This week, Dan Roberts senior consultant at Cambashi, discusses Big Data analytics
Experts at industry analyst and market consulting firm Cambashi contribute a regular blog for TM titled Silos Changing exploring how new software applications enable manufacturers to implement business initiatives in the new economy.
In previous blogs, we have written about the advances being made in the harnessing of social media. It is a rich source of data about how a company’s products are perceived in the marketplace. However, social media is just one of a whole host of sources of data about how products are used, perceived, bought, sold and even made.
Many of these data sources are readily accessible to manufacturers. For example, modern cars are filled with a host of sensors that are used in the engine management, internal climate control and other management systems.
But the data that these sensors produce is normally used in real time and then discarded. If the car was able to analyse this sensor data, it would get a valuable insight into how it is being driven.
This could have implications for when the car needs servicing, enabling dynamic servicing scheduling, where the car determines how often it needs a service. High-mileage cars that are driven smoothly on long journeys may not need servicing as often as medium-mileage cars used for short, sharp bursts.
For companies that manage fleets of cars this could provide a valuable source of cost reduction, especially if it also reduces the instances of cars breaking down through insufficient service. This data could be collected during servicing. Car manufacturers could then track aggregated data from their customers to understand how to improve their products.
Similarly, machine tools could flag when they need to be serviced, enabling more uptime on production lines by eliminating both break downs and unnecessary services. Much of this data is what the data company Teradata refers to as “lots of data” rather than “Big Data”.
The distinction that Teradata draws is to do with how difficult it is to extract useful information out of the data. For example, much of the data produced by production lines is well-structured and easy to query using SQL. However, time-series data is difficult to extract with standard SQL queries, but can often give vital insights into how events occur. This could allow companies to anticipate and prevent production outages. Alternatively, they could discover positive outcomes, such as how to load machines for reduced maintenance requirements.
Using the MapReduce programming model allows data scientists to uncover patterns in a series of events. Teradata’s core client list includes banks, utilities and retailers. These customers are able to track interactions with customers that allow them to learn, for example, that a pattern shows a customer is likely to defect to a competitor.
Manufacturers could use similar techniques to anticipate when machines need servicing, rather than waiting for the sensors to tell them. This could allow the schedulers more leeway to route around the machine, or schedule servicing in the most efficient time slot. For discrete manufacturing, or batch manufacturing, this could drive new levels of efficiency in the production line. For non-stop production lines, it would be hugely valuable. It is now possible to spot the patterns in real time and prevent problems. Whilst most companies will still be left scheduling maintenance at more regular intervals than is absolutely necessary, the data-savvy companies will be forging ahead.
As companies start to look for the patterns in their sensor data, they will start to discover that they know far more than they understood. The role of data scientist, who is the person that looks for the patterns in unstructured or semi-structured data, will start to become more prevalent. The traditional data-heavy industries have already started along this road. The forward-thinking manufacturing companies that are looking for their next competitive advantage should be following them.
In future blogs we will go on to write about other potential deployments that enable manufacturing companies to make use of their valuable data.
