The Anatomy of a Real-time Prescriptive Analytics System


In the last post Sanjiv discussed about the 4 types of Analytics and the way Real-time Prescriptive Analytics is helping organizations become operationally agile. But what goes on inside such a prescriptive system? To discuss that, we will leverage the OODA framework developed by military strategist Colonel John Boyd. Though developed for achieving success in air combat, today the framework is applied to a number of fields including business decision making.


As the figure above illustrates, decision making consists of 4 steps – Observe, Orient, Decide and Act. The idea is to traverse these steps faster than the adversary. As Harry Hillaker (chief designer of the F-16 Falcon) aptly put it – “Time is the dominant parameter. The pilot who goes through the OODA cycle in the shortest time prevails because his opponent is caught responding to situations that have already changed.” Of course for the operations manager there is no one adversary – anything that disrupts the day-to-day deliverables is an adversary. So how does OpsVeda help traversing the OODA loop faster? Let us look at each step:

Observe: The aim at this stage is to collect as much information as possible. Extractors installed in the various operational systems transmit transaction changes to OpsVeda in real-time, which in turn gets stored in a Process Agnostic Data Store (so that processes that are very organization specific can be easily mapped). Change the delivery date of an order and OpsVeda will know immediately… Put a credit-block on a customer – again OpsVeda will know immediately. In short the users can do anything on the transaction systems but the extractors will ensure that OpsVeda has all the information about the current operational situation.

Orient: This stage is about analyzing the current information in the light of past experiences and other background information. To ensure speedy analysis, OpsVeda keeps the current & historical data, and the expected path of a transaction (configured) in memory. In real-time, proprietary algorithms in the analytical engine of OpsVeda compare the information sent by extractors against historical patterns, and does the calculations to assess if the latest changes would disrupt the process (now or few steps ahead). For example, historically if it takes 2 days to remove a credit-block and a shipment is scheduled the next day; it is an indicator that the customer needs attention.

Decide: At this stage, the goal is to evaluate available options and to decide on the course of action that would prevent the disruption or minimize its impact. The Analytical Engine again digs into historical patterns – what is the current CSAT for the customer and how was it impacted in the past when delivery was delayed by a day, how fast has Finance lifted the credit block when the Account Manager got an email assurance from the customer. Based on the patterns and constraints (eg. penalties, bounds on CSAT etc.), the Analytical Engine will prescribe the best course of action for the operations personnel to tackle the situation.

Act: In the business world most actions around exception situations involve collaboration. OpsVeda’s Collaboration features enables personnel to exchange unstructured information and get the necessary approvals where needed. When action is taken, once again the extractors sense and relay it back to the Analytical Engine immediately. Thus the loop is closed and the patterns are updated immediately.

The key enabler is the real-time capability of all the components. Unlike batch jobs, extractors send data when the transaction changes. The Analytical Engine analyzes and prescribes the moment data from the extractor comes in – thus the operations dashboard is updated with potential exceptions and prescribed action in real-time. Powerful radars, ergonomic cockpit, well-trained pilots etc. enable faster traversing of the OODA loop in air combat. OpsVeda does that for Order Fulfillment, Supply management, Manufacturing & Distribution Operations.