Lost in statistics

An open discussion on issues related directly or primarily to the service or help desk.
Post Reply
User avatar
viv121
ITIL Expert
ITIL Expert
Posts: 117
Joined: Fri Dec 14, 2007 7:00 pm

Fri Mar 19, 2010 4:35 am

Hi All,

Came across an interesting query. Like all organizations, my organization is also working on improving on the total age of the incidents. For the baselining purpose, we took the information for the last quarter of 2009 which were around 180,000 helpdesk incidents. To find the age of an Incident, we subtracted the open time and close time to get the age of one incident. We summed up all the ages and divided the number by 180,000 to get a value and we termed it as Baseline MTTR value.

My question is that what we call as MTTR and expand as Mean time to repair, is it actually the average time to repair or should be have the median( 50th percentile of 180,000 datapoints) or midspread/inter-quartile value to be closest to the actual MTTR. In the number of tickets, we will have a number of outliars ( resolved in 1 minute, resolved in 30 days etc) and average might actually not work. From the purist point of view, MTTR is expanded as Mean Time to repair. How would that be expanded from a practical point of view.

We tried some lies and damn lies to the business. Statistics appears to be the last resort.

Not an ITIL question surely but may be a best practice question. Appreciate any help.


regards,

Vivek
"the only statistics you can trust are those you falsified yourself"
Winston Churchill
User avatar
UKVIKING
ITIL Expert
ITIL Expert
Posts: 3639
Joined: Fri Sep 15, 2006 8:00 pm
Location: London, UK

Fri Mar 19, 2010 5:01 am

Viv

incident 19123 is different from incident 22045

you can do some LDLaS on the following which would treat all incidents the same - all cats are grey at night

key points -
service impacted
raised
initial action by SD
escalate to resolver group
communication with service customers
service restored

This would depend upon whether your tool differentiates in your ticket
between the service impact / restore DTG and the ticket open, last touched, closed DTG

If you merely use the open/close DTG to calc age of the ticket and then gather them all and do standard deviation, average, mean, median, high, low overall.

Take the following into consideration

what tickets are created by SD Staff, via autogenerated email from customer users, automated by NMS / Monitoring / Alerts / etc ; etc

You should divide these into the distinct groups that the LDLaS would reflect more accurately

Thn you take the statistical data that you have gathered, arrange in a periodic manner - monthly, weekly. daily so that you can see if the STDDEV, average, mean, median, high, low for the entire range and each of the periods

The above can be done for tickets that have a classification system of some type associated with it

And then rinse lather repeat for management

But.... what does this really do besides fullfilling some MBA Graduate in a senior mgmt desire to see numbers... {Note: The same can be achieved using random number generation]

The real issue and LDLaS usage should be used this way

Volume by day or hour - where the ticket volume grows - this helps with planning staff numbers

etc
John Hardesty
ITSM Manager's Certificate (Red Badge)

Change Management is POWER & CONTROL. /....evil laughter
User avatar
Diarmid
ITIL Expert
ITIL Expert
Posts: 1894
Joined: Mon Mar 03, 2008 7:00 pm
Location: Helensburgh

Fri Mar 19, 2010 5:29 am

Viv,

didn't like my response yesterday then?

John, of course has it. You have to turn your question on its head. There is no such thing as the "actual...average". Mean, median and mode are all equally valid statistical averages, but they tell you different things.

So, as John said, you identify what you actually want to know and what you want to do with what you want to know and then you apply operations of data recording, gathering and analysing to meet that.

Just to get a bit pedantic (I lost another game of chess last night - that's the only two games I have played for the club this season!) if the purist definition is not the same as the practical definition, then one of them is incorrect.
"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
User avatar
viv121
ITIL Expert
ITIL Expert
Posts: 117
Joined: Fri Dec 14, 2007 7:00 pm

Fri Mar 19, 2010 4:09 pm

Diarmid,

How are you managing loss after loss in chess? Who's (check)mating you mate?

John, liked the concept of LDaS really. I'll get into this report generator utility to get some control charts in place. The control chart will help me understand why 19123 is different from incident 22045.

Cheers

Viv
regards,

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