Report on Top Incidents Reported using Short Description

[lberthold]

Does anyone have any suggestions for determining what the top issues reported are within ServiceNow? Our incident form doesn't have a ton of meta fields to choose from to help with this. We do use Category and a custom field called Product. But we are looking more to compare the Short Descriptions of incidents to see what times of issues stand out - possibly by comparing and counting similar words. I just have no idea how to go about doing that.

Any thoughts would be appreciated!

Thanks!

Lori

[Pradeep Sharma]

Hi Lori,

My thoughts here are that, you should report it based on the priority field.

Thanks,

Pradeep Sharma

[peterwestergaar]

It sounds to me as if you're looking to categorize your Incidents and then group similar sorts of Incidents together for   reporting.

One of the best ways to do that, one that is well in line with an ITIL framework, is to define your Services such that your Incidents are clearly linked with interuptions of a provided Service, or with failures of specific CIs.     Then you can report on the Services that have large volumes of Incidents associated with them, or types of CIs that have large volumes of Incidents associated with them.

[lberthold]

Thanks for the responses!

Regarding Priority, I'm not sure how to use that - just because we have high priority incidents, doesn't mean they are the most reported. Is there some other way that field would be used to indicate top issues? I would expect most incidents to fall under the Low Priority value.

With regards to defining our services, our CMDB is set up slightly differently. Our ServiceNow instance is externally facing...most of the CIs that have issues are a software version or product of our software. So the CI field is actually a filtered set of our product/version list. But that alone wouldn't tell us what the issue is. For example, are the issues reported related to crashes, software latency, login issues, etc. Those issues can be across product.

That's why I was wondering about the Short Description and if there is anyway to correlate similar words and their occurrence. Even if that means exporting data and using some third party tool to help with analysis.

Thanks!

Lori

[peterwestergaar]

Lori,

It sounds like you're well aware of the potential for miscategorization by using free-form text in that way, so I won't further belabour that point, other than to say it would probably be more reliable and consistent if the people resolving these incidents tagged, categorized, or in some way sorted the tickets into buckets.

But if attempting to pull signal from the noise of short description is the focus, I'd probably lean either towards building a set of business rules that look for key-words and assign a category based on those (for example, you could have a rule that assigns a hidden category field to "crash" if you see key words like "crash", "BSOD", "blue screen", "dies", etc., and another rule that assigns that hidden category field to "login issues" if you see key words like "password", "login", "userid", "let me in", etc).  

Another method might be to look into something like "Bayesian" sorting, which would allow you to define a number of buckets, and train the system by having human beings carefully put several hundred incidents in each bucket, after which the system gets steadily better at guessing which bucket a new short description belongs in.     This is one of the techniques some spam detection software has used.  

The problem with both methods of course is that they're just attempting to sort the Incidents into pre-defined Categories.   Other brainstorming ideas: allowing free-form "tagging" for the Incidents, like what ServiceNow uses on this community;   or even stripping out "stop words" from the short descriptions and then running all remaining text through a "tag cloud" generator to pick out commonly-entered individual words without trying to collect synonyms together.

Sounds like an interesting challenge, hopefully you find a good solution!