I find it interesting comment and transcribe a recent discussion on a blog of specialists in QlikView. The topic was why it is necessary to model and build a data warehouse and not just build the whole application to support decision making using only the tool QlikView, as promoted marketing of the company.
The truth is that most companies are not ideal conditions for data to be used directly by analytical applications. It is common to see quality problems, coding and / or classification of products or customers that are not consistent between different systems, the need for data transformations, etc..
For those wishing to review the original discussion can see this link . Then mention the most relevant:
1 .- QlikView marketing people seem to always forget that to achieve such a beautiful way to display and analysis REQUIRED datawarehouse design a solid and well built . The data, mostly, are not available for the ideal way to show the demos.
2 .- You can build tightly integrated with QlikView applications but when it comes to metadata, maintenance, integration from multiple sources, data lineage, history management, falls short. It's a great tool but should be used as a front-end and not as a replacement data warehouse.
3 .- The data should be stored in a vendor-neutral format and so openly as possible.
4 .- The data must be available to be worked by other tools with features that users require.
5 .- QV Excel replicated the effect. Proliferenado data in various containers
6 .- create multiple sources of truth, something to be avoided in a successful implementation. Data Marts also end up generating not integrated with each other favoring the multiple truths.
7 .- should be considered extreme about privacy regulations in governments that are crying datamanagement sophisticated architectures and accompanying data and BI solution.
8 .- lack of control and metadata management. A pillar on which any management architecture must be built.
9 .- In reality many, especially to large companies NEVER allow direct access to transactional systems which is fine as is necessary to construct a datastore and refresh periodically. To the extent that increase the data will become a challenge to the architecture of the repository.
QlikView The truth is that, regardless of its innovative technology and it is not just a display but also has a powerful engine that correlates information in-memory "is just another weapon in the arsenal of tools that can be used to exploit the data contained in corporate data warehouse.
is interesting that the marketing message has been so attractive to end users and many IT managers. The message is not complicated to design and build their data marts, data warehouses and cubes. Do not waste valuable time on these activities and empowers end users and build their applications directly connected to their data. Obviously there is still a debt to both end users ease of use of BI tools and self-service capabilities to obtain results in lower implementation times and greater satisfaction.
do not agree with it more time consuming to make a cube that an in-memory db for QlikView. An example is Cognos PowerPlay from their old versions. Enough to indicate whether the source data in flat files or tables in a DB and the wizard of designing Transformer module dimensions, hierarchies, and with another click the cube was loaded and visualized. Instant cubes faster than proprietary languages \u200b\u200bQlikView and without extraction!.
users want to build their own analysis and have the ease of integrating their external data. Do not want to wait months for their analysis are available. These aspirations are trying to fill the well with PowerPivot companies like Microsoft and MS SQL Server 2008 R2 which blends self-service end users with managing IT professionals.
decision systems are not just hubs such as Cognos PowerPlay or databases in-memory as QlikView. We need clean data, standardized, accessible open formats by different software tools. We also need the "single source of truth" which can not be all contained in PowerPlay cube one or QlikView database.
0 comments:
Post a Comment