This article proposes looking at Predictive Analytics from a conceptual standpoint before jumping into the technical execution considerations. For the implementation aspect, organizations need to assess the following keeping in mind the contextual variances:
Pros and Cons of Technology Infrastructure
Providing Enabling Tools to Users
Describing Predictive Analytics
Predictive Analytics is a branch of data mining that helps predict probabilities and trends. It is a broad term describing a variety of statistical and analytical techniques used to develop models that predict future events or behaviors. The form of these predictive models varies, depending on the behavior or event that they are predicting. Due to the massive amount of data organizations are collecting, they are turning towards Predictive Analytics to find patterns in this data that could be used to predict future trends. While no data is perfect in predicting what the future may hold there are certain areas where organizations are utilizing statistical techniques supported by information systems at strategic, tactical and operational levels to change their organizations. Some examples of where Predictive Analytics is leveraged include customer attrition, recruitment and supply chain management.
Gartner describes Predictive Analytics as any approach to data mining with four attributes:
Emphasis on prediction (rather than description, classification or clustering)
The rapid analysis measured in hours or days (rather than stereotypical months of traditional data mining)
An emphasis on the business relevance of the resulting insights (no ivory tower analyses)
An (increasing) emphasis on ease of use, thus making tools accessible to business users
The above description highlights some important aspects for organizations to consider namely:
More focus on prediction rather than just information collection and organization. Sometimes in organizations, it is observed that information collection becomes the end goal rather than using that information to make decisions.
Timeliness is important otherwise organizations might be making decisions on information that is already obsolete.
Understanding of the end goal is crucial by asking why Predictive Analytics is being pursued and what value it brings to the organization.
Keeping in mind that if the tools are more accessible to business users then they would have a higher degree of appreciation of what Predictive Analytics could help them achieve.
Relationship of Predictive Analytics with Decision Support Systems or Business Intelligence
The University of Pittsburg describes Decision Support Systems as interactive, computer-based systems that aid users in judgment and choice activities. They provide data storage and retrieval but enhance the traditional information access and retrieval functions with support for model building and model-based reasoning. They support framing, modeling and problem-solving. While Business Intelligence according to Gartner is an umbrella term that includes the applications, infrastructure and tools, and best practices that enable access to and analysis of information to improve and optimize decisions and performance. These descriptions point to the fact that Decision Support Systems or Business Intelligence are used for decision making within the organization.
Interestingly, it seems like Predictive Analytics is the underlying engine for Decision Support Systems or Business Intelligence. What this means is the predictive models that result in Predictive Analytics could be under the hood of Decision Support Systems or Business Intelligence. It should be noted that organizations should proceed with caution with regards to the Decision Support Systems or Business Intelligence since if the underlying assumption is incorrect in making the predictive models then the decision-making tools would be more harmful then helpful. A balanced approach would be to create expert systems where Decision Support Systems or Business Intelligence is augmented by human judgment and the underlying models are checked and verified periodically.
Implementation Considerations for Predictive Analytics
As the descriptions above have indicated that the aim of Predictive Analytics is to recognize patterns and trends that can be utilized to transform the organization. This requires organizations to first educate themselves on what value they want and what can be derived from Predictive Analytics. Predictive Analytics is about business transformation and it needs to show what value it brings to the organization. In this regard, we have to assess people, processes and technologies of the organization in terms of the current state (where the organization is right now) and future state (where the organization wants to be). Typically, this revolves around Strategies, Politics, Innovation, Culture and Execution (SPICE) as shown below.
The assessment of people for Predictive Analytics means understanding what users will be leveraging Predictive Analytics and if they understand that simply relying on Predictive Analytics is not enough but in order to have an effective system, they need to be part of the system. This means that analytics insights need to be augmented by human expertise to make intelligent decisions. The assessment of processes for Predictive Analytics entails looking at how organizations make decisions right now and how future decisions would be made if Predictive Analytics is put into place. This includes having appropriate governance structures in place. The assessment of technology entails looking at what technologies exist within the organization and if they could be leveraged for Predictive Analytics. If not then looking at what Predictive Analytics products are in the market that would work for the organization and are they flexible enough in case the underlying assumptions for the predictive models change and when predictive models become obsolete.
