Big Data has led to a discussion of predictive analytics in supply chain management. I’m not sure we have a good understanding in the field of how predictive analytics differs from forecasting.
Here is how I would differentiate the two.
All forecasting involves making a prediction, but not all forecasts are based on Big Data. Using historical data to make a regional forecast for most companies would not require Big Data. Forecasting demand at the store level probably would. The more inputs, the more granular the data, the more granular the forecast, the more we are in the realm of Big Data analytics. In short, when I hear predictive analytics I am apt to think this is a forecast based upon Big Data.
In logistics, we have not seen all that many examples of Big Data analytics. Using a demand forecast at the store level to create transportation plans –Terra Technology and JDA do this in different ways – would be two examples. Until recently, I would have been hard pressed to come up with another example.
That has changed. I recently had a conversation with Adam Compain and Will Harvey, the cofounders of ClearMetal. Mr. Compain is the CEO, Mr. Harvey is the Head of Technology. ClearMetal is applying predictive analytics to help Ocean carriers improve their profitability through better placement of ocean containers and more efficient vessel space planning. Even though the issue of Ocean carrier profitability is not at the top of my interests, the way ClearMetal is solving this sector’s problems is very interesting. And it is easy to see that many shippers may have deep data sources that could be mined in similar ways.
I’ll focus on ClearMetal’s container placement problem in this article. The company is endeavoring to make more accurate bottom up forecasts that go out 8 weeks on where containers are likely to end up and where they will be needed.
The company is using booking, container event, and vessel voyage data from carriers. They enrich this data by combining it with third party data such as currency rates, commodity trends, weather data, and other forms of data as well. They then create a chain of events to support an end to end shipment. If an electronics goods manufacturer wants to ship out of Minneapolis to the Port of Santos in Brazil, what is the probability distribution associated with having an empty container available in Minneapolis in eight weeks? Past order history for shipments into and out of Minneapolis, combined with commodity information (what goods are in those containers and is demand growing or weakening for those commodities) can help make that prediction. But containers into Minneapolis may have started as voyages from ports around the world. What are the chances (the probability distribution) these feeder voyages will be delayed at the port and intermediate rail heads?
The next step is to take the different supply chain lanes into Minneapolis and run simulations. There may be a certain range of chances that a certain set of containers that would eventually end up in Minneapolis will be delayed being unloaded at the port of Long Beach, delayed being loaded on the train in Long Beach, delayed in transit in Chicago, etc. Different estimates are used: a one day delay at Port of Long Beach is one simulation event, one day quicker through Chicago than expected is another, and so forth until millions of simulations are run.
Clearly this is a Big Data solve. It is interesting math. And it could provide a lot of value to Ocean Carriers that have been facing tough economics for years.
Does the math work? ClearMetal says it does. Based on some pilot engagements, top ocean carriers are telling them this solution has the potential to save them tens of millions of dollars in reduced container inventory. They believe they have also proved they can enhance revenues by making sure containers will be in the right location to support door to door deliveries.
But whether ClearMetal succeeds in the market or not, it is clear that interesting, new predictive analytic solutions will increasingly be available in logistics.