Originally published on Manufacturing Business Technology
A hi-tech manufacturer recently reduced recall costs by $20M annually and reduced test cycle times by 50 percent. A fuel-injection pump manufacturer reduced pump calibration times by 30 percent. A global steel manufacturer identified up to $1.3M of annual savings once they managed to aggregate data from over 20 different sources. This is the emerging and lucrative market for advanced analytics in manufacturing operations — and we’re just at the tip of a very large iceberg.
Manufacturers are constantly striving to improve their performance — and over the past 20 years — have employed innovations into their operational strategies, such as continuous improvement (Kaizen), six-sigma/statistical process control, lean synchronisation and Total Quality Management. However, there is still considerable variability in performance both within and between organizational enterprises. Industry 4.0 is hailed as the next industrial revolution — top performing companies are already implementing their strategies for digitalization, automation and extracting value from data.
The vision of the Smart Factory that incorporates the Internet of Things (IoT), Internet of Services (IoS) and automation across the value chain will emanate ever increasing quantities of Big Data for which an advanced analytics platform will be a key enabler for converting this information asset into value. Successful manufacturing companies will be able to leverage their Big Data assets and apply advanced analytics technology in order to improve production yields, increase throughput, reduce cost of quality and improve customer satisfaction by reducing the number of field failures and warranty claims.
Bridging the gap between the copious amounts of “Big Data” that will be collected over the next 10 years and using that data to deliver value at the operational level is where the rhetoric meets reality. Often the data sits in different information silos, typically in place to run operations — incoming materials and component data, manufacturing process parameters, laboratory analyses, testing results and customer experience data. These disparate and eclectic data sources are often in different formats, include different genealogical reference points and are owned by different parts of the organization. Joining them together into a single, coherent data table suitable for analysis can therefore be an almost insurmountable challenge.
However, this is exactly what the companies referenced at the beginning of this article are doing. Not only are they investing in platform architectures that allows them on-demand access to coherently joined data tables from across the value chain, but also in the advanced analytical tools necessary to identify value and the real-time environments necessary to extract it. This includes the deployment of real-time predictive models, enterprise-level quality dashboards and case management systems for rapid access to root-cause diagnostics.
A leading hard-drive manufacturer has recently built an advanced analytics platform that enables them to predict which batches are likely to have failure rates above specific control limits and thus to intervene before these hard drives take up valuable space in the testing bay and bottleneck production. Data from across the value chain was extracted which included supplier data, assembly line data, test-bay parameters and quality records. In all, over 3000 variables were collected and analyzed in order to ascertain which parameters were key to predicting an increased likelihood of drive failure. Advanced analytical models were created that could predict failures hours and days before the drives actually reached the testing bay resulting in higher throughputs worth millions of dollars annually.
A comprehensive platform architecture was developed and implemented that created an end-to-end solution that included data collection, data aggregation, model re-training, case management, root-cause diagnostics and management reporting. This adoption of state-of-the-art advanced analytical technologies is not limited though, to state-of-the-art industries such as hi-tech hardware manufacturing. Traditional industries such as steel, pharmaceuticals and the automotive industry are heavily investing in this area.
According to a recent market report by Transparency Market Research the multi-billion dollar predictive analytics market is expected to grow at 17.8 percent CAGR from 2012 to 2019 with the retail and manufacturing sectors expected to grow faster than any other segment. Companies with a roadmap and vision for digitalization and Industry 4.0 are well positioned to maintain and improve their position as this new competitive force separates those companies that embrace the age of analytics and those that don’t.
Key to Operationalization Success
The key to success in the application of advanced analytics in manufacturing environments is the recognition that the data is only part of the story. Simply throwing data at advanced analytical algorithms won’t lead to value. At best, it will waste time and at worst, it will result in decisions being made on unreliable information. Process engineering and domain knowledge is equally critical in ensuring that the analytic algorithms generate credible information that is actionable in what is often a safety-critical environment. Analytics is a decision-making aid, poor decisions in manufacturing can have costly implications. Credibility is key and this can only be achieved by the synergistic team-working of several disciplines which includes process engineers, shop-floor personnel, business unit managers, quality management, IT group, maintenance teams and, of course, the statisticians and data scientists. A highly effective team will have access to each of these stakeholder groups and will actively involve them in any project.
A global steel manufacturer adopted this approach on one of its European steel mill facilities. A team involving veteran steel domain experts, quality/lab personnel, operations management, IT group resources and data scientists worked in collaboration to both collect, join and most importantly, understand the context of the data and the outputs of the analytics. The result was the identification of up to $1.3M annual benefits based on identifying major contributing factors to product defects that could be reduced with actionable changes in operational procedures.
The operationalization of advanced analytics is a journey with several steps along the way. Many companies evolve their strategies in an agile-manner such that each progression provides tangible benefits that can fund the next step. Others, explicitly develop a comprehensive platform from the outset, in order to tackle a specific operational need (such as yield improvement or predictive maintenance).
The first step on this journey is typically operational reporting — which tells operations and managers what is currently happening on the plant in a format that is efficient to digest (current sight). The next step is historical reporting — where past performance can be critically analyzed and ideas generated for future improvements (hindsight). The third step is where the historical data is mined for relationships in order to identify why things happened (insight) and the next step is to use predictive analytics that identifies patterns and behaviors from historical behaviors, in order to predict future performance (foresight). The final stage on the analytics journey is the actionable piece (prescriptive analytics) — how do you control what happens — this is where the real business outcomes are realized.
Advanced analytics can provide root cause diagnostics that can be used either in a supervisory manner (i.e. advice to operations on what actions to take in order to mitigate the forecasted event) or in an automated manner where the advanced analytics has some level of autonomy to take automated actions. Often the solution is a hybrid of the two.
Clearly, leading manufacturers are preparing for this journey and many are already on their way and realizing the benefits that advanced analytics can bring to manufacturing operations. The information age is maturing into a more intelligent world where data isn’t just collected and reported, it is analyzed, understood and leveraged to support operational decision-making across the enterprise.