Pavlenko Michael

Pavlenko Michael

Quote

The most important obstacle of mass distribution of BI — mistrust to yielded and to results. In addition distribution of BI is prevented by complexity of use of the BI tools and mediocre performance of the systems intended for independent work of users. Therefore many BI projects in Russia come down to creation of some standard set of reports. But it is impossible to call it an operational business intelligence in any way.

First, results in new tools differ from results which users received in own practices earlier. Secondly, the report long forms, but the user does not want to wait. He does not understand that the relational DB needs much more time for data preparation for the analysis, than for transaction updating. Such sluggishness guards the user, and he does not trust BI. Besides, data are unclear to the employees as they are not reflected in the BI system in a type, usual for them. It turns out that the company buys expensive BI systems, IT department implements them, and people do not use them. It badly affects distribution of technology of a business intelligence in the company.

First of all, the interface should be not only intuitive, but also simple in use: the less clicks, the better. Then data and communications between them should be clear. If the user understands everything and it is easy for it to work, then it likes the tool and it will apply it.

Further, it is necessary to provide high high-speed performance. Traditional relational DBMS are able to process quickly transactions, but are not adapted for processing of analytical requests at all. For this purpose special DBMS which release some companies are required. So, Sybase still in the mid-nineties developed the IQ system working with the SQL language, but inside organized not as relational DBMS at all.

The specialized data model for creation of storage in which two data types should be stored is necessary: detailed information, for example all orders for all clients in ten years, and the aggregated data necessary not to recalculate key indicators every time again. Process of aggregation is called Extract, Transform and Load (ETL) also are applied to it special tools. With their help, for example, it is possible to aggregate all orders on clients or income on a product, i.e. to prepare all key performance indicators. And when the report will be required, it is rather simple to print these indicators, to construct the diagram on the client or on a product and also to carry out deeper analysis, by consideration of the parts creating higher units (drill-down).

Business users shall participate in the BI project. In Russia often it is not observed though there are also exceptions. For example, in one of large private Russian banks the chief BI project manager is the first deputy chief director, everything becomes under requests of business users and result positive.

If business does not participate the project, by IT shniki do storage as find necessary, and a system which does not answer tasks of users turns out. As a result only IT work with this system.

It is necessary to use ready data models which offer several companies now. For example, Sybase offers Industry Warehouse Studio (IWF) — a set of physical data models for the different industries (banks, a telecom, retail, insurance companies, etc.). And a core at all models identical because all apply the same attributes — orders, deliveries, clients, employees, agreements, etc., and go the superstructures specific to each industry further. Such model can be unrolled on any DBMS including on relational. But (for example, Sybase IQ) everything will work at the specialized database much quicker.

Over this model it is possible to use any BI tool. Such approach considerably simplifies and accelerates creation of BI and also reduces the total ownership cost as it is not necessary for each new report to build separate data marts or an OLAP cubes. The detailed analysis (drill-down) is easily available because everything is connected among themselves.

Thus, for implementation of BI specialized DBMS, over it — a data model, and atop — the BI tool is necessary. In such construction there is nothing superfluous therefore its implementation will be inexpensive, and the system received as a result will quickly work.

• Why BI has no mass distribution in Russia?