The Dirge of the MTBF Bias
We use our biases every day to make choices.
We select the beige sweater because we have a color bias concerning our sweaters.
Many of our biases help us quickly make decisions. We rely on biases to move through the day. Many of our biases are under the surface, unconsciously guiding our daily decisions. Mostly, biases are good or at least inconsequential.
The problem is the bias that shields us from achieving our goals.
MTBF Bias Blindness
When making decisions concerning reliability, if we have a bias toward using MTBF, we inadvertently thwart our ability to make informed decisions.
If we assume a constant failure rate or the exponential distribution applies, we then fail to look for contrary evidence. We are, in some cases, unable to recognize the changing failure rate behavior when it actually occurs.
It has to be well described by MTBF. That is our bias speaking.
We relegate the decision to use MTBF to an un-thinking act.
I once heard the rationale for the use of MTBF as ‘we want to compare our current performance to past performance’. Since they had been using MTBF for years, they did not have any other measure. They did not have reliability (probability of success over a duration).
Since the team had assumed MTBF was good enough, they summarized the time to failure data and lost the necessary information to check if the bias was helping or hurting them.
The NoMTBF Bias
I have a strong bias. I do not like using MTBF, for anything, ever.
It has limited my ability to communicate clearly at times. Maybe softening my stance would help the overall goal of making informed decisions concerning reliability.
Be aware of your bias concerning reliability measures.
Be aware of the bias of those around you, in your organization and industry.
Is the use of MTBF serving or hindering your ability to understand and improve the reliability performance of your product or system? If not, do something about it.
Help those around you check their MTBF bias at the door and recognize the harm the MTBF shackles inflict.
Expose the MTBF bias when you find it.
What is the first metric you think of concerning the performance of your system over time?
How do you and your team or industry talk about reliability? A couple of good questions that may reveal a latent (and misguided) affection or bias for MTBF.