Challenges related to the implementation of Business Intelligence

In publications, Business Intelligence is more and more often replaced with the term analyst.  This is because today’s systems can do much more than just deliver business information. Analytics is a broader concept, it means, inter alia, processing data from multiple sources and implementing artificial intelligence machine models. What does the implementation of such a conceptually extended Business Intelligence look like? How should you prepare for it? What risks must be taken into account when implementing BI? We answer these questions in the article.

What is the structure of the BI platform?

The beginning of the data analysis process is obtaining information from data sources. Then the data is cleaned (verified) and integrated. In the next step, it is moved to the data storage (i.a. data lake, data warehouse, etc.). Only now is it possible to use it further in analytical models translating data into the language of business.

At this point, it is possible to perform an intermediate step, i.e. machine learning, in which artificial intelligence algorithms are used. These activities may introduce modifications to previously created models. At the end, we get ready-made reports or data to be used in self-service tools that give users the freedom to conduct analyzes.

There are no wonderful BI tools to solve all your business problems

There are many companies on the market with aggressive marketing declaring that they will provide us with BI tools that will be able to solve all our problems related to data analysis. Unfortunately, such tools do not exist. Firstly, because they do not have direct access to data sources or operational databases, and each time they need to be adapted to the existing systems. Secondly, because the data in the sources is usually stored in an application structure that is incomprehensible to business. Thirdly, such tools do not exist, because it would require their creator to know the data structure of every possible source, in every system.

Each time, the adaptation of ready-made analytical tools requires proper data modeling and creating the full path that the data travels: from its appearance in the storage to the self-service or reporting tool.

To sum up: introducing BI requires, first of all, understanding and often modifying business processes in the company, which cannot be done by any wonderful tool “straight out of the box”.

Assumptions and strategy of BI implementation

It is necessary to develop assumptions before implementing a BI solution. When designing a solution, you must take care of the infrastructure and adapt to the business processes that are responsible for data creating. This approach is the key to implementing effective analytics in the BI tool and consistent data obtaining. It should be remembered that the implementation of BI is not only about introducing new layers of information processing. An approach that takes into account business logic is necessary, therefore the implementation of BI requires adaptation to business procedures, and often also modification of these procedures.

Another aspect that should be kept in mind is the data management strategy, i.e. Data Governance. Many of the company’s systems provide information only about the product and we need to combine it with data about their organizational structure. Data must be managed in a way that ensures the consistency of all areas in an enterprise (Master Data).

It is also necessary to ensure data security (Data Security). In this area, it is particularly important to define the rules of access to information. This access can be defined in an object-oriented manner, designating groups of employees who have access to a specific range of data: reports and models.

Another way can be vertical access through attributes. It sets the access limits based on e.g. customer attributes. A separate group of employees will have access to information about the customer’s data and address, and another group to information about their purchases and services they use.

Access to data can also be defined horizontally, based on the records contained in the database. In this case, one group of employees may have access to information about customers from Poland, and the other one, for example, about customers from Germany.

It is very important to define your data access policy in detail, but care should also be taken that it can be managed and changed as needed.

When designing a BI system, data quality should also be taken into account. By establishing standards and processes ensuring high data quality in the analytical platform, analysts can be sure that the obtained data is correct and that it can be used to make the right business decisions. With this in mind, one should take into account, among other, data life cycle, versioning and data overwriting. The acceptance of new tools by the analysts themselves is important. If the analysts do not have confidence in the analytical system, they will not use it or even avoid participating in the project, which may ultimately cause the implementation failure.

BI implementation example

We will take a closer look at a case study of BI implementation in pharmacy.  A company in this industry was struggling with the problem of a large amount of distributed data sets. Finding answers to business questions was very difficult because the data was not well integrated with each other. To facilitate access to information, it was decided to build a central warehouse based on Microsoft cloud solutions using Azure Synaps Analytics tools and the Power BI platform.

The implementation of this task was undertaken by Inetum. Inetum has designed a new architecture based solely on cloud services, it developed security rules applicable in this architecture and rules for data access (Data Governance).

An analysis and reporting service fully functional, moderated self-service for business users, enabling them to visualize and report on their own was implemented.  Thanks to the created solution, users can share their analyzes with other community members within teams via Microsoft Teams, using the capabilities of the Power BI platform. Access to the platform from the level of a browser and mobile devices, as well as a dedicated reporting and analytical space for sales representatives were provided.

The biggest challenge during the implementation was the diverse data structures integration. In countries where customers had branches, business processes were implemented differently from each other. Among other things, the goals of sales representatives were defined differently. In some countries, the quality of a sales representative’s work was determined by the number of meetings, while in another country it could be determined by the number of presentations. The customer classification systems and billing periods were also different. First of all, it was necessary to make this information uniform, so that the process of comparing and analyzing individual areas would reflect the real picture of the implementation of sales plans, and not only their very inaccurate estimation. As a result of the implementation, the company obtained a valuable Business Intelligence tool that allows them to track the state of business in the entire geographically vast organization in near real time.

This task was completed following the One BI philosophy. By implementing this concept, Inetum helps customers to build a uniform space for data analysis, and also helps to improve business processes. The tools, cloud and local solutions used by Inetum provide users with access to analyzes and reports from a computer, tablet or phone.

Marek Czachorowski

Head of Business Intelligence Practice at Inetum in Poland. For 10 years Marek has been involved in BI and broadly defined data analysis and processing. Since the beginning he has been mainly associated with Microsoft solutions and tools. Since 2017, a certified specialist in the area of data warehouse design and SQL Server platform management. He is currently developing primarily in the area of cloud analytics. As a consultant, he helps clients define company processes, establish rules for processing and access to data.