We had a large organization that used an internally developed solution for recording and approving changes. Changes could have 1 to many approvers and 1 to many task activators along with a variety of different artifacts. This tool was generally the bottleneck for many initiatives and had an estimated impact of 2-3 million USD in cost over a year in wasted resources due to these inefficiencies.
Dynamic Duo interrogated change records for trending of changes over a multi-year period in respect to multiple metrics using Machine Learning. We found multiple correlating trends that we investigated and determined to have a high correlation with expected results. We proposed that we could develop a data driven solution that could use a dynamic model that we could build and refresh on relatively infrequent intervals that would provide necessary output to assist in reducing the inefficiency.
The model was created according to the data collected in the analysis phase. We then built a scalable Cloud Application that utilized Machine Learning technologies with a light footprint that would adjust during periods of high utilization and then scale down to reduce infrastructure costs during non-peak hours of operation. Application was introduced and utilized by key stakeholders with little to no training. Initial application was expanded using an Agile methodology until all key features and functionality were fully developed.