MTTR measurement

[viv121]

[color=blue]Guys,

Not again an ITIL question but a statistical one. I got a report on some MTTR values for the a team. For example- support group A resolves the twenty incidents assigned to it in 2 hours , 4 hours, 6 hours,8 hours and so on. The calculation given to me as the final value is the averge of the all the values of the incidents. am wondering should there be a better approach towards calculating the MTTR. May be the mid-spread calier which is the difference between the inter-quartile range should be taken up as a MTTR?
_________________
regards,

Vivek
"the only statistics you can trust are those you falsified yourself"
Winston Churchill

[Diarmid]

viv,

no answer without context.

There is a clear definition of mean in statistics, but you are talking as if you take it to be "average" i.e. one of mean, mode, median (or something even more esoteric). One expects MTTR to refer to the mean, but if not:

What is MTTR going to tell you?
What is MTTR going to be used for?
What does the distribution look like?
If you classify your incidents according to, say, the quartile they lie in, how well do the classes correspond to types of incident categorized in some other useful way (like apps server, network, ... or service a, service b,... or user type 1, user type 2,...)?
Is twenty a large enough number to make valid statistical statements in this case?

If MTTR is in the SLA, how is it defined there? ( if it is there but not defined, then invent several plausible definitions and use whichever one looks best at each review )

Have you got rules for excluding "exceptional" cases?

If you are doing this to measure staff;
Are you weighting the measures according to, say, their experience?
Are you allowing for teams receiving different profiles of incidents to deal with?

I hate statistics.
_________________
"Method goes far to prevent trouble in business: for it makes the task easy, hinders confusion, saves abundance of time, and instructs those that have business depending, both what to do and what to hope."
William Penn 1644-1718

[viv121]

Hi Diarmid,

I was clueless when you mentioned that I wanted to hear from John. Now I realize that what you meant. I never recalled that I put this question already but I don't really know why did I put this question in the problem management forum. Few of my nocturnal activities are inexplicable. It happens when we mix statistics with psychotropy. Nevertheless, I understood the concept of LDLaS completely. I should not go for tendencies, central or whatever. I shall be working around the mean. For example, I have a list of daily Service Minutes lost for a particular application for the last quarter. I have ninety data points. I'll get a mean to find the Mean SML. This would quite incomplete without being more elaborate. Mean or average means nothing without standard deviation. So, what I have thought of producing is the Mean value and all other values and their respective dispersion distance from the mean. Control chart in short with 3 standard deviations on either side. What I will try to take as a baseline is the respectives distances from the mean and the goals or targets will be reduce the dispersion so much so that most of the data points are around the mean. By control chart, I'll not give a tendency but complete study of all the values and not just one value.

What are your thoughts on this Diarmind?
_________________
regards,

Vivek
"the only statistics you can trust are those you falsified yourself"
Winston Churchill