Blogs

The sky is the limit: Flexible, scalable predictive analytics in the cloud

Portfolio Marketing Manager, Predictive Analytics, IBM

Businesses today are finding increasing opportunities to integrate predictive analytics and cloud computing into their processes. These technologies each have the potential to change how organizations conduct business, and they are opening up exciting new possibilities.

When applied correctly, predictive analytics has demonstrated the capability to double return on investment (ROI). But many organizations have yet to leverage the full breadth and depth of enterprise-scale predictive analytics. Fearing large infrastructure investments, complex installation and configuration, and a very long timeline before value is realized, some organizations limit their deployment scope and, paradoxically, the potential for enhanced results. Cloud computing has opened up new opportunities for creating an enterprise-scale predictive analytics environment, and organizations no longer need to delay reaping the benefits that predictive analytics can offer.

Despite the hype that still exists for cloud-delivered processes, the cloud has actually been around as a concept since the 1960s—but it wasn’t terribly practical then. It didn’t become practical until there was a backbone that could support the processing needs of multiple requests that all demand a speedy response. The definition of speedy in this context morphed from weeks to days to hours to minutes to instant as requirements moved from real time, near real time and right time. Today, the cloud has become an important vehicle for deploying technology—not just predictive analytics, but just about any business technology.

Finding the right cloud strategy

Probably because of the hype, the word cloud is often applied without much explanation, and the customer is left to figure out what it means. In the absence of clarity, a host of misconceptions emerge: “It’s cheaper.” “It’s an either/or decision.” “It’s not secure.” These misconceptions are an attempt to lay out the practical considerations to take into account when thinking about cloud-based delivery of predictive analytics. The word practical is defined as follows:

http://www.ibmbigdatahub.com/sites/default/files/definition.png

Keep this definition in mind when considering what IBM has to offer and what you need to consider when you are thinking of mixing predictive analytics and cloud deployment:

  • When you think predictive analytics, are you more concerned with how you get there—the analysis itself—or how you benefit from it—the outcome of the analysis?
  • What parts of the analytics process—analysis, data cleaning, deep analysis, deployment, model building, model testing and so on—are you doing as part of your predictive analytics work? Which parts do you share with others? What parts do you do yourself? And what parts do you not want to do at all?
  • Who are the stakeholders that you need to take into account? Do you have a single project or process in which predictive analytics needs to play a role? Or are you looking at something broader?
  • When, where and how are your users working with the tools of predictive analytics? Online, offline or mobile? Shared, individual or in between?
  • What kind of skill set do you have in your organization for predictive analytics? On what part of the analytical process do you want your resources to spend the bulk of their time?

Taking the optimal deployment route

When considering the technologies that allow an organization to apply predictive analytics to their business, deployment options can be as varied as the predictive analytics techniques and approaches they support. Not every organization should deploy cloud analytics in the same way. There are nuances that need to be considered when deciding how the cloud fits the broader strategy:

  • Those organizations just getting started with predictive analytics are typically casual users just trying out predictive concepts. Packaged solutions powered by cloud-based delivery of predictive analytics are a great option, as long as a path beyond the initial foray into predictive analytics exists, whether on the cloud or not.
  • Those organizations looking to broaden and deepen their use of predictive analytics should look at how to do more—more techniques, more data—and start thinking about how to deploy analytics. These organizations may be looking at the ability of the cloud platform to embed predictive analytics into both software-as-a-service (SaaS) and on-premises systems easy and widespread.
  • Those organizations with an established predictive analytics factory capable of an industrial process for creating predictive analytics models, need to take advantage of the elastic compute and cloud-based data sources and big data environments that underpin cloud solutions. These organizations need to look to cloud computing as a way of broadening usage and the use cases to which predictive analytics can be applied. In the process, organizations are likely to find that different parts of the organization are at different levels of sophistication.

Selecting the right blend

An organization-wide strategy for cloud-based delivery of predictive analytics may involve all of these approaches, within different parts of the organization and at different levels of sophistication. The right blend of solutions with public, private and hybrid clouds can actually improve predictive analytics adoption across the whole organization. IBM offers practical cloud-based solutions for just about any organization or application by providing a range of approaches suited to different types of requirements: 

  • http://www.ibmbigdatahub.com/sites/default/files/skyslimit_embed.jpgBusiness users can explore and get answers to questions with IBM Watson Analytics, which provides a cloud-only tool for predictive exploration.
  • Analysts looking to build and apply sophisticated models can leverage IBM Predictive Analytics on Cloud, which makes optimizing business processes possible through cloud-based creating, testing and modeling—with no software to install, upgrade or maintain.
  • Developers interested in using the outcome of a model within an application containing multiple services and components can find the tools they need with IBM Predictive Analytics for Bluemix

These approaches are just a few of the ways organizations can take advantage of cloud-based predictive analytics. Hybrid cloud environments are also becoming quite prevalent, and they are particularly attractive to organizations that need to keep the data they analyze within their own walls. No matter what are the requirements, IBM helps organizations create a solution that can take advantage of the benefits predictive analytics offers.  

Exploring cloud-based predictive analytics deployment strategies

Attend IBM Insight 2015, 26–29 October 2015, in Las Vegas, Nevada, for a closer look at cloud solutions and how to determine the most practical approach for your organization. Jane Hendricks, technical marketing manager, IBM Predictive Analytics, at IBM, and Lance Nichols, IBM SPSS offering manager, at IBM are presenting “DPA-1990: Practical Predictive Analytics in the Cloud.” This session covers the many deployment options available for predictive analytics including on-premises, cloud-based and hybrid platforms. In addition, see how cloud-based delivery can accelerate the adoption of predictive analytics. The presentation, “DPA-3197: Predictive Analytics Platform as a Service at Boeing,” by Chris Stinnett at Boeing looks at how this IBM client has addressed the need to make analytics easy for its employees to adopt and for its IT staff to maintain. If you are attending Insight 2015, stop by the Predictive Analytics area of the expo floor.

And continue your predictive analytics journey at the IBM Analytics resource page.