The advanced techniques mentioned in the book, Seven Methods for Transforming Corporate Data into Business Intelligence would be applicable to Predictive Analytics. These methods are:
Data-driven decision support
Technologies Used for Predictive Analytics
Gartner has been publishing its Magic Quadrant on Business Intelligence and Analytics Platforms since 2006. Due to the increased importance of Predictive Analytics in the marketplace, Gartner decided to create a separate Magic Quadrant for Advanced Analytics Platforms which focuses on Predictive Analytics and published its first version in February 2014. Since it is the first version of the Magic Quadrant, all vendors listed are new and no vendors were dropped.
As we can see from this Magic Quadrant that it includes well-known vendors but also vendors that are not as big or as well-known. It is interesting to note that open-source vendors such as RapidMiner (a Chicago company) and Knime (a European company) are in the same Leaders Quadrant as well-established vendors such as SAS and IBM. While there are some issues with these open-source vendors as stated in the report but perhaps this Magic Quadrant is also an indication of where the next generation of analytics would come from. Due to the very nature of open-source, there are more opportunities for cheaper customization which would give the organizations the flexibility to be as granular as they want to be. Of course code stability and lack of proper documentation are issues that organizations need to be cognizant about. Organizations may also want to “try-out” these open source tools before they make a big commitment to proprietary software to see if Predictive Analytics is something they want to invest heavily in.
Using Predictive Analytics in Specific Industries
There are many industries that utilize Predictive Analytics. The organizations in these industries either use Predictive Analytics to transform their business and/or to address certain areas that they would like to improve upon. Following is a list of some of the industries that utilize Predictive Analytics:
How is Predictive Analytics used?
Oil and Gas
Well and Field Asset Surveillance
Equipment Reliability and Maintenance
Adjust production schedules
Tweak marketing campaigns
Human Resources Allocation
Supply Chain Optimization
Electronic Health Records
Nation-wide Blood Levels
New Business Models
While there are many examples of industries that have embraced Predictive Analytics but there are other industries that have not fully accepted it as a new reality. These industries have many excuses for not considering Predictive Analytics but typically revolve around scope, quality, cost, and fear of the known. However, the tide might be changing for these industries as well since industry bloggers are beginning to insist on how Predictive Analytics could be leveraged for competitive advantages.
Predictive Analytics can come in handy in making organizations analytical and becoming a better version of themselves. However, Predictive Analytics can be a deal-breaker if organizations have attempted and failed in the past and for this very reason, Predictive Analytics should start as a discussion first. This discussion should revolve around asking which areas need improvements and among other things determine if Predictive Analytics could be something that could help. After a successful Predictive Analytics initiative, other areas could be potential candidates as well.
An important thing to note is that Predictive Analytics is an organization-wide initiative that has touchpoints across the organization and thus the maturity of the organization has to be seriously considered prior to go on a Predictive Analytics journey. No matter how good Predictive Analytics can be for the organization but if the organization is not mature enough and it does not have the right governance, processes and feedback mechanisms in place then it might turn out to be another attempt at glory but nothing to show for it.
Prescriptive analytics is used for performance optimization. This optimization is accomplished by using a variety of statistical and analytical techniques to identify the decisions that need to be taken in order to achieve the desired outcomes. The data sources used for the determination of outcomes can range from structured data (e.g., numbers, price points, etc.), semi-structured data (e.g., email, XML, etc.) and unstructured data (e.g., images, videos, texts, etc.).
