A Smarter, Faster Way to Operationalize Analytics

[twitter-follow screen_name=’@dinojain’ show_screen_name=’no’]

Operationalized analytics is the new black. Everyone wants it, even if they don’t really understand what it entails, or how to do it. The mantra is to ‘just get it done.’ But what is it? I define operationalized analytics as having a sustainable, repeatable process to quickly deploy predictive analytics models throughout the organization.

With the advent of more sophisticated analytics tools, coupled with the deluge of unstructured data, this ‘mass-production’ model development capability will quickly become mandatory in deploying successful analytics solutions. Those companies that operationalize their analytics quickly and effectively will get a leg up on their competition.

If you want to succeed at operationalizing analytics, here’s how to do it the faster, smarter way:

  1. Get good data and employ organizational data governance
  2. Build an iterative, business led development ecosystem
  3. Deploy self-service capabilities that fit all (or most) users

Good Data and Governance

If I ask you what the team name of the Major League Baseball club in Atlanta is, and you tell me, “It’s the Atlanta Braves,” I would believe you, because I know that to be true. But what happens if you tell someone else who asks that the team’s name is the Dodgers, and you tell yet a third person that it’s the Yankees? Eventually people will believe that your information is sketchy and you’re probably not trustworthy.

Operationalize AnalyticsIt’s the same with analytics systems. If people don’t get the same answers to the same questions, they won’t trust the system. That’s why it’s critically important to have consistent—and consistently good—data throughout your company’s information repositories. It’s crucial to have strong data cleansing and validation processes in place to ingest and standardize both structured and unstructured content (as much as possible).

Consistent data also requires that you put a strong governance initiative in place to ensure that data policies are set and followed. The governance effort must also be empowered—and accountable to—the C-suite. Data governance isn’t optional when you operationalize analytics. If the data isn’t good, the models won’t be either.

An Iterative, Business-Led Development Ecosystem

Operationalizing analytics is an enterprise-wide endeavor. It’s not enough to build models that work for one function or department. The models should be built within a framework that helps you embed analytics into organizational business processes, based on identified business needs. The models you build should drive faster, deeper insights into what’s really happening and how to respond effectively.

By utilizing an appropriate modeling framework, with accelerators that help you quickly construct models that reflect user needs, you can build an analytics ecosystem where you can develop, deploy, consume, and scale—in an iterative cycle—across the enterprise. As models are deployed, your framework should also give you the capability to monitor and manage performance to maintain consistent results and system stability.

Self-Service Analytics that Fit All (or Most) Users

Users come equipped with a variety of skills. Some are power users who can practically develop the system themselves. At the other end of the spectrum are those who need more than a little hand-holding. Most users fall somewhere in between. Your analytics model-building environment must accommodate as many of these users as possible, and in the most efficient way.

That’s where self-service analytics comes in. Most top-notch analytics model-development frameworks give users of all abilities the capability to construct models and adjust them, based on the results they get. Users can explore the data, ask new questions, and gain new insights based on results from analyzing different hypotheses fed into multiple models. It’s a democratized sandbox where users can play on equal grounds, despite their level of expertise and can participate in the generation of new insights that help grow the bottom line.

The wild-west days of analytics deployment, where different business functions developed analytics systems to suit their own needs, is rapidly drawing to a close. Most savvy companies have realized that, to build analytics that yield consistent answers that drive deeper, quicker organizational insights, you need to operationalize analytics—to make it into sort of an industry within your organization.

In the rush to operationalize, make sure you do it right. Make sure your data is good, that the business leads in building repeatable models, and that the capabilities can be accessed by the people who need them.

I’d love to hear what you think. You can comment or DM me on Twitter, and please follow me! [twitter-follow screen_name=’@dinojain’ show_screen_name=’no’]

You can also connect with me on Linkedin, or email me at anuraag.jain@thinkbiganalytics.com.

Leave a Reply