At-Risk Customers

At-risk customers are existing accounts — typically in a subscription, recurring service, or long-term contract model — that exhibit signals suggesting an elevated probability of churn, non-renewal, or significant revenue contraction before or at the next commercial decision point. The defining characteristic of an at-risk customer is that the deterioration is detectable before the formal notification of cancellation or non-renewal. That lead time — the window between when risk becomes visible and when the customer formally exits — is the intervention opportunity that customer success and account management teams must exploit.

The business imperative for identifying at-risk customers early is straightforward in a recurring revenue model: acquiring a new customer costs five to seven times more than retaining an existing one, and a 5% improvement in retention rates typically produces a 25–95% improvement in long-term profitability. For B2B SaaS and technology companies with annual contract values in the tens or hundreds of thousands, a single prevented churn event can represent more margin impact than a new-logo sale of equivalent value, because the retained customer carries no acquisition cost.

At-risk identification is distinct from reactive churn management. Reactive churn management responds to the customer's cancellation notice — at which point the relationship has often already deteriorated past the point of recovery. Proactive at-risk management uses leading indicator signals — product usage trends, engagement patterns, support ticket volumes, stakeholder changes, payment behaviour, and sentiment signals — to identify deterioration months before the customer reaches a formal decision.

The signals that predict at-risk status fall into three broad categories. *Behavioural signals* reflect how the customer is (or is not) using the product or engaging with the vendor. *Commercial signals* reflect changes in the customer's commercial relationship — payment delays, contract disputes, downgrade requests. *Relationship signals* reflect changes in the human connection — champion departure, reduced responsiveness, executive disengagement. Sophisticated at-risk programmes monitor all three categories simultaneously, weighting signals based on their predictive value and the specific customer's historical patterns.

For revenue operations teams, managing at-risk customers requires a combination of data infrastructure — to surface signals from across the revenue tech stack — and operational process — to trigger interventions at the right time with the right resource. Signalon's analytics platform surfaces engagement signals from deal room interactions, contract activity, and content consumption that complement product usage and CRM data in building a complete at-risk picture. Signalon pricing includes customer success capabilities across all plans.

At-risk customers are described using several related terms across customer success, revenue operations, and account management contexts:

• Churn risk accounts — emphasises the binary outcome of the risk (the customer leaves or stays)
• Red accounts — common in customer success tooling, where accounts are colour-coded by health (green/yellow/red)
• At-risk accounts — the most common equivalent term; interchangeable with at-risk customers in most B2B contexts
• Churning customers — used when the risk assessment has concluded that churn is likely (a higher-certainty label than at-risk)
• Unhealthy accounts — used in customer health score frameworks where accounts are scored on a health index
• High churn risk — used in predictive analytics contexts to describe the output of a churn prediction model
• Rescue accounts — used internally at some companies to describe accounts that have been formally flagged for active intervention
• Flagged accounts — generic term for accounts requiring special attention, broader than at-risk

At-risk customer identification operates through three layers of increasing sophistication:

Layer 1 — Lagging Indicator Monitoring

The most basic approach monitors lagging indicators that confirm deterioration: non-responses to renewal outreach, formal downgrade or cancellation requests, late payments, or support escalations that mention competitor evaluation. These signals are easy to detect but appear too late — by the time a customer submits a cancellation request, retention success rates drop below 25% in most B2B environments.

Layer 2 — Leading Indicator Tracking

More effective programmes monitor leading indicators that predict future churn before the customer reaches a decision point. Key leading indicators include:

*Product usage signals:* Login frequency decline, feature adoption reduction, active user count drop (particularly among power users), session length decrease, API call volume reduction for integration-heavy products. Usage signals are the most reliable predictors of churn because they reflect whether the customer is extracting value from the product — the fundamental determinant of renewal intent.

*Engagement signals from the deal room and communication layer:* Reduced responsiveness to CSM communications, declining open rates on renewal and expansion proposals, reduced engagement with content shared in the deal room. Signalon's analytics platform tracks these engagement signals at the deal room level, providing visibility into which accounts are actively engaging with shared content and which are going dark.

