Predictive Analytics Enters the Mainstream · Getting Started: People For years predictive analytics has been the domain of the statistician and the data scientist, but today that

Ventana Research: Predictive Analytics Enters the Mainstream

1 © Ventana Research 2014

White Paper Sponsored by

Taking Advantage of Trends to Gain Competitive Advantage

Predictive Analytics Enters the Mainstream

http://www.ibm.com/us/en/


Table of Contents

From Obscurity to Availability 3

Getting Started: Process 4

Getting Started: Information 5

Getting Started: People 6

Getting Started: Technology 7

Building the Business Case 8

Taking the Next Step 9

About Ventana Research 11

2

© Ventana Research 2014


Our benchmark research reveals that only one in five (21%) participating organizations is very satisfied with its predictive analytics efforts.

From Obscurity to Availability Predictive analytics is the descriptive label for a range of statistical analyses that use historical data to anticipate the future. It encompasses a variety of techniques including data modeling, data mining and machine learning. Until recently, predictive analytics was a complex practice conducted by specialists with titles such as actuary, statistician and marketing scientist. In order to conduct such an analysis, the specialist often had to work with multiple departments and IT staff to acquire data from various systems, painstakingly prepare the analytical data set and then build a predictive model. If the model had to be integrated with another system such as a customer database, the analyst would have to write a document that explained the model’s specifications to engineers. They in turn would

write the code to execute the model specifications in the database for the users in a line of business. It’s a messy process that invites confusion about who must respond next and often involves delays and errors. Moreover, because of all this complexity, the end result frequently fails to meet expectations. Our benchmark research on predictive analytics reveals that only one in five (21%) participating organizations is very satisfied

with its predictive analytics efforts.

Nonetheless, organizations have high expectations for future use of the technology: 86 percent of the participants in our research asserted that predictive analytics will have a major positive impact on their organization. And nearly one-third (32%) indicated that it could enable them to do things they couldn’t do before. Technology is now widely available that can help organizations realize the potential of predictive analytics without spending great sums on specialists. However, technology alone rarely solves business problems. Organizations must understand and address the challenges in each of the four key dimensions of performance: processes, information, people and technology.

3



More than two-thirds (68%) of organizations that have implemented predictive analytics said they have gained a competitive advantage from it.

Aligning these four dimensions is rarely an easy task, but the outcome is worthwhile. In this case, more than two-

thirds (68%) of organizations that have implemented predictive analytics said they have gained a competitive advantage from it. Let’s consider how to address each of the four dimensions to prepare to use predictive analytics.

Getting Started: Process From a process perspective, informal

workflows centered around exchanging working files such as spreadsheets among a few stakeholders are

notoriously inefficient. In our research 35 percent of those organizations using spreadsheets for analytics reported that data errors are common; 26 percent said that formula errors are common; and 32 percent said out-of-date information is a problem frequently or all the time. Innovative organizations choose another path. They use a formally defined collaborative process supported by technology that manages workflow among domain experts and data scientists and automates many of the steps in predictive analytics, from data collection through providing results to decision-makers. When compared to spreadsheets, these tools do a much better job of addressing the key technological challenges associated with predictive analytics: integration with the current architecture (cited by 55%), access to source data (35%) and inaccurate results (22%). Models are central to the predictive analytics process. They guide the transformation of input variables into probabilities of potential future outcomes or target variables. But over time models can get stale and produce inaccurate results. Thus to ensure effectiveness it is important to have a process to update predictive models frequently. Our research shows that this activity does not get the attention it deserves: Almost two-thirds (63%) of organizations update their models only monthly or even less often. Yet the one-fourth (24%) of organizations that are very satisfied with their analytics update their models daily or more often.

4



Organizations also should develop a process for scoring new records as quickly as possible – for example, to assist customer service representatives during interactions or to make product recommendations to online customers based on items in their shopping carts. Here again we find a relation between an effective process and satisfaction with predictive analytics: 88 percent of those that use real-time scoring regularly said that they are satisfied or very satisfied with their predictive analytics tool. Understanding the process plays an important role in the deployment of predictive analytics; a knowledgeable examination can reveal inefficiencies that may put the organization at a competitive disadvantage. It is imperative to look at the entire predictive analytics cycle – from extracting data through its preparation, model building, model scoring, model management and end-user consumption – to determine whether and how each step can be improved.

