Sponsored Content

How to make BI far more effective–and why the struggle seems so difficult

There is little argument that business intelligence–including AI and AI’s ML–has delivered profound advantages to enterprise operations, positively impacting just about every business unit. 

But there also is regrettably little argument that BI is not always implemented with the business user and/or the consumer end-user in mind. That means that the promised ROI of those BI rollouts can be compromised. 

The good news is that most of these challenges can be fixed. BI insights today are too difficult for most LOB managers to master.  

In the earliest days of BI, the goal was self-service. BI enthusiasts wanted to take IT and data analysts out of the equation as much as practical, instead allowing end-users–who presumably understood the business far better–to run their own data-crunching. It was a great idea then and it still is today. 

Humans are emotional

That said, companies are populated with humans–for better or for worse–and that brings emotions, fears and irrationality into the picture, both from IT and the end-users. There are, however, some easy ways to reduce the friction and to realize those stronger results.

The problem is that the initial tools were sophisticated and, from the end-user perspective, overly complex. In an ideal environment, the person making the inquiries is an expert in the business, the data, how far analytics can and can’t go as well as the most effective and efficient ways to structure queries.

In a realistic business environment, there are very few–if any–people with all of that experience and knowledge. There is the theoretical possibility of extensive training, either training IT people and data analysts to understand every corner of the business intimately or to train LOB operatives to understand the analytics business and what their databases have and what they don’t have.

How much training is practical?

The best route is to make the software more approachable to the businesses. Boiled down to the basics, analytics is typically little more than answering questions that the LOB executive needs answered to run the business better. That reflects about 80 percent of LOB inquiries. Those are the easier ones.

The other 20 percent need far more investigation, leaps of insight, and often a skilled mathematician.  If we don’t call out that there are hard questions worth answering with specialists, we’ll be hammered by those specialists for “not getting it.”

Most of these questions are simply not that complicated. Why are costs going down? Why is this geography suddenly losing customers? What do we need to change to boost our margin by 15 percent and what is the best route to do that?

Many of these LOB managers know precisely what they want answered. What they don’t know is how to ask it properly, in the sense that they express it in database language or even in terms that exist in a data repository somewhere.

Those business managers know how to express their question in business wording, but what they don’t know how to express it in database language. The software usually insists that the phrasing is in terms that exist in a data repository somewhere.

Then there are the data issues themselves. Business managers rarely know what data the company currently saves and how to extract the needed answers.

Enterprises need to bring the data to those knowledge workers, embedded into a familiar context that delivers relevant insights where they’re needed most. These capabilities must not be placed in an isolated environment that will be unfamiliar to these business executives. Instead, it needs to be seamlessly integrated into the tools they already use, such as a workflow tool, ERP, Excel or PowerPoint. 

Many businesses are already trying to do that, but it’s not working because the answers they receive deliver data, not actionable insights or the drill-down context that allows the creation of said insights. The apps are bringing the data–reams of it–in a form very difficult for the average worker to navigate.

1. IT needs to stop trying to govern data out of a fear that they’ll get blamed for mistakes

There is nervousness approaching paranoia from IT and data teams when it comes to giving LOB end-users access to all of the data. The fear is that when end-users ask the wrong questions or query the wrong data and understandably get wrong or irrelevant answers, IT will get blamed. “Our data was obviously not clean,” an end-user might complain. To which IT and data reply: “When they get it wrong, they blame us.” 

This again points to the need to make the interfaces as intuitive as possible. Hide the complexity, reveal the answers – and allow users to explain how they reached their conclusions. There are better ways to run a business than teaching everyone to be a statistician.

2. Use KPIs in such a way as to make them far more actionable.

Use KPIs to project outcomes and suggest course correctionsLet’s say that the project goal is to improve productivity in one workgroup. The analytics could report back that one employee is 40 percent less productive than various peers. 

That detail, although interesting, may not help change behavior to improve productivity. But what if the report instead told that employee the specific work habits and behaviors that were shared among the most-productive employees in the group? That gives that employee a concrete and actionable way to improve productivity.

3. Embrace change

Nothing at all about the analytics equation stays the same. 

The Data

The data itself changes in two directions. The data and IT teams are always coming up with new ways–often in conjunction with marketing–to gather information about customers, prospects and every other aspect of business operations. That means far more data.

At the same time, compliance and cybersecurity managers are trying to delete as much data as practical, so that the enterprise is retaining only what it truly needs. The less data retained, the less that can be stolen, argues the CISO’s office. The less data retained, the fewer headaches with GDPR’s right-to-be-forgotten and other compliance and legal privacy worries, argues compliance and legal. That means far less data.

Then there is the nature of the data, which also constantly morphs.

The Analytics

Like any other software, new versions–as well as entirely new applications–materialize constantly. This means that the kinds of insights that can be gathered also change constantly.

The Business

With new business units being acquired and others being sold, with geographies entering and exiting, with remote sites, expansion of cloud, the increased/decreased use of contractors–nothing about the questions that need to be answered is static.

This means that executives must plan for what they have now but also plan for those requirements to change and evolve with little to no notice. The data landscape will always change.

The business intelligence problem will continue until the software interface and output is reenvisioned with a focus on the end-user. To make BI useful to the LOBs, developers must make their answers useful and actionable for them. 

To see how we’ve gotten this process to already work at scale, please visit www.sisense.com/infusion.