Thesys Group The Data Mining Company - Factoring & Debt collection workflow optimization

Typical sponsor

CEO
Head of Debt Collection Division
Risk Managers

Typical participant

Debt Collection Workflow Experts
Analysts
Data Administrators

To whom do we recommend?

Collection agencies, debt buyers, creditors and companies with debt collection and/or factoring “issues”

Aim of the project

Support the process of small and unsecured debt collection with statistics and data mining models.

Methodology

There are several statistical methods to assess not only the process but also the relevant participants. In statistics, logistic regression uses numerical and/or categorical variables as predictors to forecast the outcome of a given event (will/won’t pay). Decision tree model creates a graphical decision support tool as an output that uses a tree-like graph to display the decisions and the possible chances of event outcomes, resource costs, and utility. The variables of the living area are essential to improve the precision of our models.

Results

It can be used to validate the existing methodology or create a new methodology and a system of recommendations.

Improves the knowledge base: Confirm or refute association rules, connections assumed by the experts of the area
Better return rates: The return rate of the collection actions can be measured, predicted, planned and optimized, the outcome is a higher profitability
Rationalized actions and reactions: It can be implemented like a decision supporting system (for example should we allow installment or ”discount”)
Improve the collection process, which leads to well-founded planning and higher returns

DOs and DON’Ts

Consider the possibilities of purchasing external databases, because you always want to know more about your debtors
The projects need a very close cooperation, so calculate with the additional duties of your human resources
Accept the results as guidelines and recommendations to consider and not strict rules

Don’t

We don’t recommend you our services, if you do not have the “critical mass” of data about the debtors or the quantity of debtor addresses in your database is low (missing values)
Do not be “afraid” of change and preserve processes that are strictly defined, and not flexible enough to adapt the recommendations of the created models

The output of the decision tree model is a tree-like graph that represents the results of the decision rules given by the model. Through the number and percentage of good and bad cases on a leaf we can evaluate the goodness and the strength of the split.

Why we are different

Expertise in data source management (collecting, managing, cleaning and connecting data sources)
Multi-aspect approach considering the bank-debtor and the debt-handling information methodologies together and simultaneously
We not only use your data from the process of collecting (internal database), but also the relevant databases from the market (for example, attributes of the executors, courts, regions, etc.), in order to improve the created models

Factoring & Debt Collection Optimization