How to get useful and valuable insights from your IT cost data
Insights. These are the gold nuggets everyone wants from their IT cost data. Yet no one believes their data is good enough to provide them. You strive, sometimes with Herculean efforts, to get everything right. But what’s interesting is that no matter what state it’s in, data always provides insights.
Of course, it would be easier if insights simply jumped off the pages of reports. The unfortunate reality is that understanding requires intuitive reasoning. Insights evolve from things that simply look like outliers, and they often require a bit of research. (Interestingly, Merriam Webster defines insight as an act or result of “apprehending” the inner nature of things.)
So, how do you know you have useful and valuable insights? There are three characteristics: they must be consumable, defensible, and actionable.
One of the most skeptical classes of people on the planet are CIOs looking at IT spend reports. Their combined knowledge of the IT landscape and financial spend mean they need more than just top-level numbers. They need details too.
To make insights about IT data consumable, you’ve got to present your findings in a format that allows reviewers to understand what it is and why it’s relevant. My point is this: you’ve got to know your audience.
For example, let’s say you pull together the following data set on servers:
Looking at this, you might say “Sheesh, we have a lot of Windows servers!” But there are likely only a couple of people who are interested in this data: the person who put it together and the person who owns the server infrastructure—who are probably the same person. If you were to share this with executive management, they might say: “So what? Is that bad or good? What do I do with it?”
Now, consider what happens when you bring in a little extra research to make your data more consumable for this audience.
Example of consumable insights
When you can add information you know will be relevant to a bigger audience, you offer larger value. In this example, you can now say, “Gee, not only do we have a lot of Windows servers, but they are costing me double what my Linux servers do. How can we make better use of Linux as we go forward, to optimize and reduce our infrastructure spend?”
With this spend data, you’ve opened up your audience to include not only the folks who own servers but the CIO trying to figure out how to reduce his run budget, those responsible for I&O, the folks responsible for developing and deploying applications, and more.
Think about what happens when you put a report in front of somebody for the first time (like the CIO looking at the IT spend report). You get the finger point—either at the report or at you. “Where does that number come from? That can’t be right. I know it’s X, not Y.”
You need to defend the numbers. This doesn’t mean you need to have a high degree of detailed accuracy—you don’t have to work your data out to the half penny. What it does mean is this: you have a full understanding of where the insight comes from. You can answer questions like “What does the number mean? Where does the data come from? Has the data source been vetted?”
Before you present your data, you’ve got to make sure you’ve answered your own questions: Does it pass the reasonable person test? Would you stand up in front of your boss or a room full of people you care about and feel confident about it? Would you stake your reputation on it?
To get to something defensible, you may need to do a little more research and refine your results.
Example of defensible insights
Using our server example again, this table shows cost by class—the standard “t-shirt” sizing of servers (S, M, L, XL). It also depicts cost per hour and AWS cost of instance hours.
It’s very tempting to say: “Wow, look at how much cheaper AWS is! I can save $10.56 an hour for every XL server I move to AWS!” Now, think about your boss and that room full of people. Would you really say this? Probably not.
Of course, there are different costs that make up the total cost of a server. So, if you enhance this table with cost breakdowns like labor, depreciation, and maintenance, you provide the detail necessary to see the cost impact of migrating servers to AWS.
As you look at labor, you may realize there aren’t many savings opportunities there. You’re probably not going to start firing people as you migrate servers. These people probably do other things, regardless of whether those servers exist or not, so it’s safe to say the labor costs aren’t going anywhere (unless you’ve got some external contract work in there or etc.)
Also, you’re probably still depreciating some of those servers. Even if you toss a server out tomorrow, you still have that expense to manage.
It’s when you evaluate maintenance contracts paid by server that you land on a defensible insight. Now this is money you don’t have to spend any more if you move an XL server to AWS. There are real and achievable savings for every server you move to AWS. With more research, you could take this even further, looking at lower risk servers that support dev and QA environment, things for extra capacity only, etc.
You can see how a little more research and refinement makes the insight you’ve gleaned about servers more useful by being defensible. This is something you might be willing to stand up and present to an audience of decision-makers, because you know you could pull it off if your audience says go do it.
The final critical piece of the data insights puzzle is actionability. So far, you’ve asked yourself, “Is my data consumable? Yes. Is it defensible? Yes.” Excellent news. But there’s one final question you’ve got to ask: how does this insight help the organization optimize, make a better decision, or achieve an objective? If you can’t point to some improvement as a result of your work, you’re just wasting your time.
Example of actionable insights
Here’s a practical example: Let’s say you’re looking at a budget variance report for six months. The report shows how much you’ve spent according to plan. You notice one month sticks out. “Wow. We really overspent in April.” Interesting piece of data, right? However, there is no way to go back to April to do anything about it.
So, how do you make this actionable? One way is to review the overall spending trend over the last six months. Let’s say you discover a steady and consistent increase in variance. “Whoa, we’re getting ahead of ourselves in spending. If we don’t reverse this trend, we’ll spend way outside our intended budget this year.”
This is actionable. Your next step is to determine if you need to re-forecast and ask for more money or if there are some adjustments you can make to get back on track.
Your data just needs to be “good enough”
When I first moved to London, I relied on the London Tube and Rail Map to get around. Note: this map is not to scale, nor is a perfect map of London. Yet it was definitely good enough to get me where I needed to go. The alternate, “more perfect” version might be a physical or topographical map, which would indeed give me a lot more information about rivers, green spaces, and elevations. But having all of that data would make it MORE complicated to figure out how to get around, not less.
The moral of course is don’t let perfect be the enemy of good—or good enough. Thinking everything has to be all filled in will only work against you. You will never get to perfection. Let consumable, defensible, and actionable dictate your data needs.
Bill Balnave is Regional Vice President of Solutions Consulting at Apptio. This is the first installation in a 3-part series on data analysis.
Read Bill's next installation in this series, 3 best places to start your search for cost optimization insights.