Join Alan Simon for an in-depth discussion in this video Understanding prescriptive analytics, part of Foundations of Business Analytics: Prescriptive Analytics.
- The primary reason that organizations of all sizes and all industries are collectively investing billions of dollars today in analytics is a rather simple one. They recognize the importance of objectively driving decisions and actions by real data, not just highly subjective experience and intuition. Today's analytics are actually the culmination of more than 30 years of steadily advancing core technology and software products as well as how we bring those technologies and products together to build increasingly sophisticated systems.
We can go back before the early 1990s and we find earlier efforts, such as MIS, management information systems, DSS, decision support systems, or EIS, executive information systems. All of these efforts did provide significant business value, but all fell short of what we're able to do today with analytics. Since about 1990 the discipline known as business intelligence or BI has dominated the analytics space and we'll talk more about BI later. Regardless of whether we embarked on decision support systems or executive information systems or business intelligence, we've had one common objective no matter what we've tried.
We're trying to get insights driven by our data. When we talk about the data that we use for our analytics, traditionally it's come primarily from our internal applications, our financial management systems, our customer systems, our order entry systems, our supply chain management systems. Different applications from all over our enterprise regardless of what industry we happened to be in. Increasingly though we're bringing data from outside the enterprise in alongside our internal data. That's providing an entirely new generation of data driven insights beyond what we've typically been able to do in the past.
Also when we talk about data, it comes in many different forms. Most of what we've used in the past falls under the category of structured data. This is the data that you're probably most familiar with, things like relational databases and file systems. The types of data you have here are things like numbers and dates and short-length character strings. That's typically been the type of data most conducive to analysis. When we look at today's analytics though, we're able to add in semi-structured data, things like emails and forms, even tweets and blogs as well as unstructured data, images, audio and video.
The power of all three of these forms together, again, is opening up an entirely new realm of analytical capabilities for data-driven insights. When we talk about insights, there's a number of different facets that will describe what we're after. We need to look at insights from different timeframes. We need to know what's happened in the past. We need to know what's happening right now. To the greatest extent possible, we would like to know what's going to happen in the future. We also deal with insights of different precision.
Most of what we've worked with in the past have been cold hard facts, a sales report, a customer activity report, a statement of something that has absolutely happened for a fact. With analytics today though we not only deal with facts, we deal with hypotheses. Some of these hypotheses are rather strong in nature. They're very precise, they're very fine-tuned whereas others are kind of fuzzy. They look at certain data, they find certain patterns, which may or may not be significant and we will analyze them further and see if indeed they do have any significance and importance to us.
Our analytical capabilities need to produce insights from different paradigms. Some of them come from very specific questions, whether it's the past, the present, or the future we're looking at. We're asking a very specific question and looking for a very specific answer. Other insights come from turning our analytical models loose on incredibly large volumes of data, and then looking for hidden patterns and causal relationships and things that we may not necessarily ever have thought about. But the power of our technology ferrets out those things and shows them to us and lets us take action upon them.
Sometimes we actually try to rewrite history. We'll take historical information and ask questions along the lines of what would've happened if we had reduced our prices by 10% three years ago? It's more than just straight line math to come up with that answer. That's where all the sophisticated analytical models bring all that data together and come up with the answer of what might have happened if we had done something differently. Some of our insights are very precisely focused, maybe on a specific business process within our company. Others are broader in nature.
They're enterprise-wide. They affect many or even most of our organizations, many or even most of our business processes. Regardless of the types of insights we've been after, for the most part we've fallen short in the past. There's a number of different reasons ranging from insufficient data management technology, the capacities of data that we really need to analyze just haven't been able to be supported. Maybe the performance of the databases hasn't been as fast as we've needed it to be. Some of the software tools we've had to work with over the years haven't been as mature as they need to be.
Beyond technologies and products, we've often fallen short in how we've used those tools to build systems. We mentioned we needed insights into the past, the present, and the future, but most of the systems we built over the years have had an unbalanced emphasis looking back into the past, that historical information with far less insights into what is likely to happen and what we should do about them. Also many of the systems we've put together have been poorly architected for a number of different reasons. They work for a little while but they're not very well structured.
Eventually they run into performance issues and quality issues and other problems, which causes them to go out of favor. The biggest shortcoming beyond all of those? Workflow failures. Most people when they think about decision support systems and business intelligence don't really think of the workflow aspect, but as we'll see for prescriptive analytics, workflow is every bit as important as the technology and the analytical models. What happens a lot is that we do have a lot of available data, but it's not fully processed and analyzed.
Or we do process the data but not far enough to form a hypothesis that we can then do something with. Maybe we do form hypotheses but we don't carry them through to the point at which they're either definitively proved or disproved, so the hypotheses sit there and nothing really happens as a result. Or we may get to the point where we do produce critical insights but we never take action upon them and it's like they never happened in the first place. Essentially what happens to the detriment of all these systems is too often things fall through the cracks.
That's where prescriptive analytics comes in. Prescriptive analytics is a framework that carries our data all the way through from the first moments it makes its way into our systems, all the way through to the point at which that data's used to drive actions. By having a formal workflow as part of prescriptive analytics, we help prevent things from falling through the cracks as they have happened so often in the past. When we look at the idea of data-driven insights, it becomes a reality, not just a catchphrase. A final word about prescriptive analytics.
When you see the term with the P and the A capitalized, it may refer to a specific solution from a vendor known as Ayata. When you see prescriptive analytics in lowercase or maybe in sentence case, that's usually a general term. It's not really referring to any specific solution, and that's the term that's widely used by consultants, vendors, and industry analysts. That's the focus of what we're covering in this course. One last point though, sometimes because the trademark for Ayata is not that widely known, sometimes you may see the capital P and the capital A, or it could be referring to the general approach to prescriptive analytics rather than that vendor-specific solution.
- Exploring the analytics taxonomy
- Understanding prescriptive analytics fundamentals and workflow
- Looking at data warehousing and business intelligence
- Exploring big data
- Collecting and processing data
- Exploring triggering events
- Formulating business hypotheses
- Refining and enriching business hypotheses
- Reaching definitive conclusions
- Putting the finishing touches on prescriptive analytics