Getting Started: Information Predictive analytics is fueled by data, most prominently customer (69%), marketing (67%), product (55%), sales (54%), financial (51%) and employee (34%) data. Needless to say, data sources for predictive analytics are multiplying rapidly, as are their formats. Increasingly the data comes from outside the company firewall as well as within it.

The significantly increased volumes of data from more diverse and distributed sources is commonly referred to as big data. Big data is linked to predictive analytics: The largest percentage (64%) of participants in our benchmark research on big data analytics said applying predictive analytics is a critical capability to have when analyzing big data. This research shows that the most common types of data used for big data analytics are transactional data from

5



applications (60%), external sources (50%), content such as documents and Web pages (49%) and event-centric data (48%). Our research on information optimization shows that managing all this data is an issue. Analysts spend most of their time on preparing data for analysis (48%) rather than actually interpreting the data (33%). This situation takes skilled employees away from the tasks that add the most value to their organization. To address the issues in managing information and preparing data sets for predictive analytics, organizations must be able to extract data from various sources, merge the data, reconcile different data formats, and enrich and clean the data. Automating these tasks can free analysts to do what they are trained for.

Getting Started: People For years predictive analytics has been the domain of the statistician and the data scientist, but today that is changing. While personnel

from data science and data mining (32%) are still the most common users, others from business intelligence and data warehousing (31%) and from lines of business (19%) now use predictive analytics as well. Even so, organizations most often are satisfied with a predictive analytics initiative when a data mining resource (70%) is involved, as compared to those in which only lines of business (65%) or business intelligence resources (59%) are involved.

This research finds that training and support for users of predictive analytics is often lacking. Fewer than half of organizations provide adequate training in any of three key areas: concepts and techniques, their application to business problems, and the use of products. Even fewer (31%) provide adequate consulting resources or adequate help-desk support (24%). Many of these skill and training issues can be addressed by using specialized tools now available. For instance, software-guided analytics

6



The biggest challenge associated with predictive analytics (cited by 55% of research participants) is to integrate tools with their organization’s information architecture.

tools that perform automated data mining as well as collaboration capabilities make it possible for business users lacking specialized skills to work in predictive analytics. Statistical libraries allow users to implement statistical processes without building them themselves. Embedded training software that provides just-in-time tutorials and access to a knowledge library around different statistical techniques can help users from multiple roles within the organization to understand and interpret data. In these ways, cross-functional teams can work together to produce the desired outcomes.

Getting Started: Technology Many organizations have complex technical environments associated with analytics. These include a variety of databases and business

intelligence, data discovery and personal productivity tools. Not surprisingly, the biggest challenge associated with predictive analytics (cited by 55% of research participants) is to integrate tools with their organization’s information architecture. Research participants want predictive analytics integrated with other kinds of tools as well, especially business intelligence (58%) and other business applications (56%). Modern tools can help in this regard, enabling more companies to use predictive analytics for competitive advantage. Built-in

connectors and data preparation software (often part of the predictive analytic tool) address the data

extraction issue. In-memory processing enables data profiling for advanced data discovery processes, cleansing and data integration. Standards such as PMML make it possible to translate models into SQL. An array of tools support writing them in R, the open source language for statistical computing, and incorporating them directly into the analytic workflow. User-defined functions and in-process algorithms allow models to be run directly in the database. And flexible deployment options enable updating models on a regular basis and deploying models into front-line applications by integrating both scoring methods and rules to provide key metrics and next best action.

7



Building the Business Case Decision-makers with authority to support and approve an initiative usually are interested in its value to the company’s performance – that is, in the business case. Therefore, we advise enumerating benefits of

predictive analytics that will appeal to those leaders. Asked in our research about achieved benefits, the largest percentages of companies named competitive advantage (68%), new revenue opportunities (55%) and increased profitability (52%). The business case should identify functional areas that are likely to benefit from predictive analytics; most often identified by research participants are marketing (65%), research and development (61%),

sales (59%) and product management (50%). It should also identify the functional use; most often cited were forecasting (72%), market analysis (67%), customer service (45%), product recommendations (43%) and fraud detection (34%). Return on investment (ROI) is a critical metric. For predictive analytics this entails defining expected output, how it will be used to achieve enumerated benefits and where in the organization these benefits will be valuable. For example, competitive advantage and thus ROI can be achieved by improving understanding of different customer loyalty segments and predicting their preferences based on behavioral, attitudinal and demographic profiles. This information can be used to help define internal initiatives such as a product development and portfolio strategy, and also to create marketing communications campaigns that target different segments and enhance the brand in the mind of the customer. More tactically, the ability to predict what an individual customer might buy next can enable the company to better meet customer needs. Proper use of this information can result in increased revenue through

