Bitemporal: Point-in-Time Data

In an age where the importance of data needs no emphasis, it is no surprise that asset managers have begun to accelerate their collection of information from multiple sources in varying amounts. These sources can range from a file provided by a broker or fund admin to a market data provider. Although it rarely occurs, there are times when historical data can be incorrectly provided or maintained by the data source. In a unitemporal scenario, a user can rectify the historical record of data for the timeframe in which inaccurate data was provided or maintained. However, in doing so, the user erases the previous record that was once thought of as correct.

For example, there might be a scenario of complying to new regulations and audits where the user might want to preserve both records for future use.

Data Governance Problem

A conventional approach to addressing the problems posed by the unitemporality of data is to make multiple copies corresponding to the time when a historical change is introduced. Multiple copies of data introduces vulnerabilities in integrity and consistency as there will be increased overheads attributing to the need for more storage and a duplication of efforts in keeping data in sync.

However, one can maintain both sets of records, the past variant which was thought of as correct and the newly rectified copy of data, without the overhead mentioned above through bitemporality.

What is Bitemporal Data?

Bitemporal data denotes values of data corresponding to two dimensions of time: knowledge date and effective date. While knowledge date is the date on which the data is entered into the application, effective date is the date for which the data is being maintained within the application itself.

Bitemporal Data: Elucidation with an Example

The below diagram represents the end-of-month price of a commodity over a January to November timeframe. As you can see, the user has made changes to the corresponding data for the months of March, April, May and June. In the month of November, they realized that the data was erroneous. From here, the user can view the data corresponding to the 31^st of March, also known as the effective date, as the user thought it was true on the 31^st of March, also known as knowledge date, corresponding to $185.16. With the help of bitemporality, the user can also view the rectified price of $188.16 for the 31^st of March date by choosing to view the data as is on the 30th of November, the new knowledge date, without affecting the storage of the previous value of $185.16.

As depicted in the diagram below, users can traverse and view the data based on both the knowledge date and effective date.