*Support and service signals:* Spike in support ticket volume, tickets escalating in severity, repeated unresolved issues, or a sudden drop in support contact (which can indicate the customer has stopped investing in getting value and is planning to leave).

*Commercial signals:* Late invoice payment beyond established patterns, requests to reduce scope or downgrade, comparison shopping behaviour (requesting pricing information beyond their current tier), or involvement of procurement in routine renewal conversations.

Layer 3 — Predictive Churn Modelling

Advanced at-risk programmes use machine learning models trained on historical churn data to generate account-level churn probability scores. These models combine dozens of signals — weighted by their historical predictive value — into a single risk score that updates continuously as new data arrives. Accounts above a defined risk threshold are automatically flagged for CSM review. This approach identifies at-risk accounts 90–180 days before renewal, when intervention success rates are highest.

Health Score Frameworks

Most customer success platforms combine multiple signals into a composite customer health score — typically expressed as a number (0–100) or a colour (green/yellow/red). Health scores give CSMs a single number to monitor rather than requiring them to synthesise multiple data streams manually. The components and weightings of a health score must be calibrated to the specific business model and validated against historical churn outcomes to be predictively useful. A health score that does not predict actual churn is worse than no score, because it creates false confidence.

Pain Points: SaaS churn is the existential risk in subscription businesses. A 10% annual churn rate means the company must acquire 10% of its customer base in new logos just to maintain flat revenue — before accounting for any growth ambitions. Churn compounds destructively: a company with £5M ARR and 15% annual churn loses £750K annually from existing customers. Without a proactive at-risk programme, this loss is distributed across hundreds of accounts, making it invisible until renewal forecasts are built. By then, it is often too late for effective intervention.

Use Case: A B2B SaaS company with 400 active enterprise accounts deploys a health score framework combining product login frequency (weighted 30%), active user count as a percentage of contracted seats (25%), support ticket sentiment (15%), CSM communication response rate (15%), and deal room engagement with shared content (15%) — the last signal sourced from Signalon's analytics platform. Accounts scoring below 60 are automatically flagged as at-risk and assigned to a dedicated retention CSM. The programme identifies 47 at-risk accounts in Q1, of which 31 receive interventions within 30 days of flagging. Of the 31 intervened accounts, 22 renew — representing £890K of ARR preserved that would not have been identified without the programme. The 9 churned accounts churn at a smaller contract value on average, suggesting the intervention successfully retained the highest-value at-risk accounts.

Pain Points: Financial services clients represent high annual contract values, long sales cycles, and complex multi-stakeholder relationships. When a client is at risk, the consequences are severe — not just revenue loss but reputational impact, potential regulatory notification requirements for data offboarding, and significant relationship management overhead. At the same time, detecting at-risk signals is harder because financial services clients tend to be formally correct in their communications even when internally dissatisfied, masking early warning signals.

Use Case: A B2B fintech company selling treasury management software to corporate finance teams identifies at-risk accounts through a combination of product usage monitoring and stakeholder mapping. When the system detects that the primary finance director contact — who has been the champion for the relationship — has left the company (identified via LinkedIn integration with the CRM), the account is automatically flagged as at-risk due to champion departure. The CSM initiates immediate outreach to identify the new contact and rebuilds the relationship before the next renewal cycle. Of 12 accounts that experienced champion departure in a 12-month period, 9 were flagged automatically and received intervention; 8 of the 9 renewed. The 3 accounts not flagged (due to the integration not detecting the departure) had a 33% renewal rate — validating the signal's predictive value.

Pain Points: Manufacturing customers on multi-year service and maintenance contracts are traditionally considered stable — but at-risk signals still appear when equipment configurations change, when procurement takes over account management from the engineering team, or when a competitor's service offering is evaluated during a broader maintenance contract review. Equipment OEMs and industrial service providers often lack the digital engagement data that SaaS companies use for at-risk monitoring, making relationship quality signals and commercial behaviour more important.

