Great article outlining issues to be considered when performing sensor data analytics: The awkward conversation about sensors and data that we need to hear
In summary, issues outlined in original articles are:
- Sensors sometimes lie: Values start drifting over period of time or because of operating conditions
- Not the whole truth: Remote processing units or edge analytics can filter out the relevant sensor information.
- It can take work to get a signal: Sometimes modeling the sensor behavior may take long time.
- Scoring: Scoring or model execution may cause problem as updating the model at remote locations is problematic.
- Sensor data can be useless on its own: Businesses need to combine sensor data with other operational data to make it meaningful.