Data Analytics Automation: A double-edged sword?
In this day and age, I don't think anyone that has been around the data analysis or business intelligence space needs to be convinced that analytics have their purpose within an organization. They are absolutely a "must-have" set of tools for any medium to large-sized company. However, many organizations that take on a large analytics project can often rush in without considering the impact that the new program will have on the business groups or individuals responsible for maintaining the systems, including remediation and follow-up.
So, what is analytics automation anyway? To put it simply, analytic tests are fully developed, have no manual manipulation of data, and are running at regularly scheduled intervals. Access to a variety of data sources has been setup in advance so that access to new and updated activity is fully automated. Many of the commercially available monitoring packages also include sophisticated web interfaces, e-mail notifications, workflows, remediation tracking, dashboards, and heat maps. The most popular areas for implementing analytics are in the areas of General Ledger, Travel & Entertainment, Purchase to Pay and Order to Cash.
This all sounds wonderful. So, what factors can possibly be of concern with analytics automation?
A key consideration is around how the organization plans to manage the exceptions identified in the analytic reports. In the beginning, most of the activity identified will be a false positive (up to 90%+), so reviewer fatigue can quickly set in without a solid process in place to deal with the volume of what is found. Have all impacted business groups and reviewers been adequately prepared for the additional work that is inevitably required to go through and review the exceptions? Good stuff can be lost in the sea of false positives, so methods of better identifying the most interesting activity need to be developed as the project progresses.
Another potential issue is that analytics can quickly become stale if the rules are not easily modified and regularly maintained. "Black box" type analytics, where the processes happening in the background are not obvious to those tasked with maintaining the system, are more difficult to work with. Organizations will typically bring in experts from the proprietary company to implement the tools, but can fail to consider the ongoing costs of using the same experts for post-implementation revisions.
All of this said, there are strategies that can be implemented to offset these and other issues that can arise from a new analytics program.
Pre-implementation meetings are vital as they serve to better prepare the individuals that will be tasked with reviewing the output of the analytic programs. Roles and responsibilities, the expected exception volumes (in the beginning and over time), and involvement with the fine-tuning of results over time should be discussed in detail. Analytics are powerful tools that can include enhanced logic to sort on dollar amounts or frequencies to highlight higher risk activity first, and the combining of multiple reports into a consolidated report can help identify entities that have a higher number of exceptions spanning multiple reports. Having an organized approach to the project is key to efficiently utilizing the full capabilities of the analytics software.
For the analytics program to evolve, it is critical that regularly scheduled logic reviews be setup with the owners of the analytics to tweak, retire, or add new analytics. Like any living entity, analytics do have a natural life cycle and will quickly die if they are not kept current. To keep the analytics up to date and the costs of maintenance down, hiring or investing in training to bring expertise in-house is one of the most important steps that should be considered.
In short, analytics automation is a valuable addition to any organization; however, do not think the work is done when the system is first up and running. Apart from an organized approach to implementation, keeping the system growing and generating timely and meaningful results requires a dedicated team to continually monitor, review and fine-tune the analytics.