Use Case: A precision equipment manufacturer identifies at-risk service contract customers through three primary signals: late payment (beyond 45 days becomes a risk flag), reduced call-out frequency relative to historical baseline (suggesting the customer has stopped using the equipment or has found alternative service), and non-engagement with renewal proposals in the deal room. When all three signals are present simultaneously, the account is flagged as high at-risk. The field service team is alerted and conducts a structured on-site visit — not a sales call — to understand the operational situation and rebuild relationship capital. This intervention programme reduces service contract churn from 14% annually to 8%, preserving £1.2M of annual recurring service revenue.

Pain Points: Professional services client relationships are inherently personal and difficult to monitor systematically. The primary at-risk signals are relational — the partner or principal managing the relationship is less engaged, the client is less forthcoming in planning conversations, or the scope of work is being reduced incrementally rather than expanded. These signals require experienced judgment to interpret, making systematic monitoring challenging. The departure of a key relationship holder on either side is a major risk event.

Use Case: A management consulting firm uses a structured account health review process — run quarterly by the account partner — to assess at-risk status across its client portfolio. Each account is reviewed on five dimensions: relationship depth (number of active relationships above associate director level on the client side), recent work quality (assessed by post-engagement survey scores), scope trend (expanding, stable, or contracting), budget position (is the client's relevant budget expanding or contracting), and competitive exposure (is the firm aware of any competitor engagement). Accounts scoring low on three or more dimensions are escalated to a formal account recovery plan, which includes a structured executive conversation, a strategic roadmap presentation delivered via the deal room, and a defined set of deliverables designed to rebuild value perception.

Pain Points: IT services companies managing managed service agreements face at-risk signals that combine technical performance metrics (SLA compliance, incident resolution times) with relationship and commercial signals. A customer who is experiencing repeated SLA breaches is at risk regardless of what their CSM relationship looks like. Conversely, a customer who is technically satisfied but whose procurement team has initiated a competitive review is at risk despite good service metrics. IT services companies must monitor both dimensions simultaneously.

Use Case: A managed IT services provider builds a composite at-risk score from four signal sources: SLA compliance rate (if below 97%, account scores amber; below 95%, red), help desk ticket sentiment trend (analysed weekly using NLP), CSM-assessed relationship quality (scored monthly on a 1-5 scale), and renewal proposal engagement from Signalon's analytics (accounts that have not opened the renewal proposal within 14 days of delivery are automatically flagged). Accounts with red overall scores are assigned to a dedicated account recovery team. The recovery programme includes a structured service review, an SLA remediation commitment delivered as a formal document in the deal room and signed via Signalon e-signature, and an escalation path to executive sponsorship if the standard CSM intervention does not produce response within 21 days. The programme reduces annual churn from 18% to 11% over two years.

