Building a Predictive CRM Value Model to Maximise Customer Lifetime Value

The Approach

The solution combined CRM strategy with predictive analytics. Instead of relying solely on historical spend, multiple behavioural signals were used to predict where each customer was likely to move next.

The process consisted of three stages.

1. Predict Future Customer Value

Machine learning models were trained to estimate each customer’s probability of moving into one of four future value bands:

• VIP
• High Value
• Medium Value
• Low Value

These predictions considered historical behaviour alongside engagement patterns to estimate expected future Gross Gaming Revenue over the next 30 days.

Rather than producing a single prediction, every customer received a probability score for each potential value band.

2. Simplify the CRM Strategy

While the models generated many possible movement combinations, managing dozens of customer journeys would have created unnecessary operational complexity.

Instead, similar prediction scenarios were grouped into a small number of strategic CRM segments.

This reduced complexity while still preserving the commercial opportunity identified by the models.

The result was a scalable framework that marketing teams could actually execute.

Benefits included:

• Easier campaign management
• Clear prioritisation
• Reduced operational overhead
• Flexible enough to evolve as model performance improved

3. Allocate CRM Budget Intelligently

The prediction scores were then used to determine how much marketing investment each customer should receive. Instead of giving every customer within a segment the same offer, promotional investment became proportional to predicted future value.

Customers with the highest probability of becoming VIPs received stronger incentives, while lower-probability customers received lighter-touch communications. This created a tiered CRM strategy where promotional costs aligned with expected future returns. One of the most valuable outcomes was moving beyond traditional segmentation.

From Prediction to Personalisation

For example, two customers might both currently sit within the High Value segment. Historically they would receive identical offers. Using predictive scoring, however, their future potential could be very different.

Customer A might have an 80% probability of becoming a VIP next month.

Customer B might only have a 35% probability.

Instead of treating them equally, CRM could prioritise investment where it was most likely to generate incremental revenue. This approach enabled more intelligent decision making without increasing campaign complexity.

Business Impact

Introducing predictive value modelling delivered several strategic advantages:

• CRM investment focused on future revenue rather than historical spend
• More efficient allocation of promotional budgets
• Earlier identification of emerging VIP customers
• Reduced bonus costs through smarter targeting
• Greater personalisation without creating hundreds of campaign variations
• A scalable framework capable of supporting millions of customer decisions

Most importantly, CRM evolved from reacting to customer behaviour into actively influencing future customer value.