Understanding the PI System 'pctgood' Function: How Is It Calculated?

The OSIsoft PI System offers a powerful suite of analytical tools to assess data quality and availability. Among these, the pctgood function is frequently used—but often misunderstood. In this post, we'll unpack how pctgood is calculated and what it tells you about your PI Points, particularly when you have partial or intermittent data coverage.

What Does pctgood Measure?

The pctgood (percent good) function computes the percentage of time within a given time range that a PI Point records 'good' data. This metric is crucial for understanding not just data value accuracy, but the reliability and coverage of your data acquisition process.

Key Concepts: Good, Bad, and Missing Data

To understand pctgood, it's helpful to differentiate between:

• Good Data: Data values with a 'good' status (valid readings, as per PI definitions).
• Bad Data: Values marked explicitly as 'bad'—for instance, statuses like I/O Timeout, Error, Comm Fail, etc.
• No Data: Intervals where no data is stored or reported at all—for example, if the PI Interface was down and no samples were recorded.

How Does pctgood Handle Missing Data?

A common point of confusion is how pctgood treats time periods where there is no data at all. Let's clarify with an example:

Scenario: You have a PI Point 'A' that collected data only from 01:00 to 13:00 on June 26th (12 hours), and was silent for the remaining 12 hours.

Question: If you calculate pctgood('A', '26-Jun-2012 00:00:00', '27-Jun-2012 00:00:00'), will it return 100% or 50%?

The Answer: pctgood Considers Time with No Data as 'Not Good'

pctgood calculates its percentage based on the entire time window you specify. If there is no data for part of the period, that time is not counted as 'good'—unless the point is marked as "stepped", in which case the last known value might be extended. For most continuous analog points:

• All values during the 12-hour stretch are good
• No values for the other 12 hours
• pctgood = (12 hours with good data) / (24 hour time range) = 50%

You will only see pctgood as 100% if there is good data throughout the entire range.

Bad statuses (e.g., No Data, Error) are explicitly counted as 'not good,' reducing the percentage further. Time periods with no recorded data—without a good status—are also not counted towards 'good' time.

References and Further Learning

• YouTube: Calculated Values and the pctgood Option (discussion at the 3-minute mark)
• Official OSIsoft PI System Documentation - Calculated Data Functions

Conclusion

Understanding the nuances of pctgood is essential to monitoring data quality in your PI System. Remember, it's not just about the absence of errors or explicit bad statuses—any gap in data is reflected in the percentage. Use this knowledge to better design your instrumentation strategy and to more accurately interpret the health of your data streams.