Introduction

A Temenos Analytics Funds Flow (FF) Predictive Model is a forecasting model used to estimate the average Income, Expense and Savings of a customer over a period – a month or a year, based on the customers transactions. Such estimations are of great business value for creating targeted and efficient marketing campaigns, making loan approval decisions as well as when used as inputs to other predictive models. Although in some cases customers’ income is tracked in the banking system, the values there tend to age and after some time become out of date and not quite reliable. Even when reliable, these numbers often contain only gross salary income and do not include additional sources of income like social payments, dividends, bonuses, extended benefits, interest income, etc.

In contrast, a Funds Flow model can be reprocessed periodically with the newest transactions’ data and thus will reflect the dynamics in a person’s income, expense and savings. The monthly variability of the latter is also tracked by the model and can provide additional insights.

Since transactional activity is the main data source for the model, the accuracy of the estimated flows (Income, Expense, etc.) largely depends on how engaged the customer is with the bank. In general highly engaged customers will use a variety of bank products, will use different types of transactions, will have most of their income sent to their accounts with the bank, will do most of their expenses through the same bank, etc. For such customers, the transactions captured by the banking system will encompass most of their financial activities and thus the estimated values will be very close to reality.

Conversely, dormant customers will have low engagement, very sparse transactions and therefore the estimations produced by the model will be quite off. The model supports the calculation of a configurable Engagement Score that ranges from zero (i.e. no engagement) to 100 (i.e. fully engaged). The score is composed of five elements with configurable weights and thus allows the bank to adjust the scoring to the specific business realities. The score is then used as a measure of trustworthiness of the estimated values. In addition, estimations for customers with scores below a preconfigured value are ignored and do not occupy extra space.

The predictions produced by the model can be configured to flow into the Warehouse by using an installer. This is referred to as ETL integration and adds the following columns to the Warehouse: