Data Science – How Do I Turn Insight Into Value?

Frank Coleman

In many of my past blogs I focused on building a data science team and getting started down the path of advanced analytics / Data Science. I originally used the slides below at EMC World this past May. They show the team and a high level processes flow. My MythBusters blog talked about our journey to the Prototype and Validate stage or “Time to Insight”.  Now I’m sharing my thoughts on getting value from your work: the Operationalize phase of that process. I refer to the Prototype & Validate phase of this process slide as “Time to Insight”. If you are fortunate enough to get this far down the path, you know it wasn’t easy. Well get ready because getting “Value” is even harder. The technical challenges of acquiring the data, although sometimes painful, are no match for the show you’re about to put on to get value out of your insight. Operationalize or “Time to Value” The second part of that process flow, Operationalize, is where you start to see the value of the insight. Here are a few lessons I’ve learned on the way:

1. IT infrastructure – You created a model that produces an output you believe can have real business impact. But the infrastructure you used to create this “Insight” may not be the same infrastructure you use to operationalize your model or output. Here are a few questions to help you figure out what level of support you’ll need:
   1. Cadence – how often do you need this model/output run? Real-time, Daily, Weekly?
   2. What other data do you want to join with this data?                            - Where are these data sets located and how often are they updated?
   1. What applications are impacted? As an example, if you’re trying to impact workflow, most likely you will be adding something or changing logic that feeds the workers.
2. Create a list of all the use cases for your output. You may have started with one use case in mind but four or five more will likely pop up along the way.                -As an example, you create a data set that helps customer service proactively address customer problem areas. But wouldn’t this information be helpful to Professional Services, Engineering, and others? Each group may form entirely different initiatives to operationalize your results. For each use case you should have an ROI calculated to help create a buzz and get funding to implement.

3. Executive Sponsorship – I’ve said this a million times in my posts around “Time to Insight”. It’s even more critical for “Operationalize” because this will cost money to implement. IT budgets are tight. If your model is considered “nice to have” you are dead in the water.  This is why the ROI is important. Hopefully you thought about this before you started the project in the “Time to Insight” phase.
4. Consider the team required to operationalize your results. It may not be the same team that got you this far. You may also want to consider transitioning this to an Operations / Program Management team.
   1. You need a really strong Project/Program Manager who can “sell” the value and influence the teams to make change.
      1. i.   Change to Process
      2. ii.  Change to IT infrastructure / Tools
      3. iii. Change to Culture
   2. Refocus the Data Science team on what’s next or expanding the existing model so they don’t get stuck in the waiting around game of “making stuff happen”.

These are just a few lessons I’ve learned along the way. I hope you found them helpful. If you are operationalizing your results I’d love to hear the challenges you may be facing that may be different from my list.

Data Science – How Do I Turn Insight Into Value?
Frank Coleman