Implementing Self-Service BI
Promoting democratic data analysis
The need for business departments to source, consolidate, and analyze data independently of IT did not just emerge with the trend toward self-service BI. Ever since business intelligence was first introduced in companies, the demand for self-determined data analysis has been growing louder. Integrating this into a company can be successful with just a few resources.
With the introduction of centralized BI solutions and data warehouses, the original data sovereignty of business departments shifted to IT and controlling departments. From then on, reports were created centrally and could only be adapted to user requirements through cumbersome request processes.
Gradually, this perception of BI in companies has shifted once again. Today, more and more users are taking advantage of modern self-service BI tools like Tableau or Power BI to analyze data and create reports independently. However, data provisioning and modeling usually remain the responsibility of the IT department. This limits the analytical freedom of experienced users. This division of labor neither aligns with the self-image of many business users nor with the requirements for agile BI.
Requirements for modern BI
Self-service BI is intended to enable business users to create their own analyses and reports and to share them with other users as easily as possible, largely without relying on IT. In addition, they should be able to connect further data sources to the provided data model as needed. In this context, IT ideally takes on the role of an internal service provider. It provides the architecture and software environment, as well as an initial data model in coordination with the business department.
Users need intuitive software for self-service BI. Furthermore, it must provide a data foundation that is prepared in such a way that the structure is understandable and the data is up to date. The growing needs of users in the context of digital transformation are increasingly presenting IT departments with new challenges. Business departments often need to expand their data basis at short notice—perhaps only once—to be able to react to changing market developments themselves. However, IT departments have standardized processes and concerns regarding data protection and performance. These can often slow down users' desire for independent data analysis. Self-service BI does not mean that all users can access all data at will. Internal and external data protection requirements must still be guaranteed. A suitable authorization concept ensures this.
For IT, the task is to introduce suitable architectures and processes. These must be able to satisfy the demands of business users while also safeguarding the legitimate interests of IT. Therefore, new forms of collaboration between IT and business departments are increasingly necessary, which can, for example, be based on agile principles.
Standard reporting meets analytical freedom
On one hand, the increasing adoption of self-service BI in companies is expanding the possibilities for conducting ad-hoc analyses. On the other hand, there remains a need for standardized, reliable reporting that is provided on a regular basis. This reporting should meet fundamental requirements and serve as the foundation for internal communication. To further develop this standard, an experimental system is required. This environment should allow business users to test and refine innovative solutions within the scope of self-service BI. However, this does not mean that a shadow BI should be built in opposition to IT within a bimodal BI framework. Rather, it calls for IT to provide a sufficiently broad framework in which the benefits of self-service BI can fully unfold.
Business intelligence must consider the target audience
The target audiences for BI and self-service BI in companies can generally be categorized into three user types, whose requirements for autonomy in data acquisition and analysis differ fundamentally.
Management primarily wants a quick overview of key figures and KPIs provided in reports and dashboards. Due to often limited IT affinity, the expectation for the data foundation is that it must be valid and performant enough to meet their information needs.
Business departments require detailed information about business processes and want to be able to view and analyze this from as many different perspectives as possible. Self-service BI and data discovery are essential requirements that the data foundation must meet. Business users themselves should be able to correctly link data so that the resulting analyses contain accurate information. The data foundation provided by IT can support this with a homogeneous nomenclature. Increasingly, modern self-service BI tools also offer features to provide meaningful support to users in this task.
Data scientists generally possess strong skills in data linking and data modeling. The expectation of this user group is maximum autonomy and an attribute-rich data model whose entities they can link according to their specific needs. Large volumes of data are often retrieved within complex queries. Therefore, it is worth considering providing a separate sandbox for these users—a "data lab"—that is decoupled from the actual BI system.
Infrastructure as a success factor
When selecting the appropriate infrastructure for the increased adoption of self-service BI, several aspects should be taken into account.
First, it must be clarified which user group is the focus and the level of detail they require to access data to meet their analysis needs. Is modeling according to Data Vault appropriate? Or are classic star or snowflake schemas sufficient? When making this decision, it is advisable to think ahead and also consider new technological developments.
Data growth and the required timeliness of data delivery are other important factors influencing the choice of suitable infrastructure for self-service BI. Should data be available in real-time or near real-time? In these cases, it may be beneficial to virtualize data rather than providing it persistently. Technologies for implementing a logical data warehouse offer corresponding capabilities and often already include a multitude of connectors.
Data model design
When designing the data model, the primary focus should be on consistent nomenclature to support users in creating analyses and reports. It is also advisable to document the definitions of metrics, dimensions, and calculations in a business intelligence wiki accessible to all users. This promotes a shared understanding and prevents misinterpretation. If users need to enrich data with additional attributes, the data model should include the natural keys required to link the data with internal or external sources.
When selecting a database system, business users should consider query performance requirements alongside expectations for data volume and growth. Based on these considerations, in-memory or column-oriented databases may be suitable options. If a separate data lab is being established, it is also important to ensure it is assigned its own hardware or at least a dedicated instance so that data scientists do not impact the performance of the BI system.
Selecting a self-service BI solution
Several criteria should be considered when choosing suitable software for self-service BI. How well does the application fit into the existing infrastructure, and what is the maintenance burden for the IT department? It is also worth questioning to what extent the application can actually relieve IT by enabling user independence. Key factors include how intuitively the application can be operated, which data sources can be connected directly, and how well the preferred solution supports users in linking and preparing data.
If the software fails to gain widespread user acceptance or fails to encourage consistent use through simplicity and speed, self-service BI initiatives can quickly lose momentum and impact. In this context, it is also advantageous if the software supports team collaboration and the easy sharing of analyses and reports. Self-service BI only reaches its full potential through team interaction and discussion.
A suitable self-service BI solution must at least meet the following requirements:
- Intuitive and rapid creation of analyses and visualizations
- Connection to a wide range of data sources without IT support
- Compatibility with IT infrastructure and departmental requirements
Ultimately, the decision for a specific self-service BI solution should always be aligned with the choice of an appropriate overall architecture to ensure the long-term success of the company's entire BI strategy.
