Originally posted 12/14/16 on TeradataVoice by Ryan Garrett, Teradata
The stunning confluence of massive computing power, incredible volumes of data, and stunning advances in analytics, now readily available to businesses in all industries and of all sizes, has changed the face of business decision-making. Data scientists and business analysts have a remarkable sandbox in which to build models, test ideas and pinpoint solutions to challenges and capitalize on opportunity.
There is no real question of whether analytics are a good addition to businesses. The great challenge that remains, however, is how to close the gap between what data scientists accomplish in their laboratory playgrounds and what is practically usable throughout an organization. The impact of a data science team’s great work will be severely restricted if there is not an effective means to distribute the results in ways that are usable, understandable, and actionable by everyone else.
Connecting the lab to the business, or more importantly analytics to the action, is an essential step for data-driven businesses. To help smooth the way, I propose five steps that will help in the conversion of analytics victories into company-wide successes:
Step 1: Document the data
When a user lays their eyes on a business solution, it should be easy to find out what data underlies that insight. That transparency is essential to establishing trust. Just because a decision maker isn’t a data scientist doesn’t mean that they don’t have great knowledge about the business. Decision makers need confidence that the data being used is properly sourced and that subsequent analytics are built on high quality, credible data. Put simply, people need to trust and understand the data.
Step 2: Let users run queries themselves
Empowering users to run queries themselves is essential. Many users who know their data just want to run their queries. Some will want deeper visibility into the underlying SQL. Those users want to see the details of the query process and trace the steps being taken. Documented queries, including any intermediate data objects created along the way, can be important to executives or managers who want to dig deeper into how queries work under the covers.
Step 3: Document and explain the analytics
Once a trusted set of data is amassed and analytics are being performed, the analytics being used needs to be made clear. While not every person downstream of analytics will require that information, the super users and decision makers who champion the application of analytics across the business ultimately establish the baseline for the trust the entire company will have in those insights and processes. With documentation that explains the analytics being applied, advanced users who want to understand how information is being derived can get that information on their own, and on demand, without having to wait for a meeting with the data science team.
Step 4: Apply illuminating visualizations
Closely related to explaining the analytics is the application of visualizations that the business user can understand. A sigma visualization might deliver stunning depth to a data scientist who understands how a network graph represents modularity and different occasions, but perhaps a simpler version of the same graphic may be necessary for people on the front line of business with jobs to do and other concerns. At the other extreme, the typical bar chart might be easy to digest, but that simplicity may lack the depth to support a business decision. This is not about oversimplifying: far from it. Finding the right visualization is an essential balancing act that will provide the business user with context to more fully understand and apply analytics insights.
Step 5: Plan for improvement
Analytics is by nature an iterative process. It is not something that is deployed once, is one hundred percent “correct,” and the work is done. Not only will insights change with time, but how analytics best mesh with the business is itself an organic, evolving construct. Having a feedback mechanism that allocates time and a process for understanding what does or does not work, and why, is a key differentiator between processes built for one-off successes and analytics that are built to deliver continuous results.
The path to analytics victory is not forged by suddenly turning everyone into an analyst or data scientist. It’s recognizing the strengths of both the data team and the business team, and then making sure that an interface exists to encourage each to make the other stronger.