Trust levels indicate, as a percent, the level of confidence in the accuracy of cell data. Trust levels can be defined for base object or source record data. The Trust level is dynamically determined based on various factors, including:

The reliability of the source system providing the data.
The change history.
Data updates and refreshes.
The age of the data. Trust can decay over time as determined by the trust decay graph.
The validity of the data. Trust levels are downgraded by a predefined percentage for data that is considered invalid according to the validation rules.

If trust is not defined for a base object column, then the column contains the value from the system that most recently updated the base object.

You can manually override a calculated trust setting if you have direct knowledge that a particular value is correct. You can also enter a value directly into a record in a base object.