Implementing a self-service BI strategy does not mean ripping and replacing your traditional BI infrastructure. It can be achieved by complementing your existing BI platform with the addition of new semantic layers of data models and end-user data discovery tools. Power BI and Azure can provide these additional capabilities more quickly and cost-effectively than other self-service BI frameworks.
The purpose of implementing self-service BI is to make it quicker and easier to provide your employees with the data they need, when they need it, and in a form that makes sense for them. This is not a new concept and is not particularly different from the definitions of traditional BI. What has changed over the last few years has been a shift in focus towards meeting the individual data needs of business users, whether they are data scientists, experienced data analysts, or non-technical users, by giving them more control over how they access and consume data.
In their 2011 paper, Claudia Imhoff and Colin White defined “Self-Service Business Intelligence” as…
"The facilities within the BI environment that enable BI users to become more self-reliant and less dependent on the IT organization. These facilities focus on four main objectives: easier access to source data for reporting and analysis, easier and improved support for data analysis features, faster deployment options such as appliances and cloud computing, and simpler, customizable, and collaborative end-user interfaces."
This definition serves as a useful discussion point to highlight how Power BI can help your organisation to achieve the objectives of self-service BI and realise its benefits.
Power BI provides the perfect platform for agile BI because it empowers authorised users to rapidly set up new data connections and create new analysis reports. This is possible because Power BI is tightly integrated with Microsoft’s other BI products and Excel, as well as having built-in connectors to external data sources.
Create a thin semantic model over existing data
Self-service BI solutions depend on data sources that are in a form that is suitable for analysis. The degree to which any data source needs to be transformed will depend on its source and format, the quality of its data, and its consistency with other data sources. This might require a single, integrated, enterprise data warehouse, a ‘logical’ data warehouse over different types of data sources, or a thin semantic layer consisting of data models.
From a business user’s perspective, the primary function of a data warehouse, or any other analytics platform, is to expose a semantic layer (or model) over the source data. This isolates the user from the complexities of the underlying data and presents it in a format that is meaningful and suitable for further analysis. It also ensures the validity and integrity of the data, and prevents ‘multiple versions of the truth’.
Power BI exposes data models as datasets which can connect to a variety of data sources. In addition to its built-in connectors to SaaS applications, Power BI has direct support for Microsoft’s tabular model which provides the best technology for building a thin BI semantic layer over your data.
Introduce an agile process to build new data models
It is the time it takes an organisation to create the required BI sematic layers that determines how agile or responsive it is. If you want to introduce an agile self-service BI process, the most effective approach is to develop your semantic layers using a combination of tabular models in SQL Server Analysis Services (SSAS) or Excel, and Power Query queries. Tabular models are in-memory databases that can source and enrich data from any relational or multidimensional data source, without the overhead and complexity of designing multidimensional databases with cubes, dimensions, and measures. Power Query queries extend the scope of these data sources further to include Azure data services, OData feeds, web pages, web services, and SaaS applications.
In short, the most agile process you can implement for designing, building, and deploying self-service BI solutions is to create Power BI reports from datasets connect to tabular models.
Power BI has tightly integrated support for tabular models, Power Query queries and third party data sources. Collectively these provide access to all the data sources your users will need and covers relational, multidimensional, structured, or semi-structured data from on-premises, cloud, or web data sources.
Connect directly to source data whenever possible
The growth in demand for data from non-traditional sources has challenged the concept of an enterprise data warehouse as the single analytics platform. It does not make sense to conform and integrate OLTP data with real-time streaming data, semi-structured and unstructured data, or third-party data. From a business user’s perspective, the primary function of a data warehouse, or any other analytics platform, is to expose a semantic layer over the source data. New types of data stores, such as Hadoop or NoSQL databases allow data to be modelled and incorporated into BI solutions without being imported into a data warehouse. Data that does not logically or easily fit in an OLAP database should be kept in a separate data repository.
Use tabular models to integrate different data sources
Extending the scope of data sources is fundamental to the concept of self-service BI, so even organisations with established BI systems will be faced with the challenge of integrating these new and diverse data sources. As discussed in the previous section, the easiest way to consolidate your diverse data sources is to develop tabular models and Power Query queries. When data is stored in a tabular model then the underlying queries to the original source data are maintained so data refreshes can be scheduled. If you are developing tabular models using SQL Server Analysis Services (SSAS) then the server will need to be sized appropriately, particularly for in-memory models. I would strongly recommend provisioning Azure virtual machines to host instances of SSAS for self-service BI to take advantage of the benefits of the Azure’s scalability, high availability, and ease of deployment.
Power BI reports and dashboards provide a wealth of features for exploring and visualising data to allow a business user to consume data on their own terms. This allows each individual user to personalise their experience by choosing which data insights are relevant to them, and to analyse them in the format and on the device of their choice.
Use visualisations that make the data more comprehensible
Analysis data is immediately accessible to each Power BI user through a web portal or mobile app. You can create your own reports from datasets or other reports, customise your dashboards by combining different reports, and interact with reports to change the view or filter the data. Even when a data visualisation tool provides a rich set of charts and other controls, it can be difficult to choose the most appropriate one. When you interact with Reports or Q&A in Power BI, it will automatically select the best visualisation for you.
Share data insights
With self-service BI, we can make a distinction between sharing data and sharing data insights or discoveries. Corporate BI is responsible for sharing data to make sure that people who need information to make decisions have access to it. This data can be distributed in different ways but it will have a predefined design and format. Self-service BI takes this further and allows data insights uncovered by individuals, who have explored the data, to collaborate with colleagues by sharing the information that is most relevant to them.
Data insights in Power BI can be shared with other team members, either by sharing your personalised dashboards with nominated Power BI users, or by taking snapshots of individual reports and emailing them from within the Power BI apps. The significance of data insights can be emphasised with report snapshots by annotating them before sharing or setting up data driven alerts.
Power BI tools can significantly reduce the learning curve for new users because report and dashboards have a simple and consistent interface that will already be familiar to many Excel users of Power View or Pivot Tables. Alternatively, experienced users, who are looking for more powerful analysis features, have the capability to build sophisticated queries and data models with Power BI Designer, or Excel’s Power Query and Power Pivot.
Explore data by interacting with reports
When users interacts with a report in Power BI, they do not need to understand the relationships between the data fields because these have all been defined within the data model. Provided that the Power BI dataset has been correctly modelled, then the end user can just drag and drop any data fields into their report, confident in the knowledge that the relationships between the data items will be correct. This is the starting point for exploring the data and discovering new insights. These interactions include selecting data fields, choosing visualisations, filtering and highlighting data, merging visualisations, and building dashboards from saved reports.
To simplify the experience even further, users can query the data and build visualisations from the results by using the natural language capability and just typing questions.
Enrich data models with key performance indicators (KPIs) and calculations
A user can just drag data fields onto a visualisation within Power BI and it will intelligently interpret the data and decide how best to represent it. This is possible because Power BI understands how fields are related, what type of data they contain, and whether they represent aggregations, calculations, unique values, geography fields, or hierarchies. This metadata is derived from the underlying data model, which can be enriched with calculations, measures, KPIs, and perspectives (subsets of the source data), all defined using a sophisticated formula language called Data Analysis Expressions (DAX).