• Significant revenue preservation at lower cost than new logo acquisition. Saving a churning account costs a fraction of acquiring a new account of equivalent value. For companies with annual contract values above £50K, a single successful retention intervention can represent more profit impact than closing a new-logo deal at equivalent value, because the retained account carries no marketing, SDR, or sales acquisition cost.
• Earlier intervention produces better retention outcomes. The success rate of retention interventions declines steeply as the customer approaches their formal decision date. Accounts flagged 180 days before renewal show retention success rates of 70–80% with effective intervention; accounts flagged 30 days before renewal show rates of 20–30%. Earlier detection from leading indicators dramatically improves the pool of recoverable accounts.
• Proactive identification improves revenue forecast accuracy. At-risk customers who are not identified skew renewal forecasts optimistically. When the renewal team assumes all un-flagged accounts will renew, the forecast overstates expected ARR. A systematic at-risk programme makes the forecast more accurate by surfacing accounts that should be discounted from the renewal assumption — enabling finance and leadership to plan more reliably.
• Root cause analysis drives product and service improvement. At-risk customer data is one of the richest sources of insight into product gaps, service failures, onboarding weaknesses, and pricing friction. Systematically analysing why accounts become at-risk — and what interventions succeed or fail — creates a feedback loop that improves the product, the onboarding experience, and the ongoing service delivery. This improvement reduces the future at-risk rate by addressing structural causes rather than just managing symptoms.
• Champion departure protection preserves multi-year relationships. Stakeholder change — particularly the departure of the internal champion who drove the original purchase decision — is one of the highest-risk events in a B2B relationship. At-risk programmes that monitor stakeholder changes and trigger immediate relationship-building interventions protect multi-year contracts from the natural risk created by role transitions. Signalon's deal room maintains a record of which stakeholders have engaged with shared content, helping CSMs understand the depth of their relationship footprint.
• Structured intervention processes improve CSM efficiency. Without a defined at-risk programme, CSMs manage risk reactively and inconsistently — some respond aggressively to subtle signals, others ignore significant warning signs until too late. A structured programme with defined triggers, escalation paths, and intervention playbooks ensures consistent, efficient responses across the team regardless of individual CSM experience level.
• Data from at-risk accounts informs pricing and contract structure decisions. Accounts that consistently show at-risk signals around renewal — particularly those citing price-value perception issues — provide actionable data for pricing strategy and contract structure. If 60% of at-risk accounts are triggered by a perceived value gap at a specific price tier, the product and pricing teams have clear direction for where to invest in value demonstration or pricing adjustment.
• Retention data strengthens investor and board narrative. Net Revenue Retention (NRR) and Gross Revenue Retention (GRR) are the primary metrics by which investors and boards assess the health of a recurring revenue business. A visible, systematic at-risk programme — with documented detection methodologies, intervention playbooks, and outcome tracking — demonstrates operational maturity and gives leadership confidence that the retention metrics being reported are the result of deliberate management rather than luck. Review Signalon's analytics capabilities for tracking retention metrics across the customer portfolio.

Product Usage Data: The most predictive signal category for most SaaS and technology businesses. Key metrics: daily/weekly active users as a percentage of contracted seats, feature activation rates, API call volumes, session frequency and duration trends. Usage data comes from the product's own analytics layer and must be integrated into the customer success platform.

CRM and Relationship Data: Contact activity logs, email response rates, meeting frequency, stakeholder role changes, and CSM assessment notes. CRM data provides the relationship quality dimension that product usage data cannot capture. Champion departure detection — identifying when key contacts leave the customer organisation — requires integration between the CRM contact records and HR data sources like LinkedIn.

Deal Room Engagement Data: Whether the customer is actively engaging with content shared in the deal room — renewal proposals, expansion proposals, product roadmap briefings, case studies — is a strong signal of relationship health. Accounts that stop engaging with shared content are often at risk. Signalon's analytics platform provides this engagement layer natively, surfacing which contacts are opening, reading, and sharing content, and which are going dark.

Support and Service Data: Ticket volume, severity trends, time-to-resolution trends, customer-reported satisfaction scores from post-ticket surveys, and NPS responses. Support data captures the service quality dimension of the relationship and can detect deterioration before it appears in product usage data.

Commercial and Financial Data: Invoice payment timing relative to terms, contract scope requests (downgrade or reduction), pricing discussions, and procurement involvement. Financial signals are lagging relative to product and engagement signals but are high-certainty indicators when they appear.

Sentiment and Voice of Customer Data: Net Promoter Score surveys, executive business review feedback, customer advisory board participation, and case study or reference willingness. Sentiment data is less continuous than usage data but provides qualitative depth that quantitative signals cannot.

Customer success platforms are the operational home for at-risk customer management, aggregating signals from multiple sources into health scores and triggering workflows.

• Health score calculation combining product usage, engagement, support, and commercial signals
• At-risk account flagging and automatic CSM task creation when risk thresholds are breached
• Intervention playbook management — defining the specific steps, owners, and timelines for at-risk account recovery
• Outcome tracking — recording whether flagged accounts churn, renew flat, or expand, enabling model calibration
• Support for Gainsight, Totango, ChurnZero, and Planhat

The CRM provides the account-level context, contact records, and commercial history that at-risk programmes depend on.

