The promise of building a dynamic organization has been as elusive as the promise of actionable information on every manager's desktop. The truth is that few companies are capable of getting the right information to decision-makers. And few software companies in the business performance management (BPM) space have been able to help them do so; instead, most of these solutions rely on "rear-view mirror" reporting and analysis. However, BPM systems that include predictive analytics help create dynamic and truly predictable enterprises.
A decade ago, business intelligence (BI) software provided a select group of business users with stand-alone data query, reporting, and analysis capabilities. But its focus on historical data handicapped the software's ability to improve business decision-making in the present. In addition, traditional BI applications isolated performance data in the hands of the organization's financial analysts. Thus, they were inherently flawed in their ability to collect, aggregate, and validate data on an enterprise scale; only power users had the access, time, and expertise to make them work.
The first generation of BPM software, viewed by many as the evolution or strategic application of traditional business intelligence tools and technologies, advanced organizations toward a truer sense of performance intelligence on two fronts. First, it brought analytics to the masses; the BPM vision included in its scope everyone from the CEO to the line-of-business manager. And second, BPM vendors emphasized the need for real-time information, stressing the limitations of viewing performance information only in the past tense. BPM became effective at centralizing data and automating the routine yet time-consuming tasks around data collection, validation, and manipulation. This ensured not only that information was accurate, but also that it was available quickly companywide.
BPM Gets Predictive
Although the move from BI to BPM represented a strong shift in the right direction, it failed to give decision-makers everything they needed in order to understand and effectively forecast their financial and operational performance. This is why a number of BPM vendors have begun talking about the potential of "predictive analytics."
Whereas traditional BI tools forced users to spend too much time on data collection, aggregation, and the like, BPM systems are sometimes accused of providing too much data. The age-old problem of analysis paralysis can hamstring users. The idea of predictive analytics, when applied to BPM, is to help users overcome information overload by accelerating the analysis process. An automated "agent" can dynamically alert decision-makers to problems and opportunities.
Like standard BPM solutions, predictive BPM tools address both historical and real-time concerns, but they add to the mix the ability for business managers to be proactive. They look to the future, while accounting for the present, and automate time-consuming processes including data collection, aggregation, analysis, and the identification of variances. They can also suggest courses of action to the business decision-makers. Exhibit 1, below, shows how the scope of BPM systems can expand to include predictive analytics.
A majority of BPM applications offer drilldown capabilities for analyzing performance variances and their underlying causes, but the analysis is typically manual, time-consuming, reactive, and cumbersome. Valuable time can elapse between the variance reporting (what happened), variance analysis (why it happened), and final determination of the appropriate action for proceeding. Conversely, predictive BPM software eliminates this time delay by automating the discovery process. It offers dashboards in which variances are reported automatically. And for variances falling outside acceptable limits, predictive BPM can provide a root-cause analysis and an explanation of the context in which the variance occurred. For example, if performance on a key business metric such as days sales outstanding (DSO) falls outside of specified parameters, predictive BPM can automatically alert decision-makers and provide them with the reasons, root causes, and context behind the variance.
Predictive BPM in Context
How does it work? Using OutlookSoft's solution as an example, the software automatically scans its data sets for the reasons behind successes and failures, based on the data hierarchies that are set up during implementation. The reasons typically reflect the component parts of the business, such as shifts in product sales, underperforming geographies, rising raw material costs, or other factors. If corporate net income comes in below expectations, the predictive analytics segment of a BPM application may scan the numbers of all the company's business units to discover which fell furthest below budget. In this way, predictive analytics is able to automate the drilldown process that is usually a manual function for BPM software users.
Root-cause analysis leverages a BPM system's access to external information to further explain the reasons behind a variance. Predictive analytics technologies delve into transactional and operational data outside of the performance management application, looking for patterns within that data which correlate to patterns in the organization's KPIs. For example, a leading-edge predictive BPM system might be able to connect shortfalls in the company's product-quality KPIs with shop floor accidents and absenteeism, problems revealed through the automatic analysis of the detailed information in the company's operational systems.
