Santam is South Africa’s largest short-term insurance company. With more than 650 000 policy holders and assets under management of R 17 bn, the company enjoys a market share of more than 22 %. Santam operates in a market where fraudulent activity can account for an estimated 6 to 10 % of all premiums. Santam wanted to find a way to improve its service to customers by settling claims faster and keeping premiums low. To achieve this, the company needed to maximise operational efficiency and find smarter ways to combat fraud. Santam worked with us to design a claims segmentation solution based on SPSS predictive analytics software. Our solution needed to integrate seamlessly with the current Santam claims process, and allow for a real-time scoring of each claim with a sub-second response rate. The solution works on a claim by claim basis. Each claim is automatically scored according to its risk level, and then distributed to the appropriate processing channel for settlement or further investigation. This allows high risk claims to be sent to the appropriate investigation unit, while low risk claims can be processed very quickly improving overall customer service. The solution offered by our team:
- Enhanced Santam’s ability to detect fraud, foiling a major crime syndicate and saving R 17 m in the first four months of operation.
- Improved customer service by enabling legitimate claims to be settled within an hour, more than 70 times faster than before.
For more information you can download the case study from Nucleus or watch the following video:
Mutual & Federal
After the success of the Santam implementation, we were approached by Mutual and Federal, South Africa’s second largest short term insurer, to implement a similar solution for them using SAS software. Our predictive analytics unit is in the final stages of implementing this solution and it looks to be a similar success story to Santam. They are also working on other predictive analytics projects for M&F at the same time. The first phase of this project (real-time scoring of motor claims) was implemented from start to finish in approximately 8 months indicating the speed of our team.
Major SA Retailer
We implemented a churn solution for a large South African-based multi-billion dollar retailer which employs close to 32 000 people. They approached us to help with extracting value from their customer datasets. The first solution that we developed was a churn model for their cellular client base. This model predicts when a customer is likely to leave their network and helps them to design targeted marketing campaigns aimed at retaining their high value churners. The model was proven to identify the churners with a high level of accuracy. We are currently mentoring and collaborating with the retailer’s in-house analytics team on further predictive analytics projects.
P-Cubed is a leading management consulting firm specialising in delivering critical and complex programmes and portfolios for governments, agencies and major institutions across all industry sectors. P-Cubed had a requirement for an automated score card development process to facilitate credit-based response modelling. We initially developed a score card within IBM SPSS Modeler which presents the credit risk scoring functionality as functional nodes within the IBM SPSS Modeler environment much like other functional nodes which ship with the product. The score card solution incorporates credit risk scoring methodologies with respect to:
- Pre-processing typical credit history data to prepare it for the analytical model.
- Ranking credit behavior predictors and producing diagnostics on the suitability of predictors
- The execution of relevant models
- Diagnostics on the suitability of the model
- Production of a score card
- The production of a range of diagnostics which aid in the determination of score thresholds
Overall, our solution provides a facility for rapid automated development of score cards alongside performance assessment utilities, and is used as a key component of P-Cubed’s business process. We have since developed an extended version of the score card facility which provides extended graphical capability amongst other improved features.
Millicom International Cellular, also known as Tigo, is a mobile phone network provider which provides affordable, widely accessible and readily available prepaid cellular telephone services to more than 30 million customers in 13 emerging markets, in Latin America and Africa. Tigo approached our team to assist them with two business challenges:
- To understand the profile of their customers and how these customers use their products
- To predict the customers most likely to leave them for the competition
We worked with Tigo to develop two solutions to address these challenges. The first solution is a segmentation model which identifies different behavioural profiles within the customer base. Such information enables Tigo to conduct highly targeted marketing campaigns and to develop products and services appropriate to each profile. The second solution is a churn model. Most of Tigo’s customers use prepaid services which allow them to churn at any time and without any warning. This poses a problem for them as the cost to acquire a new customer is considerably higher than retaining existing customers. The churn model developed by our team scores each of the customers on a regular basis to predict the probability of churn in the next few days.
Nedbank is one of the four largest retail banks in South Africa based on the number of customers. With such a large customer base, it becomes critically important to monitor the customer sentiment at any point in time. With this in mind, Nedbank approached our team to help them to develop a social media monitoring tool based on IBM SPSS software. The tool enables Nedbank to collect data on the online interactions that the company has with its clients across a variety of different social media feeds including Twitter, news feeds, blogs, Facebook and other relevant sites. All of the interactions are collected and placed into a structured database where they can be used for analysis. The interactions are stored in the form of text, which is known in the analytics world as ‘unstructured data’; data which are typically very difficult to use in any analysis. However, the solution developed by our team uses Text Mining to derive meaning from this unstructured data. Text Mining is a set of intelligent algorithms which analyses text and derives meaning from it based on key words and the context in which they were given. In this way, the online interactions are categorised and turned into ‘structured data’ which are then used for further analyses. The data are used in the Nedbank solution to drive customer engagements and upsell opportunities. There is also potential for the data to be used in predictive models to drive customer retention and other goals.