I recently stumbled on this surprising chart of parole-board rulings and their success rate by time of day. Aside from prompting increased concern for the fairness of our judicial system, it struck me as excellent illustration of some often overlooked aspects of data analysis.
It is easier to ignore instincts than data. We have long been aware of the way time of day can bias our responses. Everyone has heard presenters joke about being the last before lunch and therefore challenged. Presenters that stand between their audiences and happy hour may even express doom. Parole boards likely noticed the same potential, and I would bet they believed in their ability to overcome such tendencies. Yet a quick chart of the data shows their noble sense of fairness being defeated by the basic human limitations of fatigue and hunger. Data expressed this clearly cannot be shrugged off. This chart demands action.
Powerful insights are often derived from unusual questions. Traditional BI has focused heavily on business performance measurement for the past few decades. BI departments devote much of their energy to disciplined data management, proper data governance, data quality assurance, and operational stability. All of this is aimed at ensuring that business measures, usually in the form of key process indicators (KPIs), are correctly captured and effectively displayed on dashboards. When applying this approach to the parole process, we might focus on number of rulings per day, percentage granted versus denied, percent of violations among those granted, etc. I do not anticipate favorable ruling percentage by time of day would easily enter the picture. That question came from someone creatively pondering the last-before-lunch phenomenon and wondering what the data might show. This is why the new trends in data and analytics emphasize notions of self-service and data discovery. We want each organization positioned to ask a richer, less anticipated assortment of questions.
Simple may exceed the importance of complex. This insightful chart was easily created: small volume of data, structured with just a few fields captured for each ruling, and visualization achieved with a standard line graph. No real-time stream processing or machine learning or complex statistical analysis required. Valuable insight into a hidden injustice was produced not from asking a hard question, but the right question.
That initial “right question” starts a conversation. The alarming answer to the first query should lead us to greater inquiry. Does the nature of cases vary by time of day? Are the cases sorted by degree of difficulty within each block of the day, skewing the impression of fatigue factor? Can we change results with more frequent breaks, snacks, stretching exercises, reverse-order second appraisals, etc.? Our data analysis should be iterative, tweaked to respond to corrective actions, driving deeper and deeper toward optimum justice.
Our favorite tagline for our data and analytics services has been “better decisions faster.” We should also tout “better and better questions, leading to best results.”
That’s the analytical way.