By Cassandra Balentine
Advanced analytics enable businesses to leverage business data to provide insights on everything from customers to business processes. With this information, they are able to make intelligent business decisions, detect fraud, and calculate risk.
In its Forrester Wave: Big Data Predictive Analytics Solutions, Q2 2015 report, Forrester Research, Inc., identified predictive analytic solution market share leaders as IBM, SAP, and SAS, with strong performance from a variety of additional vendors including Angoss, Aleryx, Alpine Data Labs, Dell, FICO, Oracle, and RapidMiner.
In its report, Forrester pegs predictive analytics as a business “game changer” that is more relevant and easier to implement. Done right, it enables businesses to win, retain, and better serve customers. Enterprises seek advanced analytic capabilities to provide direct insights about customers and business processes; achieve intelligent, adaptable customer interactions and business processes; and improve customer engagement.
Chandran Saravana, senior director, advanced analytics, SAP, says industries including retail, consumer goods, telecommunications, oil and gas, sports and entertainment, banking, healthcare, insurance, and utilities and mining are leading adoption. In terms of line of business, top embracers are marketing, supply chain and operations, finance, and human resources.
When done properly, advanced analytics have applicability to just about any company, of any size, across any vertical, shares Shawn Rogers, chief research officer, Dell Statistica. He emphasizes that it is not really about how big your company is, how much data you have, or the vertical you operate in—it’s about understanding data to improve the way you do business.
Here, market share leaders as well as other predictive analytic solution vendors share their thoughts on understanding predictive analytics, adoption challenges, and the technology’s future landscape.
Understanding Predictive Analytics
Predictive analytic tools utilize algorithms to detect data patterns that help anticipate outcomes. In its report, Forrester provides the example of using predictive analytics to find a model that predicts which customers are most likely to churn. However, they caution users to realize that these models are not meant to be a one-time deal; businesses should continue to run data to ensure the best predictability.
Forrester identifies the predictive analytics lifecycle in a series of steps—identify data from a variety of sources, wrangle the data, build a predictive model, evaluate the model’s effectiveness and accuracy, use the model to deliver actionable prescriptions to your business peers, and monitor and improve the effectiveness of the model.
In addition to understanding the necessary steps, businesses have a lot to consider when implementing predictive modeling within their organizations.
“Everything starts with use cases,” states Saravana. “Find the compelling use case where predictive analytics is applied. Access the current analytics capabilities within your enterprise that include both technology, environment, and skills. Determine how to effectively manage all of your data and start small with one or two use cases that have immediate impact on your organization.”
Organizations should consider their business needs first and technology implementation second, recommends Rogers. “Though it’s powered by technology, advanced analytics is a business investment,” he stresses. As such, it is important to first understand the needs of your business. Determine which questions your business could benefit from answering. What current processes and procedures could you potentially improve upon if you better understood data?
A common misconception is that predictive analytics has to be applied to something new or something that’s not working. “That’s not at all the case,” says Rogers. Most companies already have mechanisms in place that may not be new and may not even be broken, but that can still benefit from the application of predictive analytics. So, before exploring the possibilities, it’s critical to first understand your business needs and how and where understanding data helps meet those needs.” says Rogers.
Sascha Schubert, analytics marketing director, SAS, says organizations should understand that predictive analytics is more than applying complex statistical algorithms to data. “Making predictive analytics a driver for successful business not only requires the right technology, but the right process, people, and organizational culture.”
The success of predictive analytics is measured by the business value it generates. Schubert says many organizations underestimate the challenges related to implementing predictive analytics into business processes and the change it may trigger. For example, a marketing organization might use predictive analytics to select customers to call for a marketing campaign. However, because the marketing business users have not been trained in using the results from a predictive model or do not trust the information, they might not use the recommendations. To address this, the implementation of predictive models in business processes should be accompanied with clear metrics on how to measure the value of the models. Moreover, regularized routine deployment of predictive analytics into operations requires a disciplined and managed process that is governed, has workflow, and auditable versioning.
