dannyman.toldme.com

Tracking the Value of Automation

Cfengine is an tool used by Systems Administrators to automate the configuration and management of multiple systems. I have mixed feelings about cfengine because it strikes me as overly complex, and there are alternative tools available, but the basic idea of automating systems administration is sound.

Of course, it can take a lot of effort to automate processes, and it takes effort to bring in an automation framework like cfengine to facilitate automation. (Like I said, cfengine is complicated.) Is it worthwhile? More importantly, how do you convince management?

This bit of cfengine propaganda about a new cfengine feature got me thinking:

The value_kept, value_repaired, value_notkept settings fall under cfengine transaction logging and allow administrators to attach actual monetary (or other) values to promises kept, or issues repaired, or conversely measure the loss of non-compliance in dollar terms (choose your currency). This value is summed and recorded for each execution of Cfengine, and can be turned into graphs for your management reports.

“If you combine this with system performance data, and other reports from Cfengine, you begin to build up a pretty compelling case for IT services value. Hopefully this will give skilled system administrators the leverage they need to advance in the view of the more removed managerial levels, guarding the purse strings,” says Mark Burgess, author and company founder of Cfengine.

This is a clever idea, and one that strengthens the case for getting staff to record the time required to complete trouble tickets: if it takes two days of effort to automate a five-minute process then you “break even” on the investment after the process is run 320 times.

More importantly, these cfengine bits are focussed on “business value” . . . a given process may mean five minutes of effort to a SysAdmin, but it may mean a great deal to the user depending on that process. Putting a business value on a process can be tricky to do, but if you figure that a five minute procedure is blocking an engineer from getting work done, and it takes, on average, an hour for the request to be fulfilled, then the business value of automating the process pays off after 16 runs. (More or less, depending on the “value” of the users’ time versus the “value” of the usually-Senior SysAdmin doing the automation.)

Anyway, to the extent that you can put a value on any given process, you can record the value of the expensive process of automation, and help prioritize automation efforts. The greater “return on investment” can be shown for task automation, the greater your case to management for investing time and resources into automation. (Or, it is easier to identify and explain those things not worth automating. A difficult-to-automate one-hour process with low business value that runs once a month may very clearly not be worth the effort.)