How a global B2B services company moved from reactive churn response to proactive retention across 250,000 service contracts.
This client managed more than 250,000 multi-year service contracts across global markets. Their renewal process was calendar-based: teams worked their renewal lists in date order, contacting accounts weeks before the renewal date regardless of risk level.
The problem was structural. By the time an account appeared on a renewal list, the at-risk ones had often been disengaging for months — service issues unresolved, communication declining, competitive conversations already underway. The intervention window had closed before the team even knew it was open.
Leadership lacked confidence in renewal forecasts. There was no systematic way to distinguish the 10% of contracts that needed urgent attention from the 90% that would renew without friction — so teams treated everything the same, and results were accordingly inconsistent.
No early-warning system meant the client only became aware of at-risk accounts when customers were already signalling intent to leave — weeks after effective intervention was possible.
A meaningful percentage of the 250,000-contract portfolio at risk annually. Renewal rate variability that made revenue forecasting unreliable. High cost of reactive win-back efforts.
We built predictive models incorporating service delivery performance, contract history, engagement patterns, payment behaviour, and account health signals — producing a calibrated risk score for every contract in the portfolio, updated daily via Azure ML pipelines.
Given stakeholder uncertainty about the methodology, we implemented both interpretable (logistic regression with engineered features) and ensemble (gradient boosting) models in parallel. Cross-validation and business workshops were used to validate results and build the organisational trust needed for a global rollout.
Automated risk alerts triggered for contracts showing elevated risk scores, weeks before their renewal windows opened — giving service and sales teams the lead time they needed to intervene effectively rather than reactively.
An Azure OpenAI-backed retention agent was deployed within Microsoft Teams — surfacing account-level context, risk explanation, service history summaries, and guided outreach playbooks directly in the tool sales reps used daily.
Power BI dashboards with embedded Canvas Apps gave service and sales teams a single interface to triage, prioritise, and act on retention signals — replacing manual reporting with live, risk-ordered account views.
The shift from calendar-based to risk-ordered outreach changed the team's behaviour more than the technology did. When reps understood which accounts needed attention and why — and had the AI assistant to help them prepare — retention conversations became more targeted, more timely, and more effective.
Because stakeholders could see the model logic and validate findings against accounts they knew, the system earned credibility quickly — accelerating adoption across global sales teams.
Embedding the retention assistant in Microsoft Teams — rather than a separate tool — meant reps used it as part of their natural workflow. Adoption was near-universal within the first quarter of deployment.
We can model your contract base and produce a risk distribution before the engagement begins.
Get a portfolio risk assessment