man's feet in sneakers jumping in air
PHOTO: Kushagra Kevat

I think of data capabilities as a tiered structure: starting at the ground level with basic data capture, through analytics, statistical modeling and finally reaching the buzzwords with the full range of all things big data and machine learning. So as an executive, when should you try to drive your organization to the next tier?

You Are Succeeding at the Current Tier

Moving up a tier will generally highlight opportunities for improvement in the tier you are moving from. If you are not succeeding, trying to elevate can be detrimental to organizational growth, as the existing tier will lack the capability to support the new skill players. 

The two most common challenges I see when shifting tiers are: can the data support the new paradigm, and do we have the people in place to support the next level?

When shifting from the data tier, the biggest challenge is often whether or not the data is structured to enable new uses. Initially data is structured to support vertical accounting views. Shifting to horizontal views, for example looking at trends through time, sounds trivial. However, can you tell me the last time there was a shift in how you prepared and stored the data? What about the last time your vendor instituted a data change? Both of these can throw a massive spanner in the works.

When shifting from pure reporting to any predictive modeling, the largest challenges are typically tools and people. Your organization almost certainly has people who are great at SQL, building reports and explaining any oddities in the data. Is the software engineering culture (e.g., agile, source control, etc.) sufficient to handle the increase in project management and statistical complexity?

Related Article: Stop Torturing Your Data and Other Tips to Reveal True Data Insights

You Have Useful, Inaccurate, Yet Constantly Improving Models With Unexplored Variables

Your team has built some models that improve decision making. In fact they improve it so well, there’s minimal scope for improvement. Stop. Congratulate the team. Find something else to model.

Sadly, the more likely scenario is the team built a suite of effective models that help drive better decision making. However, there’s still lots of randomness and error coming from the system. In this scenario, there may be opportunities to explore machine learning solutions.

The lack of accuracy points to the opportunity to improve. The improvement as you add data says there is more information content coming from the existing data. The model should improve through time. And finally, the unexplored data, particularly if there is some intuition it may be particularly germane, will have information a machine learning algorithm is ideally matched to find.

Related Article: Data Coming Out of Your Ears? Be Thankful

You've Identified a Problem Which Solving Would Generate Massive Returns

There are problems whose very nature requires huge amounts of data, complex algorithms to process, vast amounts of computing power to solve, which will generate outsize returns for investors by solving. For example, self-driving cars. Upgrading your CMS, CRM or customer valuation models all likely fit in this bucket as well.

There are also many problems that require huge amounts of data, complex algorithms and vast amounts of computing power. Unfortunately, the incremental benefit to your organization from them is minimal. Crystal balls are often hard to see through. At least up front, make sure you believe your problem is in the first category.

A similar error I see quite often is building the next great model when nothing will change in the business’s decision process. If a change in the model would result in no change in your behavior, there is no incremental benefit.

Related Article: 5 Ways Analytics Help You Win (and Retain) Customers

Seize the Data Opportunity

Often times we get sucked into pursuing initiatives based on buzz and feel rather than cold hard metrics. Some good guidelines for improving the outlook are that you are succeeding in your current analytics, that there is visible scope for improvement and finally, there is a clear opportunity to seize. Carpe Diem.