A predictive BPM system should also include automated context searches, which further help end users understand variances. These could involve searches of the company's supporting documentation, or unstructured data, for contextually relevant information. Just as Google might index a company's Web site, a predictive BPM system can search the corporate network for relevant information within unstructured files such as Word documents, spreadsheets, PowerPoint presentations, and e-mails. Most explanatory information within a company is stored in these types of documents -- in fact, according to Gartner, upwards of 80 percent -- and sophisticated BPM solutions can identify those that relate to a particular KPI variance and leverage that information to offer an in-depth explanation of the given variance.
BPM systems that incorporate predictive analytics technologies can also report on KPI risks and the probable impact of potential events. They can forecast performance variances and present probable reasons for the deviations, along with ratings of the likelihood that the variance will occur again in the future. Advanced statistical techniques anticipate changes in higher-level KPIs by analyzing base component metrics, serving as an early warning system for business managers.
As predictive BPM software matures, it will track not only the relationships of KPIs with other metrics, but also the connections between KPIs and business-process events. It will identify the process steps driving KPIs that are at risk and then notify managers about which ones need to be addressed to improve each metric. Even more important, the history of events, KPI values, and actions taken to influence them will accumulate within the performance management system. Eventually, when the BPM application identifies a KPI that it deems to be at risk, it will be able to present business managers with a description of the action taken in previous circumstances when the same KPI was at risk.
Predictive BPM software also can make recommendations for action based on the predicted variance and outcome. Some solutions have this recommendation logic programmed in, and suggestions for action are delivered to managers based on the nature of the variance. For example, suppose the metric for DSO comes in at an average of 90 days, versus an expected 75 days, for a certain business unit. At the same time that it automatically alerts the affected business manager to this variance from plan and identifies the root cause for the variance (perhaps a particular customer or category of customer is driving the variance), a predictive BPM solution could also recommend a course for action. It might suggest that the business manager follow a specific, predefined procedure for collecting monies owed. The advantage of this approach, versus training each employee on the company's recommended courses of action, is that the software ensures that every manager is on the same page; it automatically reinforces corporate policies and standard operating procedures.
Practical Applications for Predictive Analytics
Enterprises across a broad array of industries -- including manufacturing, financial services, retail, and technology companies -- are using predictive capabilities within performance management software to forecast product sales, plan seasonal campaigns, analyze pricing scenarios (including the potential impact of competitor price changes), and evaluate the effectiveness of their channels.
One example of a real-world company using predictive BPM is a major broadcast media organization that has implemented OutlookSoft's Insight application to improve its ability to analyze variances quickly. The setup includes a predictive analytics dashboard that focuses on costs, OIBDA (operating income before depreciation and amortization -- a form of EBITDA), and sales performance. Using the predictive BPM solution, the media company can quickly identify the top three reasons behind any major variance in costs, sales, or OIBDA. Moreover, the software focuses executives' attention on the specific media outlets responsible for each variance, further streamlining the alert/decision/action process.
Often software isn't needed to determine the primary reason behind a performance shortfall -- for example, the reason the New Orleans market struggled last fall. However, managers are finding that secondary or tertiary events impacting performance can be far less obvious. In many cases, these factors would take far longer for financial analysts or business managers to uncover manually. In addition, because this company's system leverages statistical analysis techniques, including linear regression, some of the factors it uncovers as having an impact on performance could have been easily missed, or even noticed and subsequently dismissed, if the company relied on manual analysis.
The Next Generation of BPM
As companies continue to demand more from their performance management systems, predictive analytics capabilities will play an increasingly important role in decision-making by shaping strategic and tactical planning. In a recent survey conducted jointly by OutlookSoft and BPM Magazine, 87 percent of respondents said that they think predictive analytics is important to the budgeting and planning process, but only 17 percent are employing technologies that include the capability. Fortunately for the majority of survey respondents, a handful of business performance management solutions have begun to incorporate and leverage predictive technology.
The goal of business performance management is to help decision-makers better manage, plan, understand, and leverage their performance. Predictive analytics is a natural complement to traditional BPM software and processes. It provides information about what, why, and how that helps companies understand their performance trends, anticipate future business performance, and recommend specific actions. The new generation of BPM software can make any enterprise a predictable enterprise. Now, that's a giant step forward for performance.
Christian Gheorghe, senior vice president and CTO of SAP, leads the company's alignment, development, and deployment of SOA-enabled services for business user applications.