Creating a Risk Score Model for Customer Retention

Jessica just got promoted. She’s no longer our contact. We lost a $120K account because of that one promotion. It still hurts and it’s the reason we now measure something I call the 'Champion Dependency Score.' Jessica was our champion on a $120K account. She knew our platform inside out, trained her team, fought for us at renewal. For 18 months, every conversation went through her. She fixed every problem and handled every renewal question. Her replacement? Had never even heard of us. Then came the email that made it worse: “Hi Kelly, I’m taking over for Jessica. Can you send me some information about what you do for us? Also, we’re reviewing all our software spend this quarter.” That’s when it hit me: we didn’t have a customer relationship. We had a Jessica relationship. The panic audit I did that night: Single Point of Failure Analysis for Jessica's account: - Who introduced us to the company? (Jessica: Yes) - Who runs internal training? (Jessica: Yes) - Who handles our invoices/renewals? (Jessica: Yes) - Who do we text when there's an issue? (Jessica: Yes) - Who attends our quarterly business reviews? (Jessica: Yes) Score: 5/5 = 100% dependency. We were in trouble. I checked the rest of our portfolio. Out of our top 20 accounts, 14 had scores over 70%. That’s $1.8M ARR hanging by a single person in each account. Three months later, Jessica’s replacement churned us. Despite our best efforts, they went with a “simpler” solution. Translation: no one left inside could explain why we mattered. That loss forced us to create the Champion Dependency Score: - 0–20% → We can survive a departure - 21–50% → Some risk, manageable - 51–80% → High risk — widen relationships now - 81–100% → Crisis mode — one change away from churn What we do differently now: - Every meeting: 2+ customer attendees - Map contacts in IT, Finance, End Users, Leadership - QBRs include the champion’s boss - Track “relationship breadth” as part of our health score The hardest lesson: The stronger your champion feels, the more dangerous it is. Jessica loved us so much she kept it all on her plate and when she left, so did our value. Now I ask my CSMs: If your main contact vanished tomorrow, who else would fight to keep us? If the answer is “no one,” you’re not managing a customer. You’re managing a single point of failure. Who’s lived through their own “Jessica moment”?

“Dear John, I see it all now that you're gone” 🎤 “Dear John” by Taylor Swift It’s always humbling when statistics beat your own judgment. I’ve seen this personally, as my Pittsburgh Steelers fall back down to earth and barely sneak into the playoffs. The same is often true for predicting renewal rates for #CustomerSuccess leaders. While human judgment can beat data in assessing an individual customer, the “machine” 💻 often beats our intuition in estimating aggregating Gross Retention Rates (GRR). We’ve learned this even at Gainsight. We use an internal health score methodology (which many of our clients use as well) to assess retention risk, labeled “DEAR” (inspired by “Dear John”???) DEAR stands for: 🚦 Deployment: Number of licenses deployed / number of licenses sold. 🚦 Engagement: Recency of executive connection