Our Blog

where we write about the things we love



The Reality of Universal Self-Service BI

For years, the pinnacle of BI for many has been dictated as self-service BI. I have read about it. I have heard presentations on it. I have given presentations on it and promoted it. But no matter how often it is discussed, promoted or advertised, one question remains – what is it?

Business Intelligence

If you interview 20 people across all aspects of an organisation about what they would consider self-service BI to be, you will likely get at least 10 different answers – all of which are perfectly valid!

There is a second aspect of self-service BI which is rarely discussed: the unintended consequences. In an article by Wayne Eckerson he comments on the results of a self-service BI survey. Among items noted from an end-user perspective were “Self-service BI ‘requires more training that expected’”, it “creates report chaos” and “makes it harder to find the right report”. 

End-users were not the only people who noted problems with self-service BI; IT and others also noted issues including “poor query performance, lack of adequate access control, an explosion of nonstandard BI tools and increased licensing costs.” 

For all the statements about how self-service BI is great for an organisation, few talk about the potential pitfalls in having self-service BI. The chance of those pitfalls impacting your self-service BI effort can be reduced by matching your self-service model(s) to the appropriate audience. Let me explain.

As a BI consultant, I have been asked many times for self-service BI, sometimes referred to as the ability to do ad-hoc BI. When I was younger and had all the answers – I said “Yes! We can do that!” 

This very quickly changed. Now my response to a request for self-service BI would be to ask them to describe to me what self-service BI looked like. Often I would also ask a series of questions such as “Who will be using it and what is their comfort level with databases?” 

While there are many permutations of what people envision as self-service BI, I group those visions into five broad categories:

  • Parameter driven reporting
  • Guided self-service
  • Insulated self-service
  • Full self-service
  • Visual self-service

My goal is not to state that some of them are better options than others but to point out that good self-service BI is dependent upon the audience. What is good for one audience may not be good for another. When there is an audience / self-service model mismatch, you will experience considerably more comments in regard to self-service BI similar to those referred to above.  In short, just because you have developed a great self-service tool for one group does not mean it is a good tool for another group.

Different people have different needs; line managers are very different from executives or analysts, for example. These broad categories can help in matching the correct self-service structure to the right audience. I’ll discuss each of the categories so that you get a better understanding of how each is used so model/audience matching can be done.


Parameter driven reporting 

Parameter driven reporting is exactly as the name implies: it is fixed format reporting whose output could be driven by the parameters provided. For people who have been using the exact same version of a report for years, being able to specify parameters and have a version of the report which meets specific criteria was revolutionary. 

Now, if you are like I was when I was younger (a know-it-all BI snob), you may not think this qualifies as “BI” at all, self-service or not. However, taking a broader view that the goal of BI is to “get the right information to the right people at the right time” means that this clearly fits within the BI realm: it is information delivery and making sure that information is the right information.

This type of self-service is often restricted to those with well-defined job roles – in one case it was field service workers being able to get a list of customers with a specific type of problem rather than a list of customers with any problem. It allowed them to run a report without having to wade through extraneous information to find the issues they needed to address.


Guided self-service 

Guided self-service provides a data framework for the user and allows numerous options for slicing, filtering, changing aspects of the view, sort orders, columns, and so forth. This type of self-service is highly interactive but, as the name implies, resides within the framework that has been designed. Given the choice of multiple options, this one is generally the best fit for the largest number of people.

In this category, users do not need to know any specifics about the database itself; the data element names, their tables, etc. are all hidden from the user and all the user sees are the business terms related to those data items.

In addition to hiding the database complexity, specific hierarchies are generally used to guide how drilling into detail information is performed. Often in these scenarios, the base data set, or framework, has been established and there is little flexibility in changing that.


Insulated self-service 

Insulated self-service is very similar to guided self-service in that the user is insulated from the complexity of the database. The difference is that often in insulated self-service, the user has a greater ability to create complex queries. 

The criteria for selecting the base data and which base data are selected is up to the user. Whereas guided self-service is more about guiding discovery in a specific topic, insulated self-service is about the user selecting the data they want, with the conditions they want, without having to know, or understand, the complexity of the underlying database. 

Visualisation options for the data are generally the same as those found in guided self-service. Options for charts, pivots, tables, and so forth.


Full self-service

While desired by power users, full self-service can cause DBAs to shudder. In full self-service, the assumption is that the user understands the underlying database and is given the ability to select tables, define their own queries (including join logic) and create complex SQL statements including sub-queries and correlated sub-queries. In short, the user essentially has full reign in a read-only environment. 

This model is one that some organisations choose to deploy because it can be done quickly and with relatively little planning; there is no need to apply metadata to the database to transform something like “cdetinf.c_addrl1” to “Customer Street Address 1”. Give the tool to the users and tell them they have self-service. (And people wonder why self-service BI sometimes has a bad name!)

On the other hand, some organisations refuse to deploy this model due to the fact that users can, and will on occasion, create runaway queries, queries that have Cartesian products, that return virtually an entire database and so forth, putting a huge workload on the database server and impacting overall database performance – especially if there are any system-critical applications accessing the same database.

Full self-service can be a sharp two-edged sword. It can make some users very happy and provide them with the ability to get needed information quickly. However, it can also result in database performance problems, user dissatisfaction, impact application and network performance and result in a self-service project being deemed a failure.


Visual self-service 

This is a new category that didn’t really exist a few years ago. It used to be that people often used their self-service BI tool to generate data to bring into Excel where they could chart, graph and manipulate it in a comfortable environment.

The advent of intuitive and powerful visualisation tools such as Microsoft Power View, Tableau and Qlikview has resulted in many users moving toward using these tools to visualise their data. The data set is pre-provided similar to guided self-service but here the emphasis is more on how they can visualise the data as compared to looking at the raw data itself or having basic pivot tables and charts.


What does it all mean?

The reality is that there is no such thing as a universal self-service BI model. Most organisations use a hybrid of multiple models. The same person within an organisation may have access to multiple models, or may be full self-service for one database and guided or visual self-service for another.

Self-service BI is a great goal and can provide significant advantages to organisations. Having good self-service BI can result in less reliance on IT for information, quicker access to information, better decision making and happier employees. 

As you consider implementing self-service BI, make sure you know your people, know their needs and then deploy accordingly. Some training will likely be required no matter which model you choose but the amount of training can be reduced by matching the right model to the right user. Keep in mind that the goal is to improve the business and putting the right information in the hands of those that need it is a good way to improve the business.

Posted by: Mark Worthen, Senior Consultant, Enterprise Applications | 12 June 2013

Tags: Business Intelligence, Microsoft, SQL, BI, Excel, Microsoft Power View, Qlikview, Reporting, Self-service BI, Tableau

Top Rated Posts

Blog archive

Stay up to date with all insights from the Intergen blog