The Data Analytics Blog

Our news and views relating to Data Analytics, Big Data, Machine Learning, and the world of Credit.

All Posts

Predictive Models For Your Customer Engagement Call Centre

January 22, 2019 at 8:06 AM

We apply the science of data analytics to assist our clients within various aspects of their customer-driven business and engagement process. Our products make use of predictive modelling techniques to facilitate the treatment of customers at the various stages of customer lifetime, for example during onboarding, growth and retention.

We’ve identified 4 models that can provide your call centre agent with valuable information during customer interactions.

Customer engagement models

Customer next best product model:

The next best product model aims to consider the current product holding of the customer and look at the other available products offered by the business to identify which product the customer is most likely to take up. This model therefore inherently includes product take up. The model will also allow you to be able to rank the products available for take-up so that they will also be able to identify the second and third best product.

Customer upsell/cross-sell take-up model:

This model aims to predict the probability of take up for any product. This model differs from the next best product offering in that here the model predicts whether the customer would be open to taking up another product regardless of what that product might be. Therefore, it does not provide a best product outcome but just a likelihood of taking up something.

Customer retention model:

This model aims to improve customer retention by identifying those customers most likely to cancel or close their accounts. The model identifies the profile and behaviour of customers who are likely to churn.

Customer value indicator:

This model uses historical performance data to derive the value of a customer at a fixed period. The agent can, for example, use this model together with the customer retention model. If the probability to churn is high for a customer, but their value is very low, then the agent might consider not spending a lot of time to convince the client to stay. If the customer's value is very high, the agent can put in more cost, effort and time to retain.

Methodology and benefits

A predictive model is built to be able to identify/profile existing customers. A predictive model is a statistical tool which uses historical data to forecast future outcomes. The model is made up of some predictor variables, which are variables that are likely to influence the results. e.g. life stage, product holding, account activity, balance trends, past payment behaviour etc. The model produces a score based on the variables that are selected. The model will allow you to rank your existing customer portfolio according to the likelihood of taking up an offer, responding to a campaign in the case of the upsell type models, or, in the case of looking at customer retention, the score will indicate the likelihood of a customer closing or cancelling their accounts.

 

The data required for these models include internal demographic information, customer service and/or complaint data and account and transactional level data that could be predictive. It is essential that historic campaign data is provided for a next best product or upsell model. If bureau data is used to build the models, for each of the records supplied, retro bureau data will be requested. If account and transactional data is supplied, characteristics relevant to the client can also be created. These variables are generally very good at predicting the outcome variables.

There are various benefits related to using the products stated above:

Customer next best product model:

The advantage of this model is that we are able to identify groups within the client's customer base that are more likely to take up certain products which directly translates into improved sales rates and increased profit by reduction of campaign costs.  The model answers the question of which product to sell next to the client.   As such the agent targets the right product for the right client to achieve higher sales rates.

Customer upsell/cross-sell take-up model:

In an organisation where multiple products are available, an analysis of existing customer behaviour can lead to efficient cross-selling of products. This directly leads to higher profitability per customer and strengthening of the customer relationship.  In the outbound cross-sell environment it allows the business to prioritise sales effort to those with a higher likelihood to take-up a cross-sell offer and the agent to be more efficient in his sales efforts, investing time where he expects a return.

Customer retention model:

The benefit of this model is that it enables the business to be proactive rather than reactive in their anti-attrition strategies as well as improved customer satisfaction.

Customer value indicator:

The benefit of using these types of strategies is to identify the value that groups within the existing customer base adds to the client’s portfolio. This will allow the client to be able to develop strategies that could be overlaid on various levels for example, if the cost of working a customer with a high probability of attrition is more than the value that was obtained from that customer over a fixed period of time, then it might not be worth including that client in the retention campaign. In conjunction with the retention model, a clear indication can be provided to the agent to what extent he should attempt to save the account with well-defined strategies, e.g. an escalation to management or a fee reversal vs only an apology vs close on request.

Data required

To develop customer next best product and upsell/cross-sell models the following historical data sources are necessary:

  • Dialler data indicating who was contacted together with the dialler outcome
  • Sales data indicating who took up
  • It is important to be able to match the sales and dialler data on the unique customer ID
  • Account and transactional data that includes payments, balances, delinquency etc.
  • Bureau data to be used as predictive characteristics for SACRRA members if available
  • Other demographic and geospatial data
  • For the next best product models, all these data sources need to be available per product offered

 

To develop a customer retention model, the following historical data sources are required:

  • Account and transactional data that includes payments, balances, delinquency etc.

Conveying data and information to your agents

It is vitally important, however, that your call centre agents have easy-access to the outcomes and recommendations made by these models.

With models built by Principa in our proprietary decisioning software, DecisionSmart, Principa’s Agent X Call Centre Virtual Assistant can deliver real-time call and agent specific data, decisions and information to your agents visually, to inform, guide and motivate them.

Read more on Agent X or request a demo.

Contact Us to Discuss Your Business Requirements

Muneera Salie
Muneera Salie
Muneera is an associate consultant at Principa. Muneera graduated with a degree in Chemical Engineering from the University of Cape Town. During her studies Muneera completed three years of advanced mathematics and focused on problem solving and analysis of chemical processes. Muneera has six years of experience in the consulting environment and has been active in the credit risk industry for the past two years. She has been progressively obtaining experience in the field of predictive analytics particularly in the development, monitoring and alignment of application scorecards within the South African and Middle Eastern markets and various other ad-hoc analytics and marketing services projects.

Latest Posts

IFRS9 Expected Credit Loss (ECL) Models For Retail Lending

It is expected that IFRS9 adoption should lead to an increase in provisions (initially a balance sheet / retained earnings adjustment only with commentary on retrospective impacts). Typically, the increase is mostly a result of loss provisions for all accounts (regardless of a loss event) and the extension of the loss period from a typical 12 months to lifetime (e.g. for structured loans the remaining term of the loan plus time to default/write-off plus the recoveries window).

Engaging Your Customers In A Digital Era At The 20th Customer Engagement Technology Conference

We’re excited to announce that we are a sponsor at this year’s 20th CET conference in South Africa! The Customer Engagement Technology Conference has been successfully bringing together marketing, digital/ IT and business heads to discover the latest solutions that enable greater customer experience in various industries.

Why The Human Mind Is Flawed At Making Credit Decisions

For a while, we have been running a blog series on cognitive biases and logical fallacies that data scientists should avoid. In philosophy there are a host of informal logical fallacies – essentially errors in thinking – that crop up every day. In this series we have looked at the practice of data science to determine how these same fallacies also occur. Today we will be looking at fallacies and their manifestation in credit: The Monte-Carlo fallacy and the Hot-hand fallacy with some studies in the credit world.