BI analytics value

Getting real value out of BI: closing the gap between analytics and potential

The output of BI is used in organizational decision-making, since it is not the process or technologies that are used by decision-makers but rather their output. Arisa Sholo, Copenhagen Business School, 2012

Traditional Business Intelligence (BI) is not working. BI is supposed to help businesses make data-informed decisions to improve outcomes, but the reality is that most are falling back on gut instinct to drive their actions. Thanks to a serious mismatch between existing BI infrastructure, tools and end-users, there’s a big gap between analytics and potential. While BI seems to have taken several steps forward since its inception, it has also managed to jump backward with every technological advance.  

To get a clearer picture of what we’re talking about here, we need to step back a little. 

BI grew within the IT-centric system of record in the 1980s, where those of us with business questions would go to the specialists who ran our databases, ask for reports, and through the painful, iterative dance of “that wasn’t exactly what I was looking for, I actually need *this*” eventually end up with… something, at least.

By the early 2000s BI evolved into visual-based data discovery, which offered a simple proposition: what if, instead of waiting for IT to figure out what they were looking for, they could make their *own* damn charts, using simple drag-and-drop user interfaces?

Tremendous success followed. Business Intelligence tools that turned dimensions and measures into charts and graphs flourished, and new names like Tableau, Qlik, and Spotfire started to eclipse the venerable Cognos, Business Objects, and Microstrategy. And as the technical folks responsible for implementing this new breed of tool got asked for more and more sophisticated analyses, eager product leaders converted requests into shiny new features and shipped them at a breakneck pace.

Visual-based data discovery has a simple premise: Most business questions aren’t hard to answer if you know what you’re looking for, and if a business user understands her question, she should be able to drag and drop dimensions and measures until a chart reveals the answer. Dashboards and self-service BI tools are meant to make analytics quick and easy, correct?

And there’s the rub.

Industry analyst firms are circulating a disappointing statistic: Close to 80 and 90% of knowledge workers lack the technical skills, data literacy, or access to make effective use of BI tooling. 

This is because today’s self-service BI tools are influenced by the needs of data specialists (trained business analysts, data engineers, DBAs, and data scientists) who require more advanced features and capabilities to support their use cases. So that’s why you will notice that Tableau has as many control surfaces as a 737, why Qlik has an expression editor, and why Sisense allows filter values to be expressed in code.  All the vendors in the BI space convert feature requests to software, and race to anticipate the next request of our buyers – which leads to tremendously powerful tools with features 90% of users don’t know how to use.

Essentially, self-service BI has become too complex, and too impenetrable for its target user. So now, we’ve put ourselves back into the IT-centric system of record. Not only are we back to where we started, but now we’re doing it with tools the data pros don’t like (they don’t need a visual tool, they can write SQL and Python) and the business users don’t like or don’t know how to use.

So what? Throw it all out? Go back to stone tablets? What if, instead of siloing analytics within the analyst community, you could incorporate data and insights into every workflow your team uses, to drive every decision? 

This is infused analytics, the much-needed, seamless evolution of data analytics that BI has been sorely missing. 

Here’s how we can do it: let’s hook our BI infrastructure – with its powerful cloud data warehouses, its beautiful data models, and its ability to mash up metrics from many sources – directly to the tools we *do* know how to use to analyze data. Plug a set of views and a good NLQ engine into Google Sheets, or Excel, or MS Teams, and let’s actually answer our next questions. At least for the 80% of our questions that are easily predictable (does anyone really need to guess what the VP of Sales is going to ask about?) – and for the other 20%, well, that’s why we have statisticians and data analysts on the team.

Author: Scott Castle

Source: Sisense