These companies are struggling to improve their operations to increase their production yield, operational efficiency, and product quality. This has been driving demand for Big Data analytics solutions over the past few years and as part of their evolution, electronic OEMs, particularly those highly quality-sensitive such as automotive, networking, and smart electronics, are moving in the direction of Big Data product analytics and predictive analytics.
However, the enormous amount and variety of data electronics OEMs deal with and lack of control over manufacturing data and standardization make it a challenging task for these organizations to extract value from the data they collect. Engineers waste time trying to locate the data they need for analysis and contextualizing it to perform analytics and finally act on it.
In addition to these challenges that are common across industry verticals, most electronics OEMs have adopted a highly complex supply chain typically consisting of multiple contract manufacturers and semiconductor suppliers. To realize the benefits from Big Data analytics, these companies thus require insights throughout their distributed operations, including into facilities they might not own.
Furthermore, the separate supply chains of electronics manufacturers and semiconductor companies prevent overall supply chain management and there are silos within each supply chain. Data are not shared between the different phases of the manufacturing process, even those taking place at the same location.
Despite these challenges, the benefits from Big Data analytics are simply too good to pass on. The savings generated by Optimal for customers, for example, are testament to this: analyzing customer data from over 50 billion devices, the company estimated to have saved its semiconductor customers over $250 million in only 12 months by reducing test time through eliminating redundant and unnecessary tests, reducing test costs by using advanced test methodologies, detecting product drifts that would have affected defective parts per million (DPPM) performance, increasing productivity by using automated rules that optimized manufacturing throughput and eliminated supply chain inefficiencies, and increasing visibility into their entire supply chain.
While semiconductor vendors have tapped the potential from Big Data analytics solutions, the emergence of Big Data analytics solutions spanning the entire semiconductor-electronic ecosystem signals the beginning of a new era for electronics OEMs. With bi-directional data movement across both supply chains from IC and MCP, through PCB, PCB rework, systems, to in-use and returns, electronics OEMs can easily save millions of dollars per year. With data feed forward (DFF), for example, customers can leverage prior test results within a current test to screen out suspect devices from a good population. Using the board or device ID, previous test measurements are retrieved and analytics performed while the board or device is sitting on a functional tester undergoing tests. Comparing test results between current and prior test operations in real time enables manufacturers to identify issues such as excessive parametric drift between certain test results which can be used to identify a good device that has passed all the required tests but will probably fail prematurely. The ability to quickly and automatically retrieve historical test data that can be used in real time to screen out bad devices from good populations make it possible for electronics OEMs to meet aggressive quality goals while maintaining high yields.
This being said, the promises from Big Data analytics solutions will not materialize with the use of any solution labeled Big Data analytics solutions for manufacturing. A number of solutions are available in the market from different companies with various capabilities, different price points and models, and deployment efforts required. Electronic OEMs should put high emphasis on reviewing available solutions closely to get the highest return on investment (ROI) possible.
Important aspects to consider include the reach of the solution, the types of data used by the solution as well as the source for the data and who collects, manages, and governs it. The analysis is as good as the data used for it and hence customers should aim to use the most complete and reliable data to realize the most benefit out of the analysis. OEMs should also consider speed issues including the real-time capability of the solution as well as its ability to take action automatically.
OEMs should also seriously evaluate vendors taking into consideration credentials in delivering tangible results to customers. Vendor experience has an impact on how innovative their solution is, and is an indicator of how it will evolve over time to provide customers more value.
Last but not the least, OEMs should also pay attention to the support, if any, provided by the vendor not only to reduce the time to customize the solution to their needs and time to results but also because usable insights require an iterative process between machines and data scientists with domain expertise.