If done correctly, Prescriptive Analytics is the Holy Grail of analytics. However, if done incorrectly, it can result in misinformed decisions that can be outright dangerous. Individuals and organizations have to understand that even if the data is correlated that does not mean that there is some sort of causation. A general example of this is when in a news report, the host(s) says that the survey has shown that x is correlated with y but then they go on how y was caused due to x. This is simply what I call “jumping the data gun” and organizations that are not aware of this can fall into this trap.
Another thing to be aware of is that after the Prescriptive Analytics gives you certain courses of action and you apply those actions, keep track of how well your Prescriptive Analytics is performing as well. In other words, you have to measure the performance of your performance optimization ways. The reason to do this is that over time you can see if the models presented by your Prescriptive Analytics engine are worth following, re-doing or dumping.
To get you started, here are a few questions to ask:
Who uses prescriptive analytics within, across and outside your organization?
Who should be using prescriptive analytics within, across and outside your organization?
What outcomes do prescriptive analytics tell you?
What outcomes prescriptive analytics should tell you?
Where is the data coming from for prescriptive analytics?
Where should the data become from for prescriptive analytics?
When prescriptive analytics is used?
When prescriptive analytics should be used?
Why prescriptive analytics matter?
Why prescriptive analytics should matter?
When you are asking the above questions, keep in mind that Prescriptive Analytics uses data to create a model (aka a data version of the world) that is used by individuals and organizations to make real-world decisions. But if the model itself is flawed then you are bound to get answers that although might look visually appealing are completely wrong. It is not all doom and gloom though. In fact, Prescriptive Analytics is used in determining price points, expediting drug development and even finding the best locations for your physical stores. Companies like Starbucks have been using Prescriptive Analytics in the last few years to determine the best locations for their next coffee stores. Interestingly, some have claimed that wherever Starbucks goes, the real-estate prices also increase. While there is some correlation between a Starbucks coffee store opening with increased real-estate prices but this does not mean that because of Starbucks coffee stores the real-estate prices increase.
Predictive Analytics is a branch of data mining that uses a variety of statistical and analytical techniques to develop models that help predict future events and/or behaviors. It helps find patterns in recruitment, hiring, sales, customer attrition, optimization, business models, crime prevention and supply chain management to name a few. As we move to self-learning organizations, it is imperative that we understand the value of Business Analytics in general and Predictive Analytics in particular.
It turns out that Predictive Analytics is about Business Transformation. But in order for this Business Transformation to take place, you have to take into account the organizational contexts in the following ways:
Strategic Perspectives: Not all organizations are the same and thus what works in one organization might not work in yours. Based on the knowledge of your organization’s maturity, you have to decide if Predictive Analytics is going to be a top-down, bottom-up, cross-functional or a hybrid approach. Additionally, take into account what should be measured and for how long but be flexible in understanding those insights might be gained from data that might initially seem unrelated.
Tactical Perspectives: One of the key factors in Business Transformation is change management. You need to understand how a change would affect your organization in terms of people, processes, and technologies. You have to take into account the practical implications of this change and what kind of training is needed within your organization.
Operational Perspectives: It is all about how the execution of Predictive Analytics is done within your organization. To fully integrate Predictive Analytics into your organization, you have to learn from best practices, learn the pros and cons of your technology infrastructure and determine if the necessary tools are intuitive enough for people to make use of them.
Now that you understand the different organizational perspectives, it is time to ask the following:
Who uses Predictive Analytics to make decisions?
Who should use Predictive Analytics to make decisions?
What happens to decisions when Predictive Analytics is used?
What would happen to decisions if Predictive Analytics will be used?
Where does the data for Predictive Analytics come from?
Where should the data for Predictive Analytics come from?
When is Predictive Analytics relevant?
When should Predictive Analytics be relevant?
Why Predictive Analytics is being used?
Why Predictive Analytics should be used?
When you ask the above questions, keep in mind that the reliability of the information and how it is used within the organization is paramount. A pretty picture does not guarantee that the insights you get are correct but you can reduce decision-making errors by having people who understand what the data actually means and what it does not.