David Socha follows up his recent blog on the missed opportunities left in the Data Science lab with a discussion on how to operationalise those insights and deliver new value to the business.
Last month, I talked rather disparagingly about those little Data Science projects that many of you already have running. Some are in back rooms where only young people in ripped jeans go. Others are in fully-fledged new Digital Divisions of your business…populated almost entirely by young people in ripped jeans. Sorry, I’m doing it again, huh?
I don’t really have a problem with these small-scale Data Science projects at all. What I do have a problem with is the fact that so many of us in the industry appear to think they’re an end in themselves. What’s the real value of these projects, if none of them ever get further than that little back room? The answer is simple. There is no value. Data Science is a method for discovering new business insights. But business insights of any kind only have value if they are acted upon. If they are applied. So why aren’t they?
Little ideas and big problems
Data Science projects should start small. Those little experiments should most often begin by using just enough resources to quickly but credibly prove a concept. And nothing more.
Your Data Scientists might use a laptop, some free Open Source software and a small and easily accessible data set to prove in a week or two that – conceptually – they can use Machine Learning to predict when Circuit Breakers will fail. They might also show that therefore, they could – conceptually – provide information on when CBs should be maintained based on condition, not just time since last maintenance or number of trips.
Next, they might find evidence to suggest we could reduce the number of maintenance visits we currently do, while also reducing maloperations and disruptive failures. Wins all round, right? Improved safety; reduced maintenance costs; fewer outages…the list is endless. But how do we actually prove their theories and then actually make those things happen?
This is where things start to get hard. Because to make something happen in the real world and not just on a laptop requires effort. And will. And budget. It usually means technology and process change. And it often means Change Management too. Here we see the stumbling blocks stopping so many great little ideas from ever becoming great big new ways of doing business-as-usual.
Getting over the technology
We’re going to need to move off that laptop for a start. We’re going to need a solution that can scale to support and analyse all the data, not just the sample data sets from the PoC. And can we really afford to do all this work to operationalise these new Circuit Breaker insights as a standalone, silo’d solution? What about the next insight we’d like to implement in the real world?
It’s time to think big. It’s time to implement a single, scalable, flexible analytics platform that can support the rollout of all new analytical insights into real, live business operation. One that provides a single, secure, shared source of data. One that can support all kinds of users and technologies, from those Data Scientists experimenting with their ever-changing toolset to business users with their existing dashboards and BI tools and even to customers and business partners, given selective and secure access to insights relevant to them.
But all that sounds risky. And expensive, right?
Wrong. Thinking big doesn’t need to mean investing big. A genuinely scalable solution is one that can start small, only scaling up as required. Thinking big really means thinking strategically. That means choosing a platform that meets your needs today, but that is also flexible enough to service all your current and future analytics needs and able to scale from something small and affordable today to something that one day, potentially sits at the heart of your business, bringing value to every Department.
Investing in such a platform versus more and more Departmentally-funded point solutions will need vision and commitment. But once you do, one key enabler is in place to facilitate that transition from back-room insight to real, live better business, again and again.
Getting over the process
On to our second perspective. A couple of years ago, I wrote a blog about how implementing new insights from analytics has to become business-as-usual (or as I wrote at the time “a way of life”) in itself. In other words, we must operationalise not just insights, but also the process of migrating from initial insight all the way to BAU. Then, we can do it again and again with minimum effort and maximum comfort that what goes live will go live quickly. And will work.
Hmmmm. That sounds expensive too. Are we talking about a new Department here?
Let’s look at it a different way. If we take a leap of faith and believe that we might find more than just one nugget of value in our data ever, then which way of operationalising the value will be more effective: a series of one-off projects, starting from a blank sheet each time, or a set of trained people, understood processes and ratified Governance checks that are focused on delivering the same way each time? Your call.
Getting over the change
Finally, even with all of the above in place, we must still recognise that change is scary. And that properly implemented Change Management is still required. Going back to the Circuit Breakers: which Managers are going to have their budgets cut due to the savings we’ll achieve? Are they OK with that? Is the Chief Engineer signed off on changing our Asset Management practices? Who is responsible for stating our formal position the first time a Circuit Breaker fails after we’ve migrated to the new regime? Are they comfortable with that? Which tradesmen are going to feel that computers are stealing their jobs? How are they going to react to that?
None of these issues have anything to do with Data Science whatsoever. But they’re just as critical to delivering the value of Data Science as any algorithm; any software; any smart kid in ripped jeans. Neglect Change Management at your peril.
So…it’s not actually about Data Science at all
In summary: getting the value from Data Science is all in the operationalisation. You may have the best Data Scientists in the world, but apart from a little kudos on the conference circuit, they’re about as valuable to you as a chocolate teapot. Unless…you take their most valuable insights and actually apply them in your business. To do so, you’re going to need scalable, flexible technologies. You’re going to need repeatable processes. And you’re going to need to recognise and manage the fact that change is still scary, even right here in the 21st Century. It’s all achievable. Other industries have been doing it for years. Isn’t it time you did it too?