relevant? Are we sure about:
- The impact of process variability
- The stability of materials in the supply chain
- The aging effects of humidity, dust, insects, fungus, cleaning agents, etc.
- The use loading and frequency (are our assumptions really the worst case?)
- Are we measuring the right parameters? Do they matter to customers?
- Are the measurements accurate enough to detect subtle and important defects?
Do we know what we need to know to account for the major risks involved with this product?
Reliability and Influence
At some point, someone will make a decision. The risk is acceptable or not.
How is that decision made? What information is included and considered?
I don’t really know, as I’ve never been in that situation. Been involved in many of those decisions as a reliability engineer though. And, have found that providing a clear understanding of what the statistics does and does not reveal is a complete picture of what we do and do not understand.
In short, we need to provide a full accounting of the risk.
It often comes down to our ability to influence.
Our experience, information, and analysis often rests on our ability to be trusted, believed, and understood. It doesn’t serve us to use fancy statistics alone without providing the context and assumptions.
We build trust by being certain when we can, open about what we don’t know, and candid about the remaining risks. We build influence by being honest.
Thus, it is my contention that reliability is so much more than simply statistics. Sure, it is part of what we do, yet is our character that provides the type of uncertainty that is considered.