Time is Money
The temporal dimension yields richer analyses and more insight.
Time dimension is key to any organization’s insight in an increasingly competitive and ever-changing business environment. Understanding how the business has evolved over time, as opposed to just delving into what it looks like today and what transactions occurred in the past, will lead to much better insight and smarter decisions.
A temporal database is a record of the past and present, and even a glimpse into the future that provides the strategic link to the time dimension.
With a temporal database, a business can track fluctuations and patterns in production or other activity over months or even years to economically understand the business at any point in time.
The data in a traditional enterprise data warehouse represents the business as it is. The data in a temporal data warehouse represents all aspects of the business, including how attributes of the business have changed over time and how the business has evolved—opening a whole new dimension to analysis.
A temporal database allows a business to examine a point in time, compare two points in time or perform time-series analysis. It also provides simple, quick support for audit trails and reporting assistance, eliminating the time-consuming need to reconstruct history to show why or when a particular decision was made.
With a temporal database, a business can track fluctuations and patterns in production or other activity over months or even years to understand the business at any point in time. For example, instead of just asking, “How many customers do I have now?” business users can ask, “How many customers did my records show were active as of December 31 when I calculated year-end bonuses last January 31?” In addition to better quality data, a temporal database provides the opportunity for much richer analysis.
Without temporal support in the database, it is extremely complex and expensive to maintain the necessary time-dependent historical data in IT processes and ask time-based questions. These complexities force compromises that generally prevent temporal analysis.
Who Needs It?
Most organizations can recognize the enormous value of maintaining history through time and supporting time-sensitive analysis in a data warehouse. Examples are virtually endless:
A temporal database provides value when the data or the business it represents changes over time, as is often the case in sales. A salesperson’s territory can change regularly—customers are added or taken away; territories are divided or expanded. If these changes are not considered, business views become distorted.
A sales manager, for instance, could appear to perform better this year than last year when, in fact, he simply gained salespeople or territories. Likewise, a salesperson’s numbers may be down this year not because he performed worse, but because his sales territory shrank. A temporal database understands how the territories changed over time and allows the organization to examine sales while adjusting for organizational changes.
Sales this year and last year can be calculated using the territory definition as of this year, or the definition in effect when the sales were made, thus isolating the salesperson’s true performance. While the blind comparison of absolute sales this year versus last year is still available, the organization now has the option to truly understand performance over time.
Common time scenarios in the insurance industry occur when policy changes are made and when events are processed out of order. As an example, a policyholder may incur an expense before, but file a claim after, a policy change is made. If policy terms are updated without maintaining how they evolved over time, the claim cannot be processed according to the appropriate terms.
In another case, when an employer waits until mid-month to supply a list of new employees covered as of the first of the month, knowing both when an individual was covered and when the coverage was reported can be critical to claims processing and accurate reporting.
A fully bitemporal database allows the insurance company to act based on the date the coverage started and understand the impact on reports of delayed notifications by group sponsors. A temporal database also lets insurers quickly and easily file regulatory reports for a given period, a task that is time consuming and complicated using a classical database approach.
In the manufacturing industry, analysis frequently must be done as of a point in time. Bills of material (BOMs) detailing the parts used in products change over time. A temporal database lets a manufacturer analyze quality issues and answer questions based on the BOMs for a certain date.
For instance, what is the failure pattern for all products built with screws from Vendor X versus the same products built with screws from Vendor Y? When a product comes in for repair, a manufacturer can easily determine product specifications for the date it was built and act accordingly.
Recategorizing products and product category hierarchy changes are common in the retail industry and are another illustration of when a temporal database can provide unprecedented value. For example, in the retail food industry, a product might move from the snack category to dessert.
Simply replacing category information loses this reclassification history and distorts analysis. Sales reports comparing this month to the same month last year will use current product categorization if the change in history is lost; however, the category sales for last year will not match previous reports for the same month, making analysis of business progress over time difficult. A temporal database adds the capability to compare sales using product category information as of any point in time.
Financial institutions build and continually modify complex forecasting models. A temporal database gives financial experts a before-and-after view by letting them compare model projections with reality. It also provides answers to any number of questions: Which capital-markets forecasting model would have worked better for the period of June 1 to Dec. 31, 2009, using the data acquired before the significant financial event of Sept. 15, 2009? How should our decisions change based on credit report and credit rating changes over time?
Sales reporting considering continual position changes is a key date-dependent, back-office operation in this industry. Many of the factors influencing sales performance and compensation change daily. Tracking and considering this history are essential.
TRAVEL, TRANSPORTATION AND HOSPITALITY
Businesses in these industries regularly analyze airline flights, ship-ments and hotel accommodations as part of their revenue-management models to adjust pricing and maximize revenue. With a temporal database, historical analysis can determine the best revenue strategy, such as how many airline seats were sold at various points in time leading up to departure.
A temporal database provides organizations with much more flexibility to truly understand the business and optimize operations by capturing and understanding historical changes. Time is indeed money. A temporal database can help you live up to this mantra not just by saving time, but by understanding it.
Why Teradata for Temporal?
Teradata Database 13.10 provides temporal data management and query processing, allowing the data warehouse environment to present data not only as it is, but also as it was and how it changed over time. With the temporal option, the database becomes time-aware and removes the complexities traditionally associated with keeping and using historical data.
When temporal data is updated, the database will automatically keep the previous values and remember when they were valid. Users can easily ask time-constrained queries without special knowledge of effective date fields and without writing complex constraint conditions in their SQL. The temporal semantics and optimizations built into the database allow IT to economically deliver new time-based analytical capability to the business.
The Teradata temporal option includes:
- The period data type with special temporal meaning to track date ranges
- Full bitemporal capability to manage and query valid time (when something was valid in the world being modeled) and transaction time (when it was entered into the database), enabling the full breadth of time-based questions
- Built-in logic to track historical data when updates are completed
- Upward compatibility for existing queries
- Simple temporal SQL syntax options to make temporal use easier and to add temporal logic to existing data warehouses and applications
- Several temporal optimizations to improve temporal update and query performance and maintain the best possible performance of nontemporal queries
- Special time-series analysis extensions