Analyze

Question and objective

How do you go from identifying an anomaly to arriving at a verifiable answer? Choose the type of analysis based on the task at hand: targeted review, individual evaluation, or continuous data processing.

Implementation in Bytefabrik

Problem	Appropriate function
Filter and Aggregate Stored Events	Queries and Charts
Explore a specific question using code and visualizations	AI Notebooks
Apply a rule to new events on an ongoing basis	AI Pipelines

AI Notebooks and AI Pipelines are Bytefabrik extensions in the IoT Data Hub. They speed up the creation of analyses and processing code, respectively. Technical review remains an integral part of the workflow.

Best Practices and Tips

First, specify the subject, time period, metric, and comparison. “Compare the cycle times for Product A before and after the adjustment” is more testable than “Analyze production.”

Start with a limited amount of data, monitor interim results, and keep track of the filters you use. A reproducible answer consists of the result and the data source. If you need to run this analysis regularly, decide which option—a chart, a reusable notebook, or a running pipeline—requires the least maintenance.