There is a lot of discussion these days about running business "by the numbers." Writers such as Tom Davenport and Jeanne Harris have analyzed companies that are beginning to use analytics as a form of competitive advantage. Sometimes, he has found, this kind of approach is described in terms of ensuring that decision-makers have lots of data and analysis before they make a decision.
Davenport, however, has a more precise definition, one he shared when talking with Jim Ericson: "The analytic competition I think of is primarily based on process optimization, doing business better than anybody else."1 This means I need to run my overall business by the numbers, as well as my operational processes. If I can, run my processes more precisely and consistently by using my data to improve the way I execute my high-volume, operational processes, then I should execute better than my competitors. And execution, as Larry Bossidy and Ram Charan pointed out, is " the missing link between aspirations and results."2
Given that we need to have "sufficient volumes of high quality data"3 to consider using an analytic approach, we can focus on automated processes, as only the use of IT to automate business processes will generate large amounts of data. Today, companies typically implement one of three methods to automate a process - they use custom code, an application suite, or a business process management system (BPMS). In general, analytics get applied to this environment in one of two ways.
Most companies have a business intelligence (BI) infrastructure in place to report on and analyze historical data and trends. If they are really lucky, this information is presented in an accurate and timely manner. Some are even starting to give business owners and other knowledge workers predictions of likely future behavior and results. This kind of BI might be delivered as dashboards or reports, and might be entirely driven by the business user, by specialized analysts, or by some mixture of the two.
Where a BPMS or a process component of an application suite is being used, additional process analytics and reporting might also be available. These use similar analytic techniques and BI infrastructure but focus on the actual execution of the process (which steps were executed, how often and how fast). The intent is to then use this kind of analysis to improve the process by evaluating its flow either manually or automatically and changing the process to eliminate bottlenecks and so forth.
We are not just talking about analytically improving processes, but also analytically driving them. An analytically driven process has a couple of characteristics.
- It is transaction-centric in that the information about an incoming transaction and the information that can be linked to or derived from that transaction, is used to determine what happens. Instead of the process definition (do this, then do this, then check that and pick one of these two options) driving execution, the requirements of the transaction do. You have now inverted the process - it flows from the customer to the organization. An example of this would be an origination process where the data entered by the customer is used to drive models and rules that determine which products are available and what additional data and steps are required. Gartner has called this analytic-process controlling.
- It is personalized in that a customer's stated (or potentially inferred) preferences are constantly impacting how the process executes. Not only does the information on the transaction affect how the process executes, so does information about the customer. Ideally one has a 360-degree view of the customer and all this information is used to customize and target the process execution.
Given this definition then the capabilities of most environments will not allow me to have an analytically driven process. First, I have analytics only about the process execution. I am not using analysis of either customer preferences or transaction data/metadata in my analysis. I am likely using customer information and transactional information in my BI environment, but this is typically too focused on knowledge workers and performance measurement, not on driving my process analytically. I might use this information to derive many fine-grained segments for my customers, but operationalizing this into a production system probably means dumbing it down to a couple of processing options. Lastly, I am not focusing on the decisions within my process as units of improvement and so I am unable to replace or enhance my business intuition and policies with analytic insight.
There is, however, a way to get to a truly analytically driven process and that is to use decision-making capabilities that are embedded into the normal flow of work.4 This approach is known as enterprise decision management (EDM). EDM is an approach that involves focusing on the automation and improvement of operational business decisions - high volume, typically customer-facing decisions required as part of processing a core business transaction. So how, and why, will this approach result in an analytically driven process?
The first step to the approach is the simple act of focusing on the decisions within a process as separate opportunities for improvement in addition to the process itself. By focusing in this way, I can identify the policies, regulations, experience and, crucially, data that might impact how I make decisions. I can also use techniques like champion-challenger to see if a proposed strategy for making a decision would have better results, do a better job of simulating the impact of decisions and so on.
This separation must come first, but the next step is to automate those decisions in a way that will allow me to apply expertise, add analytic insight and maintain a level of business agility appropriate for my industry. The use of a business rules management system is vital here, as it allows a declarative approach to specifying the knowledge required to correctly make a decision, support the kind of auditing and management required by regulatory controls and because it allow business users to actively participate in the process of maintaining, testing, updating and managing these rules.
