All information systems require effective collaboration between business people and IT departments to be successful. Decision Management Systems require collaboration between business people, IT departments, and analytic teams. Ensuring these three groups work together effectively creates the “three-legged stool” that provides stability for Decision Management Systems. A three-legged stool cannot be stable with only one or two legs, and Decision Management Systems cannot be driven entirely by the business, by IT, or by analytic teams in isolation.
Decision Management Systems demand this kind of collaboration for several reasons.
- The digital transformation decisions being handled by Decision Management Systems are business decisions. Even though each decision is small, their business impact is great because there are so many of them. Because they are business decisions, the definition of what makes a good decision or a bad decision is driven by the business. The actions that can be taken, the logic that should be followed, the policies and regulations are all specified by and understood by the business.
- Decision Management Systems handle high-volume, low latency decisions— that must be taken often and taken quickly. This means automation is a necessity. Decision making must be deployed into a production or transactional environment, and, to make this work, IT must be part of the solution.
- Some Decision Management Systems implement decisions that are entirely constrained by policy and regulation. Many involve the use of analytics to predict risk, fraud, or opportunity. Some use optimization to manage tradeoffs. Even when embedded predictive analytics or optimization are not required, the decisions made should be reviewed analytically—by the numbers wherever possible and continuously improved and refined. This means analytics expertise will also be required.
In most organizations, these three groups don’t have much of a history of collaborating, and many issues divide them:
- IT departments and business users are often on different pages, with IT struggling to support changing business needs while business people get frustrated with their systems.
- Analytics people often focus on the accuracy of their model, not the business outcome. They produce models that they consider to be “best” because they have the best statistical measures—not because they produce the most profitable result.
- Analytics teams often build models that include data not available in the production environments that will use them, resulting in models that take six months or more to implement or that are too complex for IT to use. In addition, IT often does not capture the data that analytic teams require to validate and improve their models.
- IT departments design systems and specify enhancements without the kind of analytics skills that would let them see how analytics could improve the systems they are working on.
A lack of mutual understanding and an absence of collaboration may be typical in organizations but unacceptable when building Decision Management Systems. To ensure a strong stable three-legged stool for a Decision Management System, you must begin a three-way conversation and start building collaboration skills. A focus on decision discovery and business results will improve alignment, but organizational issues must be recognized and addressed no matter what.
Start a three-way conversation
Getting these three groups into a conversation and getting them thinking about collaborating is essential. The teams need to participate in identifying the decisions that matter to the business and in defining good and bad decisions. The IT team must be able to explain how the production systems that use those decisions work and be willing to discuss how they could involve business people to define and manage the logic of these decisions. Analytic teams need to understand the business need, talk to IT about the data available, and see what kinds of models they can build that will result in better business outcomes while still being deployable. If there are ways existing systems could be improved to get better analytic results (and so deliver better business results), then this too needs to be discussed. A simple awareness that these things are necessary, that the three groups must work together, is the first step.
Build collaboration skills
The most important element of building collaboration skills for these groups is simply to expose them to each other and to create opportunities for them to talk and work together. Getting the whole team together for meetings and ensuring that the project’s process or methodology reinforces this point of view is the foundation for collaboration.
Cross-training can be effective too. This can take the form of overviews of the business, information technology, or analytic approaches. These overviews can help the experts in one group get a basic understanding of the critical terms, technologies, and approaches of the other two groups. In general, everyone needs to understand the overall strategy—why is this Decision Management System being built—as well as the basics of the approach and the supporting technology.
When new technology, such as a Business Rules Management System or Predictive Analytic Workbench, is introduced, it may be possible to go further. Consider having everyone attend the first few days of training to get an overview and spend some time using the software before having the experts drill into the details. Having business, IT, and analytic teams attend training on how to develop business rules with a BRMS reinforces the mindset of a shared solution. In this case, the IT department will need some time to cover more technical aspects but it is very effective to have everyone sit in initially. Similarly, a graphical Predictive Analytic Workbench is remarkably approachable for business and IT users, even if they are not going to doing production-quality model development and refinement.
Finally, consider using brown-bag lunches or “lunch-and-learn” sessions, where experts in one group share their expertise with other groups. This builds a sense of teamwork among the three groups, provides critical know-how among team members, and helps those presenting get a sense of the level of detail that is appropriate when dealing with others outside their own area of expertise.
Decision discovery is critical
Experience with many Decision Management System projects is that decision discovery is the crucial phase for getting the three teams to work together. It is easy to think that this can be deferred until Decision Services are being built. In fact, early collaboration is better for several reasons:
- Decision Discovery sets the context for all subsequent projects and activities. If the three groups begin collaborating early, then this collaboration will carry forward into later stages. If decision discovery is carried out entirely by one group, then the others will feel that the model created is being imposed and does not address their concerns. Begin as you mean to go on.
- Having the business and IT teams work together on the top levels of the decision hierarchy is critical, as this maps the Decision Management Systems into the business processes and existing systems that both organizations are familiar with. Building a shared sense of where the decisions fit also builds long term cooperation.
- Analytics teams have a lot of experience thinking about how to use predictive analytic models to make decisions. Their engagement in the decision decomposition process will improve the model created and help them see the context for the modeling they will do later. Their ability to also say “we could probably predict X and Y, would that help?” is also very helpful at this stage, as it will guide the business and IT teams towards more analytic decision-making. They will also know what data will need to be available to build each predictive analytic model they suggest.
- Ensuring that the know-how identified in the model maps to the kinds of policies, regulations, expertise, and analytic insight available and that the information sources map to known or knowable data sources will keep the whole model grounded. This requires all three groups to be active participants.
Align around business results
One of the most effective ways to get disparate groups to collaborate is to align their personal and organizational objectives. If an analytic modeler, for instance, is motivated to produce the most precise model possible, then that will fuel a certain kind of behavior. If, instead, they are motivated to produce the model that has the most business impact as measured once the model is in production, then they will behave differently. The new motivation will help ensure they focus on making the model as easy as possible to deploy, whether or not the production environment has the data the model needs and whether or not the model is actionable in a way that makes sense for the business.
As far as possible, ensure that everyone on the team is getting rewarded and measured based on the impact the Decision Management System has on business results as measured in production. Not the estimated results or intended results but the actual results. Not IT-specific or model-specific measures, but business measures.
Resolve organizational issues
One of the ways you can strengthen the three-legged stool is by resolving organizational and reporting issues that push the groups apart. Start by identifying the reporting structure for all those involved in the project and ensuring that their managers understand the purpose and benefit of the Decision Management System. This can be used to identify the executive sponsorship that matters to each reporting structure. If the executive sponsor is someone that all three management structures care about, then organizational issues will be suppressed in favor of overall success.
The locations of teams can also be an issue, with the three groups operating out of their own enclaves. Co-locating the team and providing a war room in which business, IT, and analytics people all work together at least some of the time can make a big difference. This fosters a sense of collaboration and teamwork, while also making it easier to get answers to trade-off questions and “what if” scenarios.
Finally, the workload of the three groups must be considered and managed by the sponsors of the project. It does no good to focus the analytics team on deployed results if they are immediately assigned to a new project as soon as their model is finished. Similarly, making sure the business team can make the business rule changes it wants may require some reassignment of workloads and responsibilities. It is essential to understand who assigns work to the team members involved and how issues can be resolved when they occur.