Marc Andrews, VP, industry analytic solutions, IBM, suggests that there is no one-size-fits-all model when it comes to predictive analytics. “We believe organizations need tools that minimize effort and accelerate time-to-insights and action. Taking advantage of pre-built solutions allows business users to gain direct access to predictive insights about their customers, assets, finances, and risks—all of which are specifically tailored to address distinct use cases, challenges, and pain points within their industry. Putting powerful analytics in the hands of every business user ensures that businesses can more easily uncover patterns in data, pursue ideas, and improve all types of decisions impacting their industry.”
Adoption Challenges
Although the benefits of predictive analytics are obvious, it is an investment in time and resources to implement these processes. There are quite a few roadblocks that keep them from being implemented by the masses.
Cultural changes are necessary to become a data-driven business. The insights uncovered through an analytics initiative are only as good as the willingness of company leadership to act on those insights. “So, it’s important to have sponsorship and support from the top levels of the organization before diving deep into the given project,” says Rogers.
He also sees a lack of openness and flexibility in the overall data environment as a hindrance to adoption. “With the proliferation of new data sources and edge devices, the data environment has become exceedingly complex and many companies don’t have the mechanisms in place to facilitate the type of openness and flexibility needed to combat that complexity,” he says. “So, while they can govern and manage data within the confines of a single given platform, most companies struggle with the need to manage beyond that platform,” he adds. In order to distribute analytics through a complex ecosystem, you need a flexible and agile approach.
There is also a skill gap issue. “Enterprises have difficulty bridging the gap between their domain knowledge experts and analytics experts,” says Saravana. “On one hand you have domain knowledge experts with a deep understanding of business, and on the other you have analytical experts with a deep understanding of analytical techniques. There is a tremendous need for tools and platforms that can bridge the gap between domain knowledge experts, analytical experts, and business users.”
Lou Bajuk-Yorgan, senior director, product management, TIBCO Spotfire, argues that while many firms believe that predictive analytics is for statistical experts and is difficult to implement, modern platforms make it so it doesn’t have to be that way. “While having data scientists is definitely a plus when a firm wants to implement predictive analytics, the truth is that this is not a necessity for adoption.”
Before determining the challenges of predictive analytic adoption, Schubert recommends that businesses consider three areas of the application.
The first is making the necessary data available to the resources tasked with building the models. A large portion of the time for a predictive analytics project is spent managing and preparing the data, which resides in silos in many organizations and is not in good shape for fast and easy application of predictive analytic algorithms. “The growth of data volumes and data diversity adds to the complexity of the data preparation stage. Predictive analytics is an iterative approach in which continuous access to data that could be relevant can become vital to the success of the project,” says Schubert.
The second area is enabling quick and easy discovery of insight from data. Here, experimentation and self-service capabilities are important. “The analytics resource should be able to try different approaches to explore the data and develop the models. The access to the analytical capabilities also needs to match the skill level of the resource, a data scientist works differently from a business analyst,” stresses Schubert.
Finally, an organization’s third challenge is taking a serious look at the deployment to predictive analytics from the outset of the project. Determine the goal of the model, how it will be implemented, and how it will be used. Will the implementation of the predictive model require any changes in the current business processes? How will the organization measure the value generated by the predictive model? “Many organizations struggle today with the deployment of predictive models into operational processes as the deployment steps are often not clearly defined, highly manual, and seldom automated when the organization embarks on a predictive analytics project,” says Schubert.
Andrews suggests that implementation times and an organization’s ability to see results from predictive analytics right away largely depend on data availability.
Bajuk-Yorgan adds that companies who have not started collecting data are too immature for predictive analytics to be relevant for them. “They need to understand that capturing data around processes and events that they want to improve is the first piece of the journey towards harnessing value from data using predictive anatlyics.”
Future Landscape
Business intelligence (BI) tools, including advanced analytics, are developing and evolving at a rapid pace. Therefore, the landscape for predictive analytics today is much different from what will happen in three to five years.
Rogers explains that a paradigm shift is occurring. Instead of focusing on pushing data to the analytics, companies focus on pushing analytics to the data. “We’re already seeing this analytics-at-the-edge concept gain great traction today, and we feel strongly that within the next three to five years it will become the new normal,” he offers.
