Continue in 2 seconds

Using Advanced Analytics to Drive Better Business Decisions

By
  • Peter Arena, Stephen Rhody, Michael Stavrianos
Published
  • April 01 2005, 1:00am EST
More in

The business intelligence (BI) market is moving toward widespread adoption of advanced analytics to drive better decision management. Increasingly, BI practitioners and consumers are recognizing that advanced analytics are essential to the delivery of high-value, decision-oriented functionality (e.g., forecasting, prediction, simulation and optimization). Meanwhile, early disciples are discovering that they can build a sustainable competitive advantage through adaptive and embedded analytics.

The arrival of this market trend is heralded by incisive headlines across the major trade journals and reinforced by industry analysts whose optimistic forecasts suggest that advanced analytics will be a driving force in the BI market for some time to come:

  • "Advanced analytics tools ... will become more relevant in an organization's daily operations as the opportunity to embed such tools into analytic applications increases." (IDC, 2003)
  • "The market for packaged analytics appears to have a bright future. ... About one-third [of survey respondents] plan to add more sophisticated statistical analysis and predictive modeling capabilities in the next 18 months." (TDWI, 2002)
  • "Worldwide services for the analytics market are projected to grow from $16.8B in 2001 to $29B in 2006, an 11.6 percent CAGR." (Gartner, 2001)

This article looks beyond the sound bites to examine the real value of advanced analytics, as well as the unique challenges of integrating decision management into the BI architecture. We evaluate three competing approaches that vendors are using to meet the growing demand for analytic applications, and we offer advice to decision-makers who are exploring this investment. Finally, we articulate five strategic imperatives for successful implementation of advanced analytic applications.

Value Proposition

The driving force behind increased adoption of advanced analytics is a familiar one. Organizations that have embraced each progressive phase of data management modernization (enterprise resource planning, customer relationship management, data warehousing, business intelligence) are still looking to capitalize on their past investments. As they reach maturity in each discipline, they are rewarded only with promises about the watershed ROI offered by the next phase (in this case, decision management). Jaded readers can't be blamed for thinking they've heard this story before; but this time, it may actually be true.

Why now? Because previous phases of the modernization process have focused on tasks that are inherently foundational to better enterprise performance - data capture, data architecture and information delivery. Granted, the industry's emphasis on online analytical processing (OLAP) and other reporting disciplines as the cornerstone of BI has hastened the democratization of information assets, but it has done little to empower information consumers. Faced with a downtick in a critical business metric, decision-makers still must rely heavily on intuition and instinct in choosing how to react.

The promise of embedded analytics is to do for decision making what OLAP has done for information -- that is, to democratize decision-making assets. Believers describe it in terms of a knowledge management initiative, in which the organization's best practices for each decision-making process are pushed to the desktops of end users as embedded logic within analytic applications. These applications are typically powered either by business rules engines (which apply logical conditions to determine how a certain case should be handled) or predictive models (which probabilistically identify the action most likely to achieve the desired results).

Democratization and automation of decision making has the potential to unlock numerous benefits. At a strategic level, it helps executives make more informed decisions on fundamental business questions, such as:

  • Who will be our most profitable customers tomorrow?
  • Do recent purchasing patterns represent the start of a long-term trend, cyclical behavior or just a short-term aberration?
  • How would a price change influence the behavior of various customer segments?
  • What will be the impact on profits of introducing a new product line?
  • Where should we focus our expansion efforts?

Reporting tools alone are unable to answer these questions. Consider, for example, an executive dashboard report that shows an overall customer attrition rate of 15 percent. Even if the executive is able to drill down by quarter, region or customer demographic, more fundamental questions remain. Why is the attrition rate 15 percent? What is causing attrition for key customer segments? What actions can be taken to reduce attrition?
Business intelligence tools that merely report historical data are ill equipped to guide prospective decisions. Armed with the right tools, however, leaders can forecast and evaluate the impact of their decisions before executing them. This capability is particularly important when facing multiple (often competing) strategic objectives - and when complex or interrelated business processes create the risk of unintended consequences to a decision.

