Problems and Challenges
Our story begins with a familiar conundrum shared by many companies in the semiconductor industry who have grown rapidly throughout the years: Our employees could potentially become a limiting factor to keep pace with such high-growth trajectory, because many employees are not trained in using data and data analytics to improve business performances and many are still holding on to conventional practices and processes.
We embarked on this journey with a strong belief that the way we collect, analyze and use data have huge impact on an individual’s and organization’s behavior. These behaviors will have a strong influence on the way teams and individuals collaborate and solve problems which will result in a significant improvement in the overall business performances.
Our journey began with a series of value stream mapping (VSM) workshops to review existing operations flow and processes to identify what is needed to “stop” and “start”. Our management team also played a vital role in these VSM workshops by prioritizing the business goals while the operations team aligned Key Performance Indicators (KPI)s as leading indicators for these goals. The teams focused on connecting the dots between process and decision making.
External Data Sources
Figure 1 illustrates the architecture that enables K&S to receive data from various global suppliers. Critical process parameters are systematically captured by suppliers and automatically sent to K&S’s servers at regular intervals.
In-House Data Sources
Data from our in-house processes are captured using an organically developed Manufacturing Execution System (MES). This MES system enables our line technicians to use portable tablets to pull test flow travelers and work instructions from the server, communicate directly with the machines to run automated test scripts. Test results and data from the machine are automatically uploaded from the machines into the servers in real-time. The MES ensures that the data captured are structured and “cleaned” before sending it to the server.
Developing Data Analytics Infrastructure
Various data sources are structured and stored in our centralized database. This database links the ERP system with data from internal processes and data from our global suppliers.
Developing Data-Driven Decision Making Mindset
Our managers and engineers use the data in the centralized database to create KPIs and link these KPIs to the business goals set by our management team. Off-the-shelf data visualization software such as Tableau is used to create user-friendly and real-time dashboards, allowing stakeholders from multiple functional teams to access and monitor data.
We held regular operations meetings to bring about managers, leads and engineers from various teams: customer quality, supplier quality, manufacturing engineering and test engineering to collaborate and solve problems or to explore new ideas. The data visualization software powered our KPI dashboards and these KPI dashboards are the focal point of the meetings. These KPI dashboards provide a concise overall operational picture.
Our quarterly goals setting exercises have also been transformed into a data-driven process. Figure 2 is an example of a “Waterfall” dashboard used to visualize all the failures captured by the production line in a quarter. Managers use this dashboard to set priorities for teams. Teams provide their updates and status directly on the dashboard. Such dashboards are reviewed regularly to identify showstoppers.
We have observed that employees are much more proactive in taking ownership and fostering closer collaboration among stakeholders in solving problems. We believe the data-driven culture has allowed our employees across functional teams to rely more heavily on data as the foundation for their collaborations and making decisions. Our line managers also have a much clearer and concise picture of the problems and are more confident in prioritizing and committing resources to fix the problems.
Data Monitoring and Anomaly Detection
We upgraded the database to a “Cloud” based data storage solution. This change addresses the limitations in data security, data latency and connectivity inherent from using the in-house server system. The upgraded “Cloud” solution (Data Warehousing) also includes the capability to carry out large scale automated data monitoring of critical parameters.
The move towards data monitoring came about to address the need for process owners to painstakingly spending a lot of time to meticulously comb through a huge amount of the critical parameters coming in from suppliers and in-house processes to detect out-of-control situations and trends. We invested heavily in “Anomaly Detection”. The data from suppliers and in-house processes are automatically sent to the “Cloud” server. A series of automated algorithms developed by our data engineers were inserted into the data streams thereby allowing the algorithm to detect anomalies, trends and out of specifications automatically.
Alerts will be automatically sent to the process owners and/or suppliers once anomalies are flagged by the algorithms. Figure 3 illustrates a series of anomalies detected by the algorithms – the anomalies were color coded for easy visualization. With this new upgrade, the system will trigger suppliers and process owners to act early and prevent defective materials from flowing into K&S. Similarly the anomaly detection methodology is also applied to our in-house processes to prevent defective products from flowing to our customers, thereby saving time, avoiding costly recalls and increase customer’s satisfaction.
Key Lessons Learned
Over time, we have observed that employees are much more proactive in taking ownership and fostering closer collaboration in solving problems. The new data-driven decision-making process and anomaly detection capability have also allowed K&S’s operations to free up resources dealing with repetitive tasks, and we have since channeled those freed up resources to focus on innovating new solutions.
ABOUT THE AUTHORS
TK Loh, Wint Kyi Phyu, Vivian Soong and Koh Dong Yang