• Account health score surfaced directly in the CRM account record, enabling AEs and CSMs to see risk status without leaving their primary workflow
• Contact change monitoring — detecting when key stakeholders leave or change roles within a customer account
• Renewal opportunity management with at-risk flags visible to the sales team handling the renewal motion
• Escalation workflow integration — creating account recovery opportunities when at-risk flags are raised
• Support for Salesforce, HubSpot, Microsoft Dynamics, and Pipedrive

Product usage data is the most predictive signal category for most B2B at-risk programmes and requires clean integration between the product layer and the customer success platform.

• Real-time usage metrics by account and user, with trend calculations
• Feature adoption tracking — which capabilities are being used and which are ignored
• Active user identification and churn within the customer's user base (individual user churn within an account often precedes account-level churn)
• Usage benchmark comparison — showing how an at-risk account's usage compares to healthy accounts of similar size and industry
• Support for Mixpanel, Amplitude, Heap, and custom data warehouse integrations

Deal room engagement data provides a unique signal category — whether the customer is actively engaging with vendor-shared content — that is not available in product usage or CRM data.

• Per-contact engagement tracking on renewal and expansion proposals
• Alert triggers when accounts stop engaging with shared content or miss defined engagement milestones
• Visibility into which stakeholders are engaging (and which have gone silent) within an account
• Renewal proposal delivery and engagement analytics natively integrated
• Signalon deal rooms provide this engagement layer with full analytics visibility for customer success teams

E-signature workflows provide signals about the customer's commercial engagement and speed of decision-making.

• Time-to-sign tracking on renewal proposals — delays relative to historical patterns are an at-risk signal
• Completion triggers that update account health scores when renewals are signed (confirming recovery)
• Amendment and downgrade request tracking — early indicators of contract pressure
• Signalon's e-signature capabilities provide native tracking of contract engagement within the deal room

Email, call, and meeting activity data provides the communication engagement dimension of at-risk monitoring.

• Email response rate tracking by account — declining responsiveness is an early warning signal
• Meeting frequency monitoring — accounts that begin declining or cancelling regular check-ins are at higher risk
• Call recording and sentiment analysis — detecting frustration, confusion, or dissatisfaction in customer communications
• Slack and Microsoft Teams integration for accounts using those channels for vendor communication

• Signal selection over signal volume: More data is not always better for at-risk identification. A health score that aggregates too many signals with poor weightings is less predictive than a simpler score with three well-validated signals. Invest in understanding which signals actually predict churn in your specific customer base — validated against historical churn data — before building a complex multi-signal framework.
• Lead time as the primary design criterion: The value of an at-risk signal is proportional to how much lead time it provides before the customer reaches their formal decision point. Prioritise signals that appear 90–180 days before renewal over signals that appear 30 days before. This requires accepting some false positives (accounts flagged as at-risk that renew without intervention) in exchange for identifying the recoverable cases early enough.
• CSM capacity as the binding constraint: At-risk programmes generate interventions, and interventions require CSM time. If the programme flags 20% of accounts as at-risk but the team has capacity to intervene with 5%, the programme is incomplete. Design the risk score thresholds and intervention playbooks to match actual CSM capacity — or make the case for additional headcount based on the revenue at stake.
• Intervention playbook quality determines outcomes: Identifying at-risk accounts is necessary but not sufficient. The quality of the intervention — the relevance of the outreach, the authority of the escalation, the credibility of the proposed solution — determines whether the identification translates to retention. Invest as much in playbook design and CSM coaching as in the technical signal infrastructure. Signalon's deal room enables CSMs to deliver structured, high-quality intervention proposals backed by relevant content.
• Feedback loops are essential for model improvement: At-risk models improve when their predictions are validated against actual outcomes. Build systematic outcome tracking — recording whether flagged accounts churned, renewed, or expanded — and use this data to regularly recalibrate signal weightings. A model that is never updated against actual outcomes will degrade in predictive value over time.
• Avoid false precision in health scores: A health score of 67 and a health score of 68 are not meaningfully different. Present health scores in bands (green/yellow/red or high/medium/low) rather than precise numbers, and train CSMs to treat the score as an attention-directing tool rather than an absolute judgment. Overconfidence in score precision leads to both underreaction to borderline cases and dismissal of accounts that score marginally above the threshold but are actually at risk.