Over the past few weeks, I’ve discussed the technology and cultural changes you need to make to embed analytics capabilities into the fabric of your organization—to make it pervasive. I’ve talked about how to deal with data in this environment, and how to design systems by listening to those who know best what you need: your people. However, there’s one component that ties all of this together. It’s the philosophy, or set of guiding principles you use to deploy your analytics. That philosophy is AnalyticOps.

Before I tell you what it is, you need to know why it is. One major problem all companies of all sizes encounter when they build analytics systems is that they take too long to move the project from development to deployment. By doing this, they waste money, time, and resources that could be deployed elsewhere.

To solve this problem, it’s essential to develop analytics systems with three caveats in mind:

  • They need to move to production, at scale, and prove their value quickly
  • They need to solve the problem for which they were designed
  • They need to deliver insights quickly

Well, sure, you’re saying, of course they do. But here’s the catch: many analytics proof-of-concept (PoC) projects succeed at the PoC stage. Instead, once they’re put into production, they fail spectacularly, or they die slowly of neglect and non-use, because they don’t provide the value they promised.

Enter AnalyticOps. It truly is a philosophy—an innovative way of thinking—about developing successful analytics systems. AnalyticOps combines process and technology to close the gap from idea to insight. It helps you bring analytics PoC projects to production quickly, and at scale, to drive value. It brings analytics from theory and numbers to production and value.

How AnalyticOps is Different

AnalyticOps differs from traditional development approaches in three critical ways: development style, time to deployment, and model development and management.

Development style. With traditional approaches to development, the development and production stack are generally the same. They use the same data and there’s not much insight provided. With AnalyticOps, however, development and model exploration uses large volumes of historical data to develop the models, but once the system is in production, models are typically deployed into a real-time scoring process such as scoring engines, web services, etc.

Time to deployment. In traditional development methods, quality assurance is a laborious process, and valuable analysis time is often lost. With AnalyticOps, the review process is expansive, but it’s adaptive and the approval processes are often specific to individual models and use cases. Plus, everything is done simultaneously to model development so that the process is faster, and the time to deployment is decreased.

Model development and management. In standard development approaches, the number of applications or models in development at one time is relatively small. With AnalyticOps, hundreds, or even thousands of models are developed, tested quickly, reworked, and put into production. This enables you to deliver insight—and value—quicker than with traditional approaches.

Why it works

At its core, AnalyticsOps is driven by outcomes. Simply, the success of models, and of deployment, is measured by how well they meet your goals. If they don’t—and many won’t—they’re abandoned, and new models are developed. The key to this is twofold: one, you can’t be afraid to fail, and two, you have to learn from your mistakes and build on the learning.

What I mean by this is that failure is a necessary evil in the development process. It happens, and you deal with it, learn from it, and move on. That’s the beauty of AnalyticOps. It enables you to fail forward. You fail, learn, fix, and iterate. Then, you operationalize the results—you industrialize the process. You take the success of the model in production and spread that success—and the methods you used to achieve it—to all parts of the organization, thus embedding analytics capabilities—making them pervasive.

What it gives you

Simply put, it gives you the insight you wanted when you started your analytics journey. AnalyticOps—when done right—gives you the ability to productionalize analytics, to see across divisions and business functions and deploy analytics large-scale. With an AnalyticOps approach, you can rapidly build and deploy applications that enable you to gain insight into your most complex problems.

Once you have the insight you need, you can develop that into intelligence, which can be translated into strategy and tactics to drive those outcomes that have the most impact on your business. You can understand where you are, where you’re going, and how to get there. That changes the game.

General manager of Teradata Consulting and Go-To-Market Analytic Solutions. Thought leader in analytics, business intelligence, big data, and business transformation. My passion is helping my clients drive value through data and achieve a sustainable competitive advantage.

Leave a Reply

%d bloggers like this: