All predictive data sciences, algorithms and analytics have their day of reckoning – What happens when their insights are disregarded? What happens when actions and recommendations are ignored and the old ways followed? What happens, for example, when predictive store orders created to mitigate Retail Out of stock are consistently subjected to manual overrides?
Predictive systems are traditionally put in place after millions of $ in investments. Expensive infrastructure and tools are bought, and expensive consultants hired. In spite of this, if business users ignore recommendations,
Let us examine why:
- “It is my job, why delegate to a math model”
- “I do not quite trust it… so will override it..”
- “It is counter intuitive.. goes against my experience”
Let us look at an anecdote:
A Railroad company in America experienced snow storms and consequent downtime on account of tracks having snow and debris on them. This affected thousands of miles of railroad. The physical cleanup of the tracks was completed in a day. However, the Railroad Company struggled to get the rail traffic back on track (literally!). Each section was doing its best to optimize rail movement. However, the situation was not improving.
Finally the CEO got involved; he reviewed the ground reality; he then mandated that everyone follow the directions of the software to a T. The traffic was back in action in a couple of days.
Some traffic directors had protested about the software recommending “wrong” actions, like taking an important train into the side tracks of a station. The recommendation was counter-intuitive, yet correct.
Lesson – Trust the prescriptions from predictive data sciences.
Data sciences and predictive analytics have come a long way in understanding large amounts of data, figuring out patterns and recommending actions. After validating the analytics, learn to trust and act on it. And feedback errors so that the algorithm improves.
And to do this, in our opinion, C Level commitment and belief in data driven decision making is necessary.