(This is the final part of a three-part series illuminating the challenges of data analytics and describing a proven solution to the problem.)
In part one, Measuring the Cost of Doing Business With Insufficient Analytics, we learned that, on average, data analysts spend 4 days a week organizing data, leaving only one day per week to conduct analysis and develop reports. The result is a lag between when information is needed and when it becomes available, leading to poor decisions, and ultimately to potential profit going unrealized. I also cited evidence that companies effectively utilizing advanced analytics to drive decisions have twice the revenue growth, and 20 times the EBITDA of competitors using gut feel to make decisions.
In part two, Five Reasons Your Data Analyst Can’t Analyze, we examined the reasons why so much time is spent organizing data, and why most companies fail at analytics. While it is critical to understand each of the five issues (and worth the time to go back and read about them) we could summarize them simply with a single statement: transactional data must be transformed in a number of ways before it is useful for reporting.
Now that we know the nature of the problem, we understand the massive cost of leaving it unchecked, and we understand the underlying causes, it is time to turn our attention to the solution. With hundreds of business intelligence service providers and an even higher number of available analytics software applications on the market, there is no shortage of options. The truth is that there is no one “right way.” The challenge is to find the way that is right for you depending on the size of your business, the internal resources you have available, and your available budget. The cost and effectiveness of different options will vary widely, so perhaps the most important thing you can do is to consider multiple possible solutions before making any investment.
The good news is that for purposes of this discussion we can completely avoid making judgements about any specific application or service providers. Instead, let’s take a moment to discuss the science of business intelligence. This is a somewhat technical topic by nature, but it isn’t necessary that you understand the underlying details, just that you have a working knowledge of the critical concepts. It turns out that some very smart people have been acutely focussed on this exact problem for the past few decades. Fortunately for us mere mortals, they have devised and shared an approach that has proven successful time and again. This approach, known as dimensional modeling, gives us a framework for how to transform the data after we extract it from the transactional systems, prior to loading it into a data warehouse.
Here’s what you need to know. A dimensionally modeled data warehouse will enable high-performance reporting with any reporting tool, whether you choose to use Excel, Tableau, Cognos, Business Objects or anything else. This is the magical key to analytics success.
So, all we need to do, then, is build one of these nifty things. Or, to be a little more precise, we need to write some code (called ETL, for extract, transform and load) that automates the process of moving data from the source systems into the data warehouse on a recurring basis. Once this process is designed and automated you will have reporting-ready data available at your fingertips every day. If the underlying dimensional model is any good, it will become a trivial matter to conduct data discovery and advanced analysis. Your data analysts will be able to devote all of their time to business analysis and optimization. As a result your company will experience all of the benefits of fact-based decision making, including dramatic improvements in EBITDA and profitability.
One final note: While the star schema dimensional model is effective, there is a cost associated with its development, and this level of investment may not be appropriate in all cases. There are a variety of data solutions that can support business intelligence, and it makes sense to understand them all so you can identify the level of investment that is most appropriate given your data sources and reporting requirements.
