Predictive churn model helps a plastics manufacturer prioritise sales engagement

Predictive churn model helps a plastics manufacturer prioritise sales engagement

A multinational plastics packaging manufacturer turned a blind spot into a live signal. Its sales team could not see churn coming, because nothing in the relationship forced a customer to say goodbye: there was no contract to lapse and no cancellation notice to trigger a call. Customers simply ordered less, quietly, for reasons the sales team rarely saw until the account had already drifted away. QuantSpark built a bespoke, explainable statistical model that scored every customer's churn risk from their own order behaviour, then fed that score automatically into the systems the sales team used every day, giving every region a real-time view of who was slipping, and why.

The problem

Churn here was non-contractual, which made it structurally hard to predict. In a subscription business, a cancellation date is a clean signal. In this business, the only signal was a gradual softening in order volume, timing and consistency: easy to miss, and easy to explain away as a one-off dip until it was too late to intervene.

The customer base made this harder still. It was large and diverse, spanning different data sources and behavioural patterns across regions, so no single manual rule of thumb could flag risk reliably everywhere. A time-constrained sales team, focused of necessity on its largest accounts, had little spare capacity to watch the broad population of mid-tier customers closely, which was exactly the segment where a slipping order pattern was most likely to go unnoticed until the account was gone.

Any answer also had to earn trust, not just accuracy. The business needed a model that was credible and explainable to a non-technical sales audience, demonstrably effective, and properly integrated into the enterprise resource planning and customer relationship management systems and dashboards the sales organisation already relied on, rather than a standalone analytics exercise sitting outside day-to-day workflow.

The approach

QuantSpark started by hypothesising, with the client, the likely behavioural drivers of churn in a relationship with no contractual trigger: order timing, order volume and recent service experience. From there, statistical techniques were applied to customer-level order histories, behavioural patterns and periodicities to build a consistency index of order regularity, designed to flag customers whose future order volumes were likely to fall.

Because there was no cancellation event to calibrate against, the team used statistical methods to quantify the risk associated with particular order behaviours directly, rather than applying fixed thresholds. This picked up changes in order volume that were often small in isolation yet strongly correlated with future churn, and weighted the resulting risk score using behavioural signals such as recent service experience.

Crucially, the model was deliberately pointed at mid-tier customers, the segment a time-constrained sales team would otherwise underserve, rather than spread evenly across the whole book. It was then productionised within the client's existing systems, with risk scores fed automatically into the CRM and dashboards already used by sales teams across every region, so the output arrived inside the workflow sales already had, not as a separate report to check.

The workflow

Order and service data drawn from the client's existing systems feeds the consistency-index model, which scores and weights each customer's churn risk on an ongoing basis. Those scores are pushed automatically into the CRM and dashboards, giving regional sales teams a continuously refreshed, explainable view of who is at risk and why, ready to act on without leaving their usual tools.

The result

The model reliably identified customers at higher risk of churn from their recent behaviour, giving the sales team a real-time view of churn propensity and the leading indicators behind it, not just a flag but a reason. Sales could prioritise outreach to previously underserved mid-tier accounts and manage churn more deliberately, and the engagement built wider confidence in analytics across the organisation, a foundation that reached beyond the churn use case itself.

No quantified uplift, cost saving or churn-reduction percentage is given in the source, so none is claimed here. The value case rests on a qualitative but concrete shift: from a sales team blind to churn among its mid-tier accounts, to one working from a live, explainable, systemwide risk score integrated directly into the tools it already used.