Manufacturing analytics can provide powerful insight to help you solve problems faster, increase productivity, and drive innovation. But like any technology, there are best practices that can help you get the most value from it. Here are some manufacturing analytics dos and don’ts to help guide your implementation.
|DON’T spend money and time on company-managed infrastructure and applications. |
A DIY approach requires you to invest resources in design and development of a data environment and storage. And it can be costly to maintain and support 24/7, given the specialized on-staff resources required.
|DO look at vendor-managed infrastructure for lower total cost of ownership.|
Look for a manufacturing analytics platform that provides an end-to-end solution so you don’t have to design, purchase, manage, or scale the different layers of technology.
|DON’T try to boil the ocean (or the lake).|
Your manufacturing analytics solution should accept any data without oversight or governance. You don’t want analysts to have to start from scratch to find which data is relevant.
|DO narrow your focus to high ROI targets.|
Your manufacturing analytics solution should deploy quickly where the data is contained with as few separate sources as possible.
|DON’T just use the same storage mechanisms you’ve always had.|
SQL data models are inflexible and can’t be applied to many problems, and they can get slow when a large number of concurrent queries are made. Data historians only capture a small subset of relevant data and little, if any, analysis is performed on that data.
|DO be flexible and capture diverse forms of structured and unstructured data.|
Your manufacturing analytics platform should store, integrate, and analyze a wide variety of data, both structured and unstructured, including data types not typically found in ORDB systems. Make sure the solution will manage the data to avoid redundancy issues.
|DON’T just report on the data. |
EMI, which frequently doesn’t provide timely information and often only uses pre-set metrics, requires significant resources to capture, model, and apply analytical tools to data. Horizontal BI and data visualization software isn’t tailored for manufacturing. Users have to design all reports and data lacks semantic consistency and governed metadata, meaning you have to “wrangle” it to make sense of it.
|DO look beyond pie and bar charts with an eye toward predictive analytics.|
Your manufacturing analytics solution should use advanced analytics, such as machine learning and anomaly detection, to help discover previously unknown relationships.
|DON’T spend time and money to build your analytics application from scratch.|
You also have to maintain the entire pipeline from collection to processing to analysis to dissemination. Any bottlenecks in the data pipeline could paralyze the application.
|DO look at applications that already provide manufacturing-specific outcomes.|
Your platform should provide manufacturing-specific outcomes so you don’t need to divert resources from IT to manage and develop the big data applications.
|DON’T burn down the village just to put in a new solution.|
You shouldn’t have to stop production and tear out systems that are currently in place to use a new software tool. And you want to avoid vendor lock-in.
|DO use the data you are already generating.|
Your manufacturing analytics platform should sit on top of the assets that are already in place. It should be “low touch” with non-invasive implementation that doesn’t require downtime to deploy.
Interested in seeing a demo of a best-in-class manufacturing analytics solution? Contact us today.