8



Investigate tools that bridge the skills gap and reduce reliance on data scientists by providing intuitive interfaces, automated data mining and collaboration.

activities such as cross-selling and up-selling and impact key metrics such as revenue per customer, wallet share and customer satisfaction. The business case must address likely challenges in implementing a predictive analytics program. In general, lack of resources is the most common barrier encountered, cited by three in five (59%); two-thirds of management (that is, vice presidents) and executives said this is a barrier. Lack of awareness of the value of these analytics (45%) ranked second. Management and executives generally have a more complacent view of this than do users, who much more often are not satisfied with their predictive analytics; for the upper levels of the organizational chart the results loom larger than the effort it takes to produce them. Given this backdrop, another necessary part of the business case is to point out the insufficiency of current tools for conducting advanced analytics. Determine the extent to which people in the organization use spreadsheets and other manual processes for predictive analytics and point out their deficiencies for enterprise processes and complex analysis. Finally, clearly address each of the four dimensions outlined

above to round out the business case.

Taking the Next Step Organizations should begin by building a business case that addresses key areas of competitive advantage, revenue opportunities or operational efficiencies. Show why current legacy tools cannot attain these benefits and identify the opportunity costs associated with not moving to a more dedicated approach for

predictive analytics.

Build a cross-functional team that incorporates business domain knowledge, statistical knowledge and technological capabilities. Investigate tools that bridge the skills gap and reduce reliance on data scientists by providing intuitive interfaces, automated data mining capabilities and collaboration.

9



Due to the multitude of data sources and information challenges, determine the business areas to be affected as well as potential data sources necessary to access. Investigate easy-to-use software to extract data from a variety of sources and in all needed formats and to clean and enrich it prior to analysis. Identify other tools and applications with which predictive analytics should be integrated and make sure the software provides an efficient method for deployment. Our research shows that organizations are planning to invest more in predictive analytics. In the modern era of information and analytics, legacy approaches can be counterproductive and put companies at a disadvantage. We conclude that it is time for companies to rethink such approaches and explore the use of modern predictive analytics tools to seize a competitive advantage.

10



About Ventana Research Ventana Research is the most authoritative and respected benchmark business technology research and advisory services firm. We provide insight and expert guidance on mainstream and disruptive technologies through a unique set of research-based offerings including benchmark research and technology evaluation assessments, education workshops and our research and advisory services, Ventana On-Demand. Our unparalleled understanding of the role of technology in optimizing business processes and performance and our best practices guidance are rooted in our rigorous research-based benchmarking of people, processes, information and technology across business and IT functions in every industry. This benchmark research plus our market coverage and in-depth knowledge of hundreds of technology providers means we can deliver education and expertise to our clients to increase the value they derive from technology investments while reducing time, cost and risk. Ventana Research provides the most comprehensive analyst and research coverage in the industry; business and IT professionals worldwide are members of our community and benefit from Ventana Research’s insights, as do highly regarded media and association partners around the globe. Our views and analyses are distributed daily through blogs and social media channels including Twitter, Facebook, LinkedIn and Google+. To learn how Ventana Research advances the maturity of organizations’ use of information and technology through benchmark research, education and advisory services, visit www.ventanaresearch.com.

11


http://twitter.com/ventanaresearch

http://www.facebook.com/group.php?gid=42346537689&ref=ts

http://www.linkedin.com/groups?gid=1625427

https://plus.google.com/107660697898618790080/posts%23107660697898618790080/posts

Documents

Predictive Analytics Enters the Mainstream · Getting Started: People For years predictive analytics has been the domain of the statistician and the data scientist, but today that