Ask Slashdot: Good Metrics For a Small IT Team?

First time accepted submitter shibbyj writes "I'm a member of a small 3 person IT team for a medium sized business (approximately 300-350 employees) that has multiple locations internationally. I have been tasked with logging our performance using the statistics from our ticket management system. I've also been tasked with comparing these stats and determining if we are performing above or below what is considered optimal. I'm wondering what people opinions are on what good metrics should be in regards to mttr mtbf etc. I have had trouble finding information on this."

Comments

Metrics suck

[SJ2000]

Didn't we just have a story about how metrics suck?

Re:Metrics suck

[ScuzzMonkey]

Bad metrics suck, good metrics are useful data.

As folks have already mentioned, time to answer and time to resolve are both important, and I think you have to watch for re-opened as well to curb "how fast can I shove this under the bed?" resolution games.

My favorite is average tickets per user, though. Particularly on a small team, what you really want to gear your measurements toward is preventing incidents in the first place. It is helpful to know what your overall ticket volume looks like, then, and to aim to decrease it over time in the same way you might try to decrease time to answer and time to resolve. That's important, because as the previous article suggests, if you will get what you measure... and your overall goal should not simply be to answer tickets faster and resolve them more quickly, but to not have as many in the first place*. Every issue represents a waste of somebody's time and therefore corporate resources that could be put to more productive uses. Steadily decreasing mtta and mttr are nothing to cheer about if your ticket count is increasing.

But you can keep it simple. You can drown yourself in metrics and lose sight of why you're tracking them and what you really want to accomplish. You may not really need any more than these few; better to start small and add what you need when you need it. I know there's always tension over getting a system in place that can capture what you need for historical purposes when you realize you need to know something new down the road, but resist the urge to over-collect. Half the time you won't need it all and are just wasting time getting it in the system.

* There is a caveat to this; in some organizations, I actually DO want to see an increasing ticket/user count, at least for a time. This is something I shoot for when relations between IT and users have broken down badly enough that users have stopped reporting problems to IT because they feel it's useless, and their issues are never resolved anyway. In those cases, a rising ticket count can represent an increasing trust level, which is good. You generally won't fix issues you don't know about.

Re:Metrics suck

[CAIMLAS]

Metrics is just a management word for "bad statistics".

With a distribution of 3, it's not really possible to have statistics of meaningful nature. You've got shared responsibilities and bounce things off of each other. One person may open and close more tickets, have a shorter duration for tickets, etc. - but he's only doing the actual "work". The others may be giving him all the input necessary to complete the task(s).

Ideally, your ticketing system will reflect, very vaguely, who's doing work and who is not, but even then it's not going to be well representative of what's actually going on.

People do different types of work, of different levels of difficulty. For instance, I may do one ticket on Monday, three on Tuesday, and one for Wednesday through Friday. Why? Aside from the fact that I'm bad about actually doing tickets for my work (my god, I'd not have the time for work, and then there'd be more things that aren't getting done, making us -all- look bad), there's the reality that my tickets aren't terribly easy, often requiring hours of log perusal and research to try to fix problems. Meanwhile, the guy who knocks out 40 tickets a week - malware disinfection, workstation reinstalls, etc. - has fairly wrote work of a repetitive nature, comparably. Also, he's following instructions or asking for advice on a regular basis, even if I'm not his boss.

You said so yourself: you're a member of a 3-person IT team. The only use 'metrics' have here aside from what should be plainly obvious in a group of 3 (who's fucking up, who's not getting back to people, who's not doing work) is to keep track of what amounts to customer requests and problems. X workstation needs to be reinstalled, Y server has a crashing whatever, and so on. If you're working on and sharing a ticket queue, you are all mutually responsible for all of the tickets: if something isn't getting done, it's everyone's fault (or nobody's fault). You may consider presenting your metrics in the light of this reality (like statistics, metrics can lie, too).

Here's a few...

[HerculesMO]

Time to answer call, time to resolve ticket, abandoned tickets (unresolved).

If you google a few of those it will bring some more, but that's a simple start.

Re:Here's a few...

[suutar]

Number of calls back after initial call (measuring, in theory, how often the initial call resolved the issue) Number and duration of system outages (if you're doing sysadmin stuff as well as support stuff)

optimal?

[Anonymous Coward]

talk about flow, about bottle necks. Visualize workflow. Look at Henrik Kniberg's paper on kanban as applied to IT Ops. My guess is that your ticketing systems will provide low value data on volumes on resolution time - gear up - visualize the pipeline. check http://www.infoq.com/minibooks/priming-kanban-jesper-boeg - turn the conversation around to "business value" - don't get wrapped in the ropes of volumetrics /peace.

Re:Hahaha

[Anonymous Coward]

Yes, this.

Management wants to eliminate someone and wants to do so in an "objective" way to hide the fact that they're firing someone while probably giving the CEO a fat Christmas bonus. You're tasked with figuring out which of the three of you gets fired and how you can cloak this in enough "objectivity" that no one can object to it. Your best bet is to make this shit up. Figure out who the weakest link besides yourself is, or who you like the least, and generate a system of metrics that's biased towards eliminating that person. Use lots of acronyms and jargon. Also, make sure no one at work reads Slashdot.

totally irrelevant CAPTCHA: forgive

Re:call / ticket time is bad metrics

[DarthBart]

I got written up once because my ticket stats were radically different than the other people on my team. 15% lower "total time on tickets" but 20% more tickets closed. I was apparently fudging numbers and closing unresolved tickets.

Fortunately, a trip to HR with a ream of printouts from closed tickets proved otherwise.

Still left the company a few months later.

Re:Done in one

[nahdude812]

You cynical man. For all you know, upper management have a budget flush with cash and have singled out someone in the hard working but unacknowledged IT department for a raise and a promotion.

Don't forget the free pony.