Last month, my colleague John K. Thompson outlined five best practices for getting started with analytics. If you read the article, you probably noticed right away that none of those best practices deal directly with choosing the right analytics solution. There’s good reason for that, too. The right culture, the right questions, and the right expectations are the foundation of any successful analytics initiative. Fail to put those building blocks in place, and even the best analytics technology likely won’t get you where you want to go.
Eventually though, the rubber meets the road, and you will need to invest in a technology solution to power your advanced analytics initiatives. And just as the wrong culture can undermine the right technology, the wrong technology can likewise undermine the right culture. So while building the right culture is a critical component in the success of any analytics initiative, so too is investing in the right technology. Let’s take a look then at five key things to look for in a modern analytics solution.
1. Ease of Use
Let’s face it: Every technology vendor claims their solutions are easy to use. And in their own way, each of those claims might be true, at least to an extent. In other words, ease of use means different things in different contexts. In the context of modern analytics, ease of use is about empowering citizen data scientists – the knowledge workers, within the line-of-business driving the bulk of today’s analytics initiatives. Chances are, your company isn’t exactly teeming with highly paid data scientists. Your primary concern then when evaluating a given analytics solution for ease of use should be the extent to which citizen data scientists can actually use it to help make business decisions.
Let’s drill into a specific example. While they have a penchant for working with data and certainly understand the value of applying analytics, citizen data scientists are not likely to understand the mechanics of specific analytical modeling techniques. So, a capability that allows analytics experts to build predictive models once and citizen data scientists to reuse them repeatedly becomes critical to delivering true ease of use. The same concept holds true whether you’re dealing with data preparation, data aggregation, or even visualization. Drag-and-drop capabilities that enable non-technical users to reuse pre-built templates (think of them as the analytics version of Lego blocks), and perform tasks in a consistent, repeatable, and standardized manner are critical to the usability of modern analytics platforms.
2. Connectivity to All Data
As simple as this one sounds, you’d be surprised how many tools positioned as “data analytics” solutions lack the ability to truly connect to and analyze all data. Not just big data, but all data. That could mean sensor data from IoT environments. It could mean disparate data living on a modern big data platform. It could mean social, text, or log file data. Or it could mean plain, old-fashioned relational data from transactional systems. And that’s before getting into whether the data resides on-premises or in the cloud. Go back to the reason you’re investing in an analytics solution in the first place – to solve business problems. You can’t solve business problems unless you have a complete view into what’s happening with your business. And that means having the ability to connect to and analyze all of your data. You might not always need it, but when you do, you’ll want to have it.
3. Standardization and Flexibility
Increasingly, organizations need advanced analytics solutions that enable equal parts standardization and flexibility. The modern analytics landscape is increasingly driven by the use of Open Source languages such as R and Python. Solutions that don’t embrace and seamlessly integrate with these and other Open Source languages and technologies simply don’t offer the flexibility organizations need. Users need a solution that enables them to code and retrieve data in the language of their choice. At the same time, businesses need a solution that ensures standardization and governance, so that once models are created, they can be locked down and used consistently across the organization. Fail to deliver flexibility, and your analytics teams will be limited to a fault. Fail to deliver standardization, and you’ll spend a whole lot of time reinventing the wheel and creating multiple versions of the truth. Finding a solution that provides the necessary middle ground is thus critical.
4. Consumable Results
Once again, go back to the reason you’re evaluating analytics solutions to begin with – the desire to solve business problems. There’s not an organization around that can use analytics to solve business problems if the results and learnings of a given initiative are not easily consumable by company leaders and employees who will implement changes within the organization. Executives won’t act on what they don’t understand. So it’s critical to find an analytics solution that delivers easily consumable visualizations, with the flexibility to customize the display of information according to preference and drill down for a bit more detailed understanding if required. It’s all the better if you have the ability to pair that information with a narrative that helps business leaders understand what they’re seeing, what it means, and what actions they should take.
5. Decisions, Not Just Predictions
Speaking of actions, at the end of the day, that’s what analytics is all about, and that’s why predictions alone are not enough. Knowing something is likely to happen is only beneficial if you also know what to do about it, so it’s critical to look for a solution that gives you the ability to map predictions to the specific business decisions likely to drive the best possible outcomes. In other words, moving from predictive to prescriptive analytics. And since the right decision made too late is of little use, you’ll also want to ensure your solution enables you to deploy that decision guidance as close to the point of impact as possible. If you have a manufacturing facility in Singapore, deploying analytics on data warehouse in New York might not help you all that much. The more readily you can push analytics to the actual point of impact, the better.