In the tech community, there’s a standing joke that, if you ask four data scientists what the meaning of artificial intelligence (AI) is, you’ll get five answers. Despite arguments over what AI actually means, it is hugely popular right now. According to Gartner, the global business value derived from AI grew 70%...in a year. But why is AI so popular? I would argue it’s twofold: one, because it’s becoming well known that companies that care about data do better. As companies make the transition to being data-driven, the efficiencies of AI are extremely alluring. Secondly, AI is doing some very interesting things. Advancements in not only methods but also physical technologies (CPU cores, RAM, streaming data, etc.) mean that larger datasets can be analyzed in more ways faster. With these advancements in AI, one wonders what this means for the rest of us. If you ask me, that is the most exciting part, especially if you’re lucky enough to enable your most important data assets: your citizen data scientists.
Before we can talk about why citizen data scientists are important, let’s first define what we mean by the term. For Gartner, citizen data scientists are “power users who can perform both simple and moderately sophisticated analytical tasks that would previously have required more expertise.” This is an apt definition, and it certainly gives credence to the advancements in technologies (example, AI) that make it possible. Making analytics easier so the everyday business users can take advantage of more difficult models—even if they don’t understand the nuts and bolts underneath—is precisely the technological foundation needed so citizen data scientists can exist. But notice some of the other key traits usually not held by the “actual” data scientists: contextualized vision, been around the block and has connections, unique perspective of the business area. To me, these are the real key features of citizen data scientists. They are what give them the power to look at a unique piece of data and know what to do about it. So, to review, a citizen data scientist
- Has the ability to apply advanced data techniques, ones that—thanks to advancements in technology (AI being one of them)—were previously only within the realm of more advanced statisticians and traditional data scientists
- Has key subject-matter expertise of the data they’re analyzing
But why are they needed? It’s because the idea that AI can tell you every “why” is just not true. The value of context and area expertise cannot be overstated when the other points of data are crunched. You see an outlier. Perhaps you’ve even added analytics on top of it to get things like deviations, regressions, multivariate analysis, etc. All of these data points are valuable, and AI can help you with all of them. But where is the why? Furthermore, what is the best route to take about it? This is where the idea of a citizen data scientist comes in, to provide that area expertise needed to get you not only the “Why” but the “What to do about it.”
Imagine this scenario: You’ve rolled out some new software. You knew it was going to be slightly buggy (aren’t all upgrades), but you’ve been getting a lot of calls from your Cincinnati office. Your level 1 team has done a great job of fielding these, but it’s starting to spill over and your survey numbers are beginning to suffer. If your KPIs drop too much further, you’ll start missing SLAs, and that’s when charge-backs happen. All your AI tools have pointed out the outliers, and perhaps even given you some steps to take to help remediate the uptick in negative NPS scores. But does this give you a complete picture of what’s going on? You know the data is valuable, so you go to your citizen data scientist for input. Luckily, she has some insight on the Cincinnati office. She recently wrote some KB articles for new hires, and she’s seen an uptick in readings from that office. She thought that was interesting, so she checked out the new-account tickets and found that a new department came online and their new manager is keen to make everything a high priority. This usually means skirting the escalation process and pushing tickets up almost immediately. You also know that this person doesn’t respond to emails in a timely manner but will take a phone call if it comes from the right person. So, the data gave you the “What” (here’s an outlier, based off statistics here’s what may be affecting it to aid in your investigations), but the citizen data scientist—with her connections and area expertise—was able to give you the context and a fuller “Why.”
I ran through this in my own life, something I spoke about in one of my first SupportWorld articles, Use Custom Metrics for Insights into Your ITSM Data. If someone was just looking at the data, they might have seen an outlier and spent untold amounts of resources trying to dig into the data points affecting it. I immediately saw it, had a good idea of what it was and, more importantly, had the expertise to know where to go and look to be sure. Fortunately, I was able to rectify the situation without needing to waste the energy of more formal impact analysis. This is another example of what citizen data scientists can do for you; not only can they help you get better whys, they can use their subject matter expertise to drive efficiencies in ways that are not directly apparent from the data.
So now we know that we know what citizen data scientists are, and why they’re needed. Let’s talk about how you empower them in your environments. Remember there are two foundations for citizen data scientists: technological enablement and subject-matter expertise. On the technological side, you would need things like:
- A technology that allows them to actually do the data analysis. This can be many things...but it probably isn’t the same thing the data scientists are already using (R, Python, etc.). Remember, the whole point is to not have to rely on traditional data science techniques, so you need technologies that enable that.
- Processes that empower the use of these technologies, in an agile framework. If you have the technical bits and bytes taken care of, but only allow exploration or modification to models in very scoped use-cases, you’re going to be limiting your resources. An agile approach is also helpful to allow for the nimbleness the data exploration may need. This is not just a process requirement, but also a technological one.
On the subject matter expertise side, here are some thoughts:
Divide and conquer. No one on the team will know everything. But you can have a KB SME, a customer service SME, or someone that knows the new hire process in and out. That will take work, so you’ll need a way to divvy up that work.
Encourage social networking. Knowing how to network is valuable (the majority of first jobs out of college are based on who you know). If you ever need an answer fast, you need to have an avenue that you can leverage to get it. Make sure that along with subject matter expertise, your citizen data scientists have the appropriate resources and contacts available to assist in their explorations as needed.
Promote process knowledge. Far too often in technology (especially ITSM, which routinely handles the same amount of calls), things can become rote. This is dangerous because in IT processes quickly become assumptions. Even if something is done the same way, perhaps the intent changed, or if it did change, why? Part of the above two steps means that you need to know your processes in and out, so let’s be explicit about it.
Discovering the “Why” is never easy, and the answer will change depending on one’s context and assumptions. But one thing is clear: your data can’t tell you everything. In fact, it probably can’t even tell you “that” (whatever “that” is for you). It can point in the right direction and give you hugely valuable insights. Ultimately, you need to take that last leap and ask the next questions yourself and/or know where to go from there. Enabling a group of citizen data scientists—both technically and culturally—will get you past just the data and help you get to the really important questions: the “Whys?” and, perhaps more importantly, “What to do now?”
Your data can’t tell you everything.
Adam Rauh has been working in IT since 2005. Currently in the business intelligence and analytics space at Tableau, he spent over a decade working in IT operations focusing on ITSM, leadership, and infrastructure support. He is passionate about data analytics, security, and process frameworks and methodologies. He has spoken at, contributed to, or authored articles for a number of conferences, seminars, and user-groups across the US on a variety of subjects related to IT, data analytics, and public policy. He currently lives in Georgia. Connect with Adam on LinkedIn.