Precise Prediction scorecards can be deployed alongside traditional models, such that the shortcomings of older models are quickly highlighted. With machine learning techniques applied to real-time data, recalibration of our custom-made algorithms will quickly result in 10% to 15% improvement in portfolio performance. In the longer term, with on-going monitoring and recalibration resulting in an “ever-improving” scorecard solution, an overall performance improvement of 30% is typical.
Unfortunately not all lending decisions will result in a perfect repayment record. Customer circumstances often change and when this happens, and they fall behind with their repayments, it’s important to act quickly to understand the customer situation and take appropriate action.
As much as data can assist with lending decisions, it can equally be used to inform Collection strategy and activities. When innovative techniques using predictive analytics and machine learning are applied, the models generated can be used to determine what actions to take under which circumstances to achieve the desired outcome quickly for the lowest possible cost.
Collection scoring is a relatively new concept, which may seem odd since credit scoring to help determine a lending decision is the de-facto approach from most consumer lenders. From a collection scoring perspective lenders themselves hold the key – they have the data on past cases, they know what actions produce results, they know what the cost is to collect a 1 payment past due arrears case, but simply having the data isn’t enough. Having worked with debt management companies, including the largest in Europe, we understand the challenge of managing debt collection and how data can be harnessed to improve efficiency, effectiveness and ultimately profitability.
By way of an example, to improve the effectiveness of collection agent phone calls it is important to be able to predict when the customer is likely to answer the phone. Looking at past case outcomes and customer data on employment, demographic and location, a collection scoring model will be able to predict the best time to call and the number of calls that will be needed to achieve the desired outcome. Also whether a phone call is the right approach. There is no value in continuing to pursue a strategy that past data suggests has little chance of success.
The scoring models will vary as each clients’ customer base and market segment is different. However, the techniques of advanced analytics with machine learning will deliver benefits that over time will improve still further as the modelling platform “learns” from the result achieved and the models are recalibrated.
Such highly effective modelling dramatically improves overall portfolio performance and profitability, and at the same time helps build stronger customer relationships.
To be able to accurately predict the cost to collect is a key measure for any debt management business. By focusing on customer-specific collection techniques clients have experienced a 10%-15% increase in activity effectiveness.
The diagram below depicts a typical model structure, one that has proven successful with the largest debt management company in Europe.
Collections / Debt Management Score Models
The business improvement from our solutions will depend on the current situation of our client’s business. However, typically the initial improvement will provide a tangible ROI benefit, with further advantage being delivered as the machine learning characteristics begin to apply.
Industrial / Manufacturing
The power of Statistica is borne out by its wide-ranging use in the Industrial and Manufacturing sectors. The advanced features of the software enables clients to predict supply chain requirements, align distribution and optimise maintenance processes, which translates to minimising cost and maximising profitable revenues.
As businesses strive for competitive advantage, the winners will be those that can successfully understand what they need to do and when, ahead of time.
Across the globe Government departments are striving for efficiency and seeking to optimise their processes to save cost, or in one client case had a requirement identify and stop VAT fraud.
Through the use of Statistica software and Machine Learning models developed by Precise Prediction, the VAT office is able to identify and target likely VAT fraud cases.
Being able to predict behaviour and resulting outcomes is crucial to any business decision. In marketing the cost / benefit question is frequently to the forefront of any initiative.
For one client the challenge set was to identify opportunities to increase revenues from annual membership fees. Taking account of a large number of variable factors, Precise Prediction was able to provide a series predictive models along with recommendations, including promotional offers, that could be used to encourage annual membership renewals and a broader spend on other products and services.