# It Started With a Question

It is the idea to eradicate MTBF from common use. The first question was

How do you explain what MTBF is and isn’t to someone that misunderstand MTBF?

We talked about a couple of ideas that may have helped. Of course, that first question lead to a few more.

How many others do not understand MTBF?

How many ways can MTBF be misunderstood?

What can we do to eliminate the confusion?

A lot more discussion and a lot more clarity around the problem. By asking one question and doing the failure analysis (one of our reliability engineering tools) we then started to form paths to a solution. If we worked together and uniformly asked a question, we could make progress toward our goal.

When someone mentions MTBF, simply ask them that they really want or mean.

If someone mentions 50,000 hour MTBF, ask them if the really would rather have 98% survive over 5 years. Make those using MTBF think a little, add clarity, add details, to what reliability performance they really would like to occur.

## Campaign buttons

To raise awareness, increase the number of questions about MTBF, we created the NoMTBF campaign button. Giving away thousands of buttons has helped to create thousands of conversations.

Now there are earrings, coffee mugs, hoodies, and more available with the NoMTBF logo. Display the icon and cause conversations. Make people ask questions.

## Movement website

The NoMTBF.com site originally was a site to collect and share the questions and the wide range of answers around MTBF. It has turned into a center for those with questions about MTBF. Actually the site started with someone asking, “Is there a website for NoMTBF?” and within the hour there was.

The site also contains articles on topics that regularly cause confusion and misunderstanding. If you have a question, check the site for answers. If you don’t find an answer, write an article and share it on the site. At least ask me the question and let’s find an answer.

## More Questions

Looking back at all the questions around MTBF, it’s been great. We’re making progress. And, we have many more questions.

As so many have already done, and you most likely have too, ask about what is meant when someone uses MTBF – and look for information to help you and your team make good decisions without the fog of misunderstanding.

MTBF is simply this…

If the hazard rate is constant (and it most commonly varies with time), then (and only then) does R(t) = exp (-t/MTBF) apply. Only under these conditions does it make sense to talk about failure rate (as a constant hazard rate) and MTBF.

You can count the number of failures in any interval, and divide by the number of unit hours during that interval, and that may be useful to do…BUT…since the hazard rate usually is not constant, you are not computing MTBF.

And, if the exponential reliability distribution (constant failure rate) does apply, you probably aren’t really interested in MTBF, but reliability, R(t). And it is useful to note that R(-1) = 0.37, i.e., when one MTBF has passed, the probability that a single unit selected at random will still be working is 37%, and not 50%.

Also, if a population has any other reliability distribution than the exponential, you cannot add hazard rates. This is a major reason that reliability predictions based on computing “failure rates” by taking the N/T (number of failures divided by time) and adding them together to get a product “failure rate” don’t predict very much, and don’t allow proper planning.

It is almost always a very good idea to select a theory that matches reality as closely as possible and to understand where theory and actual conditions don’t match. Ignoring this increases risk.

Hi Paul,

Thanks for the quick comment and thoughts.

As you have certainly experience, it’s often just asking someone what they mean by MTBF that leads to the discussion you outline above. We will know the NoMTBF campaign is successful when we do not have to ask to make sure someone knows what they mean when using MTBF.

Cheers,

Fred

Fred,

I’m as frustrated as you when I have to explain that MTBF is (only) the solution to a particular model, and one that–in spite of its popularity–doesn’t apply most of the time.

Wayne Nelson in his book “Life Data Analysis” makes the observation that about 15% of the time, you can use the exponential distribution without too much error. (If my memory is correct, that is.) If the lifetime distribution of a population is normal, then using the exponential or Weibull or any other distribution is wrong.

If the real life question is “how much effort and cost will be required to keep this system running,” then naturally I’m only interested in the solution to any model insofar as it is useful to answer that question. (Of course there are other questions that can be important.) Actually doing reliability engineering is complicated. It requires thought and it requires effort to answer the business questions managers ask.

I appreciate your effort here. I don’t mean to be pedantic on these points, but so often it seems that engineers, particularly those who should know better, are enamored of methods and software tools that support them without thinking much about the assumptions being used.