Putting The B Back Into BI

4 min read

I was planning to write about putting the intelligence back into business intelligence. About making data tools smarter than 90’s style drilling and 00’s style charting. About massive analytics and graph DBs and dealing with big data. And how BI should be more AI. That would be my inner geek’s blogpost. It would be about the promise of BI – making companies smarter through the use of data.

But when I put my business hat on, I don’t care about the intelligence. So what if a tool gets smarter? So what if the company gets smarter? The fact is, being smart amounts to precisely nothing without execution. Does the tool directly help a company increase profits, execute, and reach goals faster?

The whole point of BI is (or should be) business, not intelligence. And that is the real problem with BI. Not missing the I, but missing the B.

What Doesn’t Work

Dashboards. Reports. Exploration. Self-service. Why? Because digging around increasingly large datasets is not what you want your people to spend time on. It doesn’t directly increase profits or speed up execution. Quite on the contrary. Executives today are swamped with information, 95% of which is noise. Digging through that noise distracts them from work, burdens their minds and clots their decision-making.

It’s what’s well visible across the companies we work with. Their reluctant adoption of BI on the operational level, low use of reports, failed expectations from big data projects, a sobering attitude toward data-driven promises.

So instead of focusing on the tools, we stepped back and asked our pilot customers a few daring questions. Mainly, what is the “real work” that managers do besides our beloved analytics? Is that what maximises value for the company? And why is it that BI is not helping them with the “real work” today and how could it?

Digging Deeper

Question: All tools aside, what is the first question that a manager typically asks on a Monday morning? How does he decide what’s important? How does he choose what to focus on? By common sense, this is the most important decision of the day: if people don’t prioritise well, nothing else matters. One can’t win a horse race by growing nice potatoes.

Question: How does the company detect problems? On any given day, people are working out some issues, dealing with the hot coals. Quite often, this is the bulk of their work. So how did they find out about them? Did a manager tell them to deal with it? How did he find the problem? Through a set of reports? Digging in excel? In weekly meetings? In quarterly reviews? Whichever way it is, there is a typical process for raising red flags.

Question: What happens when there are multiple issues? Who decides the priority – the manager or the analyst? Is every problem valued in financial terms? Is it easy and straightforward to assess the financial value of the problem?

Question: Does every problem have an owner? It’s no secret that ownership really helps with the resolution of problems. No ownership, no accountability, no results.

Question: How easy is it to detect the root cause of problems?

The answers to these questions are typically an intuitive part of a company’s culture. And it is in their conscious reexamination where we consistently find a lot of value. And the huge gap between what BI is and what it could become.

The Reality

After assessing the actual state in a number of companies, patterns emerge. Problems are not clearly prioritised. There’s a bias toward certain issues (those that are easily detected and those that are easily resolved). There’s no clear ownership of many problems. Solutions to previous issues are not retained or converted to experience. Decisions are not linked to their outcomes. It takes weeks to detect root causes of problems.

To maximise profit, a modern BI tool should provide valuable assistance with all of the above. It should help with a large part of the “real work” of management. And that’s achieved as follows: Automatically detect the largest problems in the company. Clearly value each problem in financial terms, so that prioritisation is straightforward. Assign each problem to the most appropriate owner.

In essence, a modern BI tool should make everyone in the company immediately see the top 5-10 issues in their area of responsibility, with a single click. That is where a manager’s work should start – in dealing with those issues (or knowing exactly who is dealing with them). Their detection, evaluation and financial prioritisation can be automated – saving time and mental energy. This is how we do it in Stories:

Objections might come at this point. How can the tool know what is important?

A simple reply: How can the manager know what is important? I offer a simple test. Go through your reports to come back with the top ten issues right now. Short term and long term. This is an analytical task of intimidating magnitude. In a company of a decent scale, it’s simply not humanly possible to reliably provide the real top ten issues. There are literally millions of individual analytical steps across the whole dataset that have to be performed to return the answer. Let alone to provide this as a personalised answer to every manager in the company. Let alone to prioritise across various siloed business units of the company. So if the methodology can be agreed upon, software is much better equipped to do the heavy lifting required.

And let’s take if further. What do you ask first when you see this detected problem?

In our experience, the first question that comes in 90% of cases is “Why is this dropping?” With a typical BI tool, people start drilling around to find the root cause. Even with only 10 dimensions, this can take an hour if someone knows what they’re looking for. And very often, in large companies, just answering this one question takes 2-3 weeks. And again, there is a lot that can be automated here: namely explanations and context.

Explanations

Why are the sales dropping? It could be down to a single customer segment. It could be a specific model. It could be customers coming from a certain city. In other words, the answer might lie in a subset of the data charted above.

Context

In other cases, a drop like this can be caused by the larger environment. Analysts checking such hypotheses ask: how is the whole market doing? Maybe this drop is in line with overall sales? Is it still an anomaly compared to other countries, other manufacturers, other car models?

Again, all these hypotheses can be tested manually, or in an automated fashion. And again, the bar should be raised for the expectations of what a BI tool can provide. It should provide a one-click overview of explanations and context as a standard service. The analysis of typical issues can be reduced from days to minutes. And that is a huge impact on the bottom line – suddenly everything in the company can move faster. By saving 80% of time needed for executing the analysis, people have more capacity and energy to execute the resolutions.

And that is the final part where BI needs to raise the bar. A good BI tool should assist in resolving every such issue. The possibility to plan next actions, collaborate around problems, see what is getting worse while it’s being resolved, see what is not being resolved, see what helped in similar situations in the past, and track decisions that have a real financial impact – those are all pieces of the puzzle that together make a company agile. And truly data-driven. And learning. And isn’t that the goal of business intelligence?