It makes sense for companies to adopt this approach. “As technology has improved, companies have become better and smarter about putting the right data on the right platform for the right reason,” explains Rogers. “That’s greatly reduced the need to pull data into a centralized data warehouse or analytics sandbox. Since the right data is already in the right place, we can apply predictive models and run analytics directly at the source, taking advantage of the compute power on that system. This not only estimates the time and expense required to transport data, but it enables immediate action to be taken in response to insights. So, we’re expending more companies to push in this direction and they’re going to be actively looking for technology solutions that help them get there.”
In the next few years, Saravana sees predictive automation capabilities growing to reach the highest level of maturity, the proliferation of predictive use cases and models, and more canned predictive applications in the cloud, addressing specific use cases.
Schubert foresees predictive analytics becoming embedded in many more applications, either based on proprietary algorithms or open source.
“Niche vendors will provide very specialized analytical applications on data created by the Internet of Things (IoT). Operational processes using analytics will provide more guided automated workflows for faster, quickly adaptive, and more democratized use of analytics,” says Schubert.
Additionally, distributed processing and storage platforms based on Hadoop are expected to provide the big data analytics innovation platforms of the future, where Schubert says organizations test new use cases for predictive analytics. “Predictive analytics platforms will integrate deeper into the data sources, including operational and even creation and the edge of the IoT networks.”
Bajuk-Yorgan says in taking advantage of the ongoing big data phenomenon, analytics is the next logical step to generate insights and monetize opportunities towards business advantage. “Predictive analytics is an integral part of analytics workflow for appropriate use cases. We believe that the predictive analytic landscape will become even more crowded as analytics firms with niche solutions built around predictive analytics come into existence.”
Market Solutions
Several vendors offer predictive analytic platforms to help enterprises glean and leverage their business data. Here are a few highlights from vendors featured in this article.
Dell Statistica delivers a full range of data blending, data discovery, and advanced and predictive analytics tools that help organizations tap into data to predict future trends, identify new customers and sales opportunities, explore what-if scenarios, and reduce the occurrence of fraud and other business risks. The solution features the newly added Native Distributed Analytics Architecture capability, which pushes predictive algorithm model-building and scoring functionality directly into the data source.
IBM’s Industry Analytics Solutions allow organizations of all sizes across industries such as retail, banking, telecommunications, and insurance to get answers to critical business questions, like how important is it to ensure certain items are always in stock, how many different product combinations should be offered in stores, are people who spend a lot dining out after 8:00 p.m. more likely to overdraft, or when should production of an oil well stop to pull pumps out of the ground for maintenance.
These predictive analytics solutions are pre-built for specific industry use cases, suggests Andrews. “Utilizing our deep industry expertise, we have pre-built predictive analytics, data preparation, and user dashboards. Each analytic solution is designed to provide businesses with predictive insights that will deliver a better understanding of customers, assets, and operations so they can take action and improve profits.”
SAP Predictive Analytics provides an integrated approach to predictive analytics that enable organizations to uncover trends and patterns from big data, the IoT, and existing SAP and third-party data sources. The solution is designed to empower a broad spectrum of users to spot opportunities in real time, make fast and accurate predictions, and act with confidence at the point of decision.
SAS provides specific analytics-based solutions for predictive analytics targeted at diverse business problems, including risk management, supply chain, customer behavior models, and fraud detection, some of them for vertical industries. SAS helps to implement data-driven decision processes in almost every industry with more emphasis on financial and communications industries.
The company also provides an integrated predictive analytics platform with data analytics products for data mining, text analytics, forecasting, optimization, simulation, machine learning, and streaming analytics.
TIBCO Spotfire is a predictive analytics platform that takes the approach of providing advanced capabilities in an easy interface for analysts. Analytic workflows consist of steps involving data collection, data preparation and exploration, building models, testing models, and iterating to improve. “There could be complexities at each step of this analytic process. It has been the focus of the Spotfire platform to enable this entire iterative workflow, providing access to a variety of data sources, data preparation using advanced R scripts, building predictive models including R scripts, and presenting results in interactive dashboards for communication,” says Bajuk-Yorgan.
The Business of Data
BI tools provide businesses with useful mechanisms to gain direct insights on customers and business processes, enable adaptable customer interactions, and better prepare for the future. While popular applications for the technology include marketing, supply chain and operations, finance, and human resources, the potential for predictive analytics is broad, spanning all industries and business cases. SW
Feb2016, Software Magazine
