While the hype of Self-Service BI is already everywhere, in our talks we find that basic expectations are not always clear, so we wrote this short intro.
Self-service is reflected more and more in virtually all products, some of them focusing exclusively on this domain, but the approach is way more important than the actual software, so this summary will be agnostic in this regard.
Contrasting the "traditional" setup with the "self-service" approach is of course exaggerated. However the difference in tendencies are very clear, and this exaggerated contrasting is needed if we want to understand the key concepts.
And to be honest, self-service does feel very different both on the business side of the projects and on the user side of the software products.
The most important feature of the traditional BI approach is that business users, those who want something out of company data, are by and large different from those who can access information.
The "business" (in marketing, sales, operations, product management etc.) doesn't have direct access to the vast amount of data buried somewhere in databases. Even if they had, they (1) would need special technical skills to get the data, and more importantly (2) wouldn't understand the machine focused structure it is stored in. Even purpose built data warehouses are most of the time very messy to deal with, and it can border the impossible to figure out that FIN_PROD(2)_B43NH_ID is actually the product ID dimension you are looking for.
True: in some cases this is due to access right philosophy, or purely to silos in the organisation. If business units keep their data to their own and don't share it with others, this probably requires a top management initiative to be solved.
But more often than not, business units are happy to cooperate, it's just the corporate setup that won't let them. The basic setup is this: if you want a report, go to BI, and ask for a new report. The dedicated BI unit -against its will- becomes a bottleneck, and struggles to keep up with the incoming torrent of questions. The paradox is this: if BI produces fast and high quality answers, this will only increase the number of questions. It is the very nature of data that an answer always generates new questions. If in the end the BI team fails to keep up, the business will turn to other sources.
No matter what, sooner or later people will start to create a spaghetti of isolated data sources, excel sheets, reports and even databases.
And let's not forget the time factor. In a traditional BI cycle you define your requirement, ask for a delivery date, perhaps get a quote from the outsourced development company, get the first report, test it, put it to production, and finally start using it. This can take anything from weeks to even months.
In contrast to the traditional approach, self service BI puts business teams in control by providing a meaningful metadata layer, through which anybody can make sense out of complex corporate data. Database tables are already joined for you, unimportant fields are filtered out, important dimensions and measures are renamed in a friendly way, common KPIs or other measures are calculated, comments are included etc. Also, template reports can give you a skeleton for creating other ad-hoc reports, so you don't even start from scratch.
Based on this metadata layer and the templates created with it, users are able to turn to the data themselves, and create, customise, transform, and also to publish their own reports.
This doesn't mean at all that the BI team would loose its importance. They just refocus on what they do best: translating between business and technology. They maintain and develop the meta layer, create templates, and run the entire system.
It's very important, that the existing investment can be preserved with the right selection of tools: data warehouses and even individual excel sheets can be fed into the metadata layer to create a "single point of truth".
In a self-service tool, users create reports by drag-and-drop, instantly analyse their data and share the findings. With this setup, the question-answer cycle is reduced to hours. And even if something very complex comes up, BI is much more likely to have the resources to create a new template for you.
From the company's overall perspective, business users gain direct access to actual data, and get the opportunity to rely on hard facts whenever they choose to.
We use Tableau in the following video to illustrate the concept with one very simple bar chart. Try to imagine how many individual reports you would need to specify to have a developer produce all this for you: all the possible levels of aggregation AND all the 72 possible permutations of the dimensions.
The illustration has two hierarchical data dimensions (Regions: Directorate, Area, City; and Products: Category, Segment, Type), showing (1) how changing the order rearranges the data break-down, (2) how drilling down changes the level of aggregation, and (3) how removing an upper level of aggregation changes the entire dimension.