We started the process with initial data analysis, which was critical as the company assets consisted of dozens of small portfolios. Each portfolio had a different structure and quality, making it necessary to identify common variables. Otherwise, prepared models could not be universally applied across the company’s operations. The project took three months to complete and was handled using CRISP-DM. During the project, over twenty models were thoroughly tested and six were shortlisted for production deployment. They varied in data usage as well as algorithms qualified for training. Shortlisted models were deployed using Brainhint AI Farm software and integrated into operations semi-automatically. Training of the models was automatic but predictions were provided on demand via a custom built GUI.
Improving Debt Collection
We have created and implemented a system that anticipates the willingness of debtors to cooperate.
Our client needed to know the willingness of debtors to cooperate. They would purchase debt portfolios but with less than optimal returns on their investment. The key factor impacting our client's business is the net return on purchased portfolios, calculated as the revenue derived from the portfolio over the purchasing and processing costs.
Our client has improved their operational effectiveness by a staggering 36% within just six months. Automating decision-making has saved around 160 hours per year otherwise spent on manual processes. Our cooperation has continued to further improve developed solutions and to innovate in other areas of operation.
increase in operational effectiveness
saved as a result of decision-making automation