Start with one operational decision
Choose a decision someone actually makes: when to investigate a stop, how to reconcile a production counter, or which energy load to inspect. Record the current process, the person responsible and the evidence they need. A dashboard is a delivery format; the decision is the reason to build it.
Week one: map the boundaries
Walk through the line with operations and automation. List the controller, software version, protocol, network owner and available maintenance window. Separate read access from permission to change equipment. Define the pilot boundary as one machine or cell, and name who can stop the trial if production is affected.
Weeks two and three: prove the data path
Test a small set of tags against a known machine state. Check units, timestamps, counter resets and quality codes. Disconnect the test link deliberately during an approved window and observe recovery. Record missing and duplicate samples. Keep a dated evidence log so an apparent success can be reproduced.
Week four: accept, revise or stop
Review the original decision with the user of the result. Can they act with less manual work, and can the team explain every number? Accept the pilot only against agreed criteria. Write the remaining issues, owner, rollback steps and operating cost assumptions before expanding to another line.
Put it into practice
Use the related tool to check your assumptions, then bring the results to your project discussion.