Deloitte recently published A Delicate Balance, Organizational Barriers to Evidence Based Management. A surprising conclusion was that the biggest barrier to success with data analytics projects is the behavior of the humans and not the quality, volume or reliability of data. Regardless of how difficult data science can be, success depends on the usual thing…us. We stand in the way of ourselves.
Deloitte’s report points to four common failure areas for Big Data and analytics projects:
There’s an enormous gulf between completely automated algorithms like the Netflix movie recommendation engine and purely human judgement, like what a physician provides in the hectic emergency room. Within that broad spectrum lie a thousand possible combinations of human judgement and machine responses. Layering in the appropriate amount of analytics alongside business process and decision points is a complex recipe and getting it wrong can be disaster for a business or a patient.
If the role of analytics is poorly understood, there is high likelihood that efforts will be met with irrational enthusiasm for all things analytical or the opposite…skepticism from those either lacking understanding or feeling threatened by a new paradigm. There needs to be a process for organizational change.
The more we value strong judgement and those we believe demonstrate it, the harder it is to accept that there can be a better way to make decisions. We believe what we believe and convincing us otherwise is nearly impossible. After all, didn’t we get here without analytics?
Equally challenging is the belief that the best data put through the perfect model will achieve an ideal outcome. This is also the best formula for organizational paralysis as we strive to reach an unattainable level of data cleanliness and model perfection. Perfect is the enemy of good enough when it has a negative impact on revenue or cost.
In the end, the answer comes down to how the organization establishes goals, educates stakeholders, overcomes understandable bias and stays humble about expectations. It comes down to a balance of process and technology. Funny enough, it always does.