Want deeper business metrics and closed-loop process optimization? Even the best BPM suites still require customization and IT hand holding.
BPM implementations aren't all that dissimilar from the summer family vacation. No matter how much time you spend planning the best cross-country route, you end up making detours that cost time and money. In the lab, our summer vacation with BPM software was no exception. More than a few times we exclaimed: "Are we there yet?"
Automating your business processes can increase productivity, but it can be counterproductive if the automation takes control away from the process owners. A good BPM suite not only involves the IT organization, but it also empowers business analysts to define, manage, analyze and optimize their processes.
Three distinct capabilities are required to create this synergy between IT and the business: modeling, analysis and simulation. We evaluated these capabilities in products from nine vendors: BlueSpring Software, Computer Associates (CA), Fuego, Lombardi Software, Pegasystems, Savvion, Tibco, Ultimus and Oracle. Every product offered basic operational reporting and analytics, but if you want deeper business activity monitoring (BAM), even the best products required IT integration and customization work. And just two products — Fuego and Savvion — excelled in closed-loop simulation for continuous process improvement, suggesting that this software category is still under construction.
Take a Snapshot
Modeling, the first step in a BPM implementation, is best performed by business process experts. Modeling involves laying out a process using a flowcharting metaphor. More often than not, eye-candy icons are used to represent steps in the process rather than the geometric shapes we all know and hate. Business process management notation (BPMN) is the proposed standard for modeling business processes, but only a few vendors (Fuego, Lombardi, Savvion and Oracle) support this blasé icon set. The modeling sets provided by BlueSpring and Ultimus had more pizzazz. CA says it will be supporting BPMN in its next release. We were most pleased with Fuego's implementation, which allowed us to "skin" our model in a number of icon sets, including BPMN, BPEL and its own proprietary notation.
Despite the eye candy, BPMN support is essential to a BPM implementation's long-term success. Adhering to standards implies the ability to export models across systems and use them in any other standards-compliant BPM solution — useful when dealing with mergers, acquisitions or even future systems replacement. So while we enjoyed the look of nonstandard icon sets, we preferred to work with standards-compliant products. Look and feel is more easily modified — as proven by Fuego's skinning techniques — than rebuilding models created with nonstandard methods.
Regardless of the product and icons used, we were able to quickly and efficiently model our business process within a few hours — including documentation and notes to would-be implementors of our process. The products all provide drag-and-drop user interfaces for creating models, and they allow properties and rules to be configured on specific activities. Most products, with the exception of Ultimus, use line connections between steps to indicate the process flow. Some products require that rules be configured on the links between steps rather than on individual steps, while others simply use the links as visual representations of the flow.
The products that configure rules on links made more sense to us and made it easier to understand the flow of the process than those products that required us to configure both pre- and post-condition rules on the step itself. While neither type of system proved more advantageous in terms of implementation, we preferred the clearer understanding we achieved by configuring rules on links, as it more closely modeled flowcharting, which most business users are comfortable with. Ultimus, however, doesn't require the use of links at all, and in most cases doesn't allow them, making visual process representation harder to understand than with other products we evaluated.
We were fascinated with the decision to include or exclude "swimlanes" from particular products. Swimlanes provide a visual representation of responsibility, with all tasks assigned to a specific group of individuals — such as departments or teams — arranged within designated swimlanes. This feature is useful during implementation, especially in shared-repository environments, where the same models created by business analysts are used by IT staff to implement the services behind the models. In this type of environment, the information about swimlane participants is configured only once and then applied to every activity placed into that swimlane. Only Savvion, BlueSpring and Lombardi provided this functionality. While CA supports swimlanes, it does so only for visual organization and documentation purposes. Even though we could drop activities into a specific lane, we still had to configure that activity in terms of what group of people or systems was responsible. We preferred the systems that allowed us to configure once and then reuse responsibilities for a given swimlane.
Once the process was modeled, in all cases it was necessary to switch drivers and let IT handle the implementation. It isn't possible, yet, for process owners to implement processes themselves, even if the services representing specific activities exist. Vendors such as Fuego and Pegasystems say that business process experts ought to be able to implement without IT assistance, but we find that's about as likely to happen as getting through Chicago without hitting traffic.
One of the benefits of BPM is the ability to analyze key performance indicators (KPIs) and adjust in-flight transactions, resources and the underlying processes. This is a significant component of optimization, as it enables process experts to pinpoint bottlenecks in the process across a variety of metrics, including process execution times, resource cost and utilization, as well as business indicators such as product volumes and costing data. Reporting on both operational and business metrics is a key component of the ROI equation used to justify the implementation costs of a BPM solution.
Operational reporting is generally built into systems and enabled as part of the deployment. Process and activity metrics are collected automatically by the process engines and stored in the appropriate repositories. This data is then available for reporting and analysis by third-party products or available within canned reports that can be viewed directly or shared through bundled portlets or interfaces to portal environments. With the exception of the Pegasystems and Oracle products, every product we tested included this base data and provided report and portal views of operational statistics. The Pegasystems and Oracle products collect the statistics but don't provide an interface or canned reports for viewing data. Pegasystems specifically requires SQL analysis services for BAM (although it's a J2EE application that can take advantage of Oracle, SQL Server and DB2 as its repository).
Business metrics are reported in all products by defining process variables. Process variables are snippets of data carried through the process that might be pulled from a database, such as the cost of a product, or data that has been input by the initiator of the process, such as the total value of a purchase order or demographic information. This data is not automatically tracked by any BPM implementation; it must be specifically marked within the process implementation as a tracked variable. This is another one of those detours into IT we experienced, as there is no mechanism for the business process owner to specify which variables should be tracked. Business process experts can take advantage of the capabilities offered by all modeling products to add notes on what type of data should be collected, but we imagine there will be some lively discussion between those experts and IT as they try to synchronize what "value of the purchase order" means in terms of data variables. Once the process engine knows to keep track of the data, it will store it in the repository along with operational statistics. This is the information the business needs to make real-time or at least near-time decisions.
Tracked process variables support the creation of KPIs based on business data. In turn, these can be complied into online analytic processing (OLAP) cubes for deeper analysis through third-party BI tools. The ability to access this data directly varied, with Tibco requiring an ETL process and Ultimus requiring integration with its proprietary events subscription interface. Most of the products, including those from Oracle, Savvion and Lombardi, allow KPIs to be accessed via Web services and integrated into the process.
CA was a standout in reporting, offering standard KPIs and the ability to configure a variety of reports. But overall, we discovered that business process data is not, out of the box, available for real-time analysis unless your BI developers have some dashboard-building experience. While most products let you view and even drill down into OLAP cubes, none of the products offered us a mechanism for specifying real-time aggregate data without breaking out the development tools. Fuego, for example, gives business users a view of data through predefined cubes while BI developers are creating more detailed, user-specific dimensions. Oracle is working on a more business-user-friendly reporting environment, but that system wasn't available for review in the current release. Other vendors simply haven't gotten around to worrying about this capability, choosing instead to focus on their products' ability to integrate easily into a service-oriented architecture (SOA) and make modeling painless for business process owners.