Ever meticulously planned a long journey only for it to be thwarted by factors out of your control? As you set off to leave your house imagine you’re greeted with a flat tyre on your car, or a motorway accident is causing a 30 minute delay; there is a whole-spectrum of uncontrollable events that could cause you to have to change your plans. Now multiply the magnitude of your journey and the list of possible unplanned events by 100-fold and we are close to comprehending the level of challenge faced by operational planners in industry and military organisations globally.
So how do you predict the unpredictable?
Predictive analytics leverages historical data and identifies crucial relationships including risks and opportunities, which are used to forecast potential scenarios. Predictions could be in the immediate future, for example predicting that a component of an armoured vehicle will need replacing the next day, or the far future, such as predicting the quantity of spare parts needed to maintain those vehicles over a year of uncertain activities.
Data analytics can be categorised into 4 key types:
- Descriptive Analytics: What happened?
- Diagnostic Analytics: Why did this happen?
- Prescriptive Analytics: What should we do next?
- Predictive Analytics: What might happen in future?
Data-driven prediction models are steered by machine learning algorithms which can process, and make sense of, multivariate data with ease. They learn what has happened previously, and generalise to new situations allowing for a forecast of figures relating to the new situation. In short, you can train a predictive model on historical data and use it to predict scenarios for an event that hasn’t yet happened.
Planning operations, when informed by predictive forecasts, can:
- Reduce unexpected costs
- Lead to improved security, certainty and reliability around risks
- Safeguard operations against unplanned situations
The real-world value of predictive analytics was demonstrated by Techmodal to the Royal Navy via analysis of historical Carrier Strike Group data. Utilising operational defect data covering a 3 month period we accurately identified 83% of equipment failures in the task group, giving a 38 day lead time on average. Predicting failures weeks in advance, and with relation to expected usage on planned operations and training, allowed stakeholders to more accurately calculate component stock needs and engineering works.
Similarly, Techmodal have worked with the British Army to predict future equipment failures onboard Terrier combat engineer vehicles using historical usage, maintenance and defect data.
Accurately planning inventory demand and engineering schedules are just some of many ways that predictive analytics can optimise operational planning. Let’s take a deeper look:
Inventory forecasting ahead of scheduled training and planned operations is invaluable when it comes to ensuring demand is met and money isn’t wasted on unnecessary inventory and warehousing costs. While looking back at past demand is useful, predictive models add sophistication by accounting for numerous complexities such as seasonality, demand growth, geographical requirements, political appetite, to name just a few. At a strategic level, factors affecting demand can be major or peripheral, and predictive models can detect key factors and account for them in the forecasting, no matter how minor they may seem to strategic planners.
Forecasting for upcoming operations is vital in itself, but it’s worth noting that understanding the drivers of demand peaks and troughs is also hugely valuable for the shaping of better processes and policies in the future.
Predictions can be made in near-real time using systems that collect data continuously and feed into automated analytical pipelines. Sensor components on machines or vehicles, for example, monitor usage, environmental variables, inappropriate use etc., and algorithms that continuously analyse those data in real-time can predict component failure and expected maintenance.
This allows organisations to be proactive in minimising faults and failures, rather than reactively dealing with them after the fact, ultimately reducing maintenance activities and costs.
Strategise for the future
Regardless of the planned operation, predictive analytics can provide insights that are crucial for effective execution. Stock levels, rising costs, best routes to travel, equipment maintenance and malfunction, are just some of the vital information that can be accurately forecasted and predicted to allow planners to efficiently strategise and plan. Building strong predictive analytical capability to forecast future events, really is the future of operational planning.
Consultant – Data Science