At an operational level, decision-making assets can eliminate the need for human involvement in decisions that are either highly routinized (such as reordering inventory), or so numerous that they prohibit case-by-case oversight (such as identifying fraudulent financial transactions). They can also enable instantaneous decision making on fleeting business opportunities, such as informed customization of up-sell or cross-sell offers during a customer-initiated contact.

Whether aimed at strategic or operational issues, advanced analytic applications can dramatically improve the consistency and quality of decision making throughout an organization -- enabling decisions that are based on quantified costs, benefits and risks, rather than pure instinct. At the same time, they accelerate the decision-making process, allowing managers to focus their time and attention on more pressing issues.

The bottom line, of course, is the bottom line - and the evidence to date suggests that decision-oriented analytic applications offer a superior return on investment compared to their non-decision-oriented counterparts. A recent study conducted by IDC, based on case studies of 40 analytic projects, found that projects involving predictive analytics offered a median ROI of 145 percent, compared with 89 percent for their non-predictive counterparts. IDC attributed the superior ROI of predictive analytics to enhanced operational processes, enabled through improved decision making. Undoubtedly, a related factor in the value of predictive analytics is that it is typically brought to bear on high-impact business processes where executives see the greatest opportunities for ROI.

Unique Challenges

To support the growth of decision management, a new breed of analytic applications is needed - tools that place decision-making assets directly in the hands of end users. The market is moving aggressively to meet this need, but the broad functional requirements surrounding these applications present a number of unique challenges:

Diverse Problem Set. The broad and diverse range of business processes that can be improved using advanced analytics is both an opportunity and a challenge. Unlike OLAP, with its homogeneous slice-and-dice reporting functionality, no two decision support tools are quite the same. While certain decision processes have common elements (e.g., within an industry vertical), there are unique aspects to each application, and customization is essential.

Specialized Functionality. A corollary of the first point is that the number of users for a given application tends to be small. Typically, decision-making authority is concentrated in the hands of selected managers. Even within a given business process, different decision-makers may require different viewpoints on the same overall process.

Highly Technical Algorithms. Advanced analytic applications are difficult for nontechnical users to develop and maintain, even with the most user-friendly development platform. Effective design of user interfaces can shield analysts from much of this complexity, but end users must still be able to interact with the application to simulate and evaluate alternative decisions.

Advanced Information Delivery. In order to support decision making, advanced analytic tools must be capable of intelligent, interpretive information delivery. Results presented through arcane statistical metrics are not actionable. Even in cases where the application proactively executes the optimal decision, analysts must still receive customized diagnostic results to ensure quality control.

Evolving Requirements. Decision support tools need to evolve and adapt as: 1) new data becomes available, 2) improved analytic methods or business rules are identified and 3) business processes change. Unlike traditional BI tools, which can be updated by adding a new report template, tuning-up an advanced analytic application can be as complicated as building it from scratch.

Market Behavior and Consumer Options

Adoption of decision-oriented analytics is nearing a take-off point. The success or failure of this next stage in the evolution of enterprise data management will hinge on decisions that are made during the design of each analytic application - decisions that will shape their ability to meet the challenges outlined previously. How is the market meeting these challenges? Typically, with one of three development methods: 1) packaged analytic applications (industry verticals), 2) analytic application development platforms or 3) customized application development. Each has its advantages and disadvantages.

Packaged analytic applications are relatively easy to use and often allow buyers to secure core functionality without having to reinvent the wheel. In most cases, however, they are unable to support all of an organization's decision-making needs - at least not directly off the shelf. Application development platforms allow greater customization and also offer tools and templates that accelerate the application development process. Ultimately, however, the quality of the production application depends on the capability of the developer. The customized "build-from-scratch" approach often yields the most innovative solutions - unconstrained by the functional limitations of development platforms - but can result in redundancy or inconsistency if multiple applications are developed for use in different parts of an organization.

In order to navigate these options and get the most from their investment, buyers should assess the decision-making environment within their organization. Which business processes would benefit most from decision-making assets? How many distinct applications are needed? Will they be deployed to users across the enterprise or concentrated in the hands of a select few? What analytic resources are available within the organization? How will responsibility for application development be shared by IT and line-of-business stakeholders?

Buyers who need several distinct applications that cut across business functions will benefit from the accelerators and integration tools offered by application platforms. Those with more "commoditized" decisioning needs - as in financial analytics and CRM campaign management - may do well with prepackaged industry verticals. Organizations targeting a small number of business processes that involve unique or highly complex decisions will do best with customized applications.

In many cases, a combination of tools will best meet decision-making needs across the enterprise. For example, vertical applications with only modest customization may prove adequate for generic business functions, such as workload optimization, financial analytics and campaign management. Meanwhile, mission-critical functionality that is highly specific to the business can be developed through highly customized solutions. This two-tiered approach to business analytics borrows a page from the management playbook of Peter Drucker, relegating "back office" analytics to packaged applications that represent the "front office" of vertical developers.

Gaining Competitive Advantage

Ultimately, the greatest competitive advantage will go to those organizations that can most effectively utilize advanced analytics to drive better decisions. We see four ways that firms can differentiate themselves from their competitors in this regard:

Plan. Organizations that demonstrate vision in deploying their analytic applications will gain an immediate advantage. For starters, this means identifying candidate business processes that are most likely to benefit from the integration of advanced analytics. It also means aligning key stakeholders behind a shared set of objectives, while recognizing their diverse objectives and constraints.

Act. Early adopters can gain an advantage over the competition by being first to market. Not only will they reap the benefits of better decision making more quickly, but they can get a jump start on the road to maturity in decision management.

Build. More experienced practitioners can build on early successes to drive improved decision making throughout the organization, taking advantage of economies of scale as they move toward more advanced applications.

Innovate. Organizations that are willing to pursue customized applications can gain a decisive edge by finding new and innovative ways to apply the principles of decision management.

Opportunities in this last category typically involve either highly complex business decisions (where solutions involve multilayered business rules, cutting-edge analytics or interactive analysis) or unique data sources that confer a competitive advantage (e.g., time-series panels, unstructured data, proprietary surveys). Two specific areas that we see as fertile ground for innovative decision tools are: 1) workforce analytics - driving better hiring decisions through contextual evaluation of resumes relative to job roles and historical performance metrics and 2) strategic enrollment management - using characteristics of student applicants to make better decisions with respect to enrollment, retention and alumni relations.

Organizations that are charting a course toward better decision management can learn from the experiences of pioneers in this field. The road ahead presents many challenges, but the destination offers valuable rewards. At the end of the day, executives can set strategy, identify key business drivers and gain a more thorough understanding of the effects of strategic decisions across the organization. Line-of-business managers can align their daily decisions with changing market conditions, customer behavior and actions of competitors. Information consumers at every level of the organization can understand why they are winning or losing business and predict how their actions will influence business results.

Regardless of whether your destination is a single customized application or an enterprise-wide decision-making infrastructure, we offer the following as strategic imperatives for your journey:

Teamwork. Build a multidisciplinary team that understands the analytic techniques, the underlying business processes and the specific end-user requirements. All perspectives are needed to guide development of an analytic application and to ensure that results are interpretable and actionable.

Targeted Solutions. Applications should be focused in their objective, aimed at guiding decisions around a specific business process. Start with processes that offer the biggest bang for the buck - typically those that involve high-frequency operational decisions.

Embedded Solutions. Provide contextual, business-oriented decision support so that the analytics are transparent to the end user. Embedded analytics make it easy for knowledge-workers to understand and interpret information as part of their day-to-day routine. The goal here is to democratize decision-making assets, enabling decisions that are more timely and better aligned with organizational objectives.

Enterprise Integration. Leverage the existing enterprise architecture for data integration, data quality and data management. Functionality will vary across applications, but the underlying data should be consistent.

Adaptive Analytics. Build solutions that can evolve (and learn!) as you discover better ways to model decisions. Avoid hard-coding static algorithms into operational processes. Instead, build closed-loop analytics directly into your decision-support applications. As new data becomes available, you will move quickly from ETL, to analysis, to modeling, to production - all in a seamless, self-documenting process.


Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access