How Bricklay built an ML-powered early-warning system that gave a market leader's C-Suite weeks to act — reversing a market-share decline and achieving a 98.5% customer retention rate across 100,000+ accounts.
C-Suite alerted before churn — not after the cancellation call
Revenue retention achieved across large & medium portfolio accounts
Siloed data sources unified into a single customer churn intelligence layer
ML-predicted churn segments: Low, Medium, High — ranked by revenue exposure
Every B2B services company has the same blind spot: the moment you find out a major account is churning is usually the moment they’ve already decided to leave. By then, the options are limited. The revenue is gone.
Our client — a market leader managing 100,000+ corporate accounts across a large B2B services operation — was experiencing exactly this. They were losing their highest-value customers to competitors, and the pattern only became visible in hindsight. After the credits, after the complaints, after the cancellation.
The leadership team knew the problem was real. What they didn’t have was a system that could tell them which accounts were at risk, how much revenue was exposed, and — most critically — with enough lead time for the right people to act. Bricklay was engaged to build that system.
The company wasn’t short on data. They had survey scores, work order histories, billing records, pricing data, recurring revenue trends, and industry benchmarks. The problem was that these datasets lived in separate systems and spoke different languages. No one had ever assembled them into a coherent picture of customer health.
As the business scaled, the gap between what the data knew and what the leadership team could see grew wider. Large portfolio customers were disengaging — signaling dissatisfaction through slow payment, reduced order volumes, and rising service complaints — but the signals were invisible until they reached a tipping point. By the time an account escalated to executive attention, the window to save the relationship had typically already closed.
"By the time an account escalated to executive attention, the window to save the relationship had typically already closed."
Three specific failure modes kept repeating across the business:
Bricklay’s engagement started with a question leadership couldn’t yet answer: of all our at-risk customers, which ones represent the most revenue exposure — and which ones can we still save?
To answer it, we unified six datasets that had never been analyzed together: customer survey scores, client support request history, invoice credit records, recurring revenue trends, work order delivery performance, and price revision data. These became the training foundation for machine learning models designed to identify departure signals before customers acted on them.
The models segmented the entire customer portfolio into three churn risk tiers — Low, Medium, and High — weighted by each account’s recurring and non-recurring revenue size. The output wasn’t a data dashboard that analysts reviewed weekly. It was a live intelligence layer that routed early-warning alerts to the right VP with enough context to intervene intelligently: what the account’s pain points were, what their revenue contribution was, and which remediation actions were most likely to change the outcome.
"For the first time, the C-Suite had a mechanism to be proactive — not reactive — about the customers that mattered most."
When a high-value account moved into High churn risk territory, Operations, Sales, and Customer Support leadership were activated weeks ahead of any cancellation signal. The right people had the right information at the right time — and the authority to act on it.
Achieved across large and medium portfolio accounts — the segment with the highest revenue concentration and competitive exposure.
Leadership was mobilized before churn occurred — not after — giving Operations, Sales, and CS VPs a defined intervention window.
Survey scores, billing, work order history, pricing, revenue trends, and industry data — unified for the first time into a single ML-ready dataset.
Data revealed the specific drivers of attrition — slow work orders, abrupt pricing, lack of follow-through — giving leadership an evidence-based CX roadmap.
The company introduced five targeted interventions — annual price revision policies, proactive credit issuance, product offering realignment, bundled pricing, and seasonal re-engagement campaigns — each one informed by the retention analytics rather than chosen by intuition. The results compounded: higher satisfaction, stronger loyalty, and measurable improvement in net revenue retention across the customer base.
More significantly, the nature of how leadership operated changed. SVPs and VPs became proactive rather than reactive. The team learned, through data, that the primary drivers of attrition were operational — not pricing. These were fixable problems. But they were only fixable because the data made them visible early enough to act on.
This capability is directly relevant to any B2B services organization where the following conditions are true:
More than 20% of your revenue is concentrated in fewer than 20% of your accounts
Customer attrition is discovered reactively — typically after the relationship has already deteriorated
Your operational data, billing data, and customer satisfaction signals live in separate, disconnected systems
Leadership has no structured way to prioritize retention effort by revenue risk — decisions are made by intuition or account manager escalation
You have the data to build an early-warning system — but it has never been assembled into one
Purpose-built on a modern, enterprise-grade stack — designed to integrate with your existing systems without replacing them.
Let's talk about what Bricklay can build for your business.
