Your health scores are lagging indicators
Most SaaS health scores are built on login frequency and ticket count. By the time those turn red, the customer has already mentally churned. Predictive models identify risk 30–60 days earlier.
+12%
Net revenue retention (Northgate Capital, 90 days)
−31%
Logo churn rate after health score deployment
+45%
Expansion ARR from AI-surfaced upgrade triggers
[ Intro ]
The SaaS metrics that matter are downstream of decisions your AI should be making.
SaaS revenue is a compounding system. Win rate, onboarding completion, feature adoption, expansion triggers, renewal timing — every one of these data points sits in your product analytics, your CRM, and your billing platform. The problem isn’t lack of data. It’s the inability to synthesise it in time to act.
Your customer success team manages 80 accounts each. They can’t read every product session, every support ticket, every billing anomaly, and every NPS score for every account simultaneously. A unified AI health model can.
The same principle applies to expansion. Every one of your accounts has a next-logical-step. AI surfaces all of them — ranked by deal size and probability — before the customer has even asked.
[ The problem ]
Why B2B SaaS RevOps breaks at scale.
Acquiring a new customer costs 5–7× more than retaining an existing one. Yet most B2B SaaS businesses invest the majority of their growth budget in acquisition, while retention runs on an under-resourced CS team and a health score that’s one spreadsheet away from being useless.
As your account base grows, individual attention per account shrinks. Expansion opportunities get missed. At-risk accounts get identified too late. NRR stagnates.
Your CSMs are managing too many accounts
With 60–100 accounts per CSM, proactive outreach is rationed by gut feel. The quiet accounts get no attention until too late.
Expansion revenue is event-dependent
Upsells happen at renewal and when customers ask. An expansion agent surfaces upgrade triggers continuously — based on usage, team size changes, feature adoption gaps.
Onboarding is one-size-fits-all
Every new customer gets the same sequence regardless of use case or activation risk. Time-to-value is inconsistent.
RevOps data is siloed across systems
Product in Mixpanel. CRM in HubSpot. Billing in Stripe. Support in Intercom. No single view of the customer.
[ AI Plays ]
Three AI plays that move NRR in B2B SaaS.
Every play targets a specific revenue lever. Most clients build the health score first, then expand to the expansion agent.
Unified Customer Health Score Model
We pull from product analytics, CRM, billing, and support to build a single predictive health score for every account — updated in real time, visible in HubSpot or Salesforce. Trained on your historical churn patterns.
Expansion-Play Agent
A continuous expansion intelligence layer that monitors account usage, team growth signals, feature adoption gaps, and CRM notes to surface a weekly prioritised pipeline of expansion opportunities.
AI-Personalised Onboarding
Onboarding sequences that adapt to the customer's use case, role, and activation behaviour. Enterprise customers with low early-session depth get automatic intervention. Power users get fast-tracked.
[ How it works ]
How a SaaS RevOps engagement works.
Data & RevOps Audit (Weeks 1–2)
We map your data sources, CRM architecture, and current health scoring logic. We identify the highest-ROI first build.
Model Build (Weeks 3–6)
Predictive model trained on your historical churn data. CRM integration. Workflow automation for at-risk account alerts.
Deploy & Calibrate (Weeks 7–8)
Model live in your CRM. CSM workflow updated. First weekly reports generated. Calibration period to tune thresholds.
Ongoing optimisation
Monthly model retraining with new data. Quarterly review of intervention effectiveness. Expansion agent layered when ready.
We built the health score model and deployed it in about 8 weeks. In the next quarter our churn dropped significantly — not because we were doing anything dramatically different, but because we were doing the same interventions earlier, on the right accounts.
London Property Group 3× Their Qualified Leads in 90 Days
↗ 3× qualified pipeline in 90 days
Vantage Hotels Lifts RevPAR 18% With AI Rate Engine
↗ RevPAR +18% in 6 months
Northgate Capital Builds Connected RevOps in 90 Days
↗ Net revenue retention +12%
Houseguest Cuts Itinerary Quote Time by 55% With AI Builder
↗ Time-to-quote down 55%
Lumen Retail Ships 4,200 Product Pages in 6 Weeks
↗ 4,200 pages in 6 weeks
E-commerce Fashion Brand Scales to £2M Revenue With AI-Powered Retention
↗ £2M ARR within 2 quarters
[ FAQ ]
Common questions.
What data sources does the health score model need?
At minimum, product event data (any analytics platform), CRM activity (HubSpot, Salesforce), and billing data (Stripe, Recurly, Chargebee). Support data adds precision but isn't required to start.
How long does it take to build a health score model?
Typically 6–8 weeks from discovery to live deployment, including data integration, model training, and testing. The first version is built to be refined — it improves with every 30-day cycle.
We're on HubSpot — will the health score surface there?
Yes. HubSpot is our default CRM integration. Scores surface as a custom contact/company property, updated automatically. High-risk accounts trigger workflows.
At what ARR stage does this make sense?
We typically start with clients at £1M+ ARR with at least 30–40 paying accounts. Below that, volume doesn't justify the infrastructure.
Can you work alongside Gainsight or Totango?
Yes. We've integrated with both, and can feed signals into either platform. Clients often find Zinerge models produce more accurate scores than built-in scoring.
The account that's about to churn is already showing the signals.
A 30-minute fit call gives us enough to tell you whether your current data stack can support a predictive health model — and what the first 90 days would look like.