Summarization Levels
What are Summarization Levels?
Typically in an executive information system, reports with aggregated data are required. In these reports the end user navigates from a highly aggregated level down to the level of detail which is of business interest to him. The selected report data is contained in the main memory and can sometimes lead to a large volume of data and long response times. There are various alternatives for improving the response times of a report. See
Optimizing Response Times In this section the alternative using summarization levels is described, and the guidelines according to which your summarization levels are to be defined are also explained. You can find out how to create and build up summarization levels in Customizing.A summarization level saves transaction data of an aspect in additional compressed forms. In such transaction data individual objects are described using a range of characteristics. Summarizing means that selected characteristics are left out. Objects from the original dataset, which are no longer differentiated once characteristics have been left out, are summarized in one object in the summarization level. The summarization allows for quick access to the data in reporting.
These simple example shows how the transaction data records can be summarized on the database in order to improve the reading times in the report.
Employee xxx in cost center 100 in company code 1000 worked 50 hours overtime.
Employee yyy in cost center 100 in company code 1000 worked 30 hours overtime.
Employee zzz in cost center 200 in company code 1000 worked 20 hours overtime.
If the transaction data were transferred in this way then three records would be written in the transaction database table:
Employee |
Cost center |
Company code |
Overtime |
xxx |
100 |
1000 |
50 |
yyy |
100 |
1000 |
30 |
zzz |
200 |
1000 |
20 |
As a rule relatively few analysis paths are used in reporting, so you can define a summarization level for the example above.
A summarization level is formed at the cost center level, meaning a summarization is made over the characteristic employee. The records are compressed together and the following structure ensues:
Cost center |
Company code |
Overtime |
100 |
1000 |
80 |
200 |
1000 |
20 |
In this case, you could set the summarization level even higher, by forming it at the company code level.
Company code |
Overtime |
1000 |
100 |
The more characteristics that are summarized, the higher the level of compression of the data and the more limited the use of the level. You must decide between the shortened response time because of the level of summarization and the possible use of the level in reporting. Highly summarized data will not be of much use to you, if it can only be used in a very few reports.
At the definition stage you have the following options.:
The key figure values are summarized above a characteristic.
Afterwards you can no longer evaluate the characteristic in the report.
The key figure values cannot be summarized over the characteristic.
Afterwards you can evaluate whatever characteristic values you like in the report.
The key figure values are only retained for a certain characteristic value.
Afterwards you can only evaluate this characteristic value in the report.
Depending on how you have specified a characteristic in your report, you decide whether a characteristic is to be summarized, given a fixed value, or left free. You define the summarization levels in Customizing. To enable the summarization levels to be used by reports, the levels have to be supplied with data. To do this, you schedule a job in the background processing to build up the summarization levels. Only then can the levels be used by reports. After the levels have been built up, the data is always updated at the same time as the aspect is supplied from the transaction database. In this way, the summarized data is as up to date as the transaction data. You can find more information about this in Customizing.
When executing a report, the system checks whether there is saved data available for the report. If data has already been saved it will be displayed. If no data has been saved for the report but a suitable summarization level exists, then the data from the summarization level will be called up. If there is neither saved data nor a suitable summarization level, the data is selected as new.
When do I use Summarization Levels?
If you are not satisfied with the response times of the reports in your Executive Information System you should consider defining some summarization levels.
However, before you decide on summarization levels you should find out about the other ways of improving response times. You can read about these under
Optimizing Response TimesGuidelines for Creating a Summarization Level
When defining a summarization level there are often several possible combinations of characteristics. Since not every summarization level can produce an improvement in response times compared to a report executed online, there are guidelines for defining summarization levels. To attain the optimum summarization effect you must analyze the desired report for each summarization level, determine the affected master and transaction data and consider the following guidelines:
Estimate the number of records to be read on the database, and the number of records in your planned summarization level. A minimum factor of 10 should be attained.
Where possible, summarization levels should be defined above characteristics with many values. You can attain a greater summarization effect if you summarize above the characteristics product and customer as these characteristics typically contain many values.
If your summarization level contains many records, the corresponding report must be defined in such a way that it does not have to read more than 10,000 to 20,000 records from the summarization level.
Since the summarization levels are updated automatically, each change to the transaction data leads to an update in all the summarization levels. This means that a change to one transaction data record leads to several writing processes in the database. If there are many summarization levels, performance in the data transfer is impaired.
If you generate a proposal for a summarization level you should edit it manually. You should also check whether a similar summarization level already exists You will find this information when maintaining the summarization levels in the overview and detail screen. You should check this data for all summarization levels occasionally, as the number of accesses, for example, can help you to decide whether a particular level is still required. You can find more information about this in Customizing.
In many cases there is a dependency between several characteristics. If, for example, the product group is derived from the product number, then the product group is dependent on the product. The volume of data in the summarization level is not enlarged by the inclusion of the product group and therefore does not impair performance. The advantage is that the summarization level can also be used by reports which contain these dependent characteristics.
The dependency between two characteristics normally exists because of derivation (see
A detailed description of how you create and build up summarization levels is given in Customizing.
For an example of defining a summarization level see Example: How can I Improve the Response
Times of my Report?.