Downside of Data
When there’s too much Big Data, the time to derive value and the cost for storing and processing Big Data can be significant, and companies are discovering that unstructured Big Data is difficult to work with.
For industrial IoT applications, we recommend a different approach:
- First, identify the desired business outcomes from the data
- Then, leverage the knowledge of domain experts to select the most likely data that drives these business outcomes
- Finally, match the appropriate data processing and analytics to processing data for business value
In this scenario, the role of domain experts in identifying and contextualizing data in industrial systems is as important as the role of data scientists in industrial IoT.
This is a simplified version of how Big Data becomes Smart Data.
Transforming Raw Data into Smart Data
In general, there are three levels of data processing:
- Device: Simple limit checking can provide useful insights into a device’s operation.
- System: Derive insights such as tension in a paper web that’s likely to result in a web break in a few cycles.
- Enterprise: Selected contextualized data, or Smart Data, from the OT environment can be analyzed and combined with data from the other parts of the enterprise to develop data mash ups or dashboards that deliver actionable insights.
Analytics and AI/machine learning at each of these levels can optimize processes and operations in industrial plants to deliver more productivity.
Big Data in industrial IoT applications will increasingly get replaced with Smart Data.
Industrial IoT solutions will leverage the scalable computing available at the device, system and enterprise levels to implement solutions that deliver business value such as descriptive, diagnostic, predictive and prescriptive analytics at the Edge and in the Cloud.
Therefore, it will become even more important for companies to use a combination of Smart Data, scalable analytics and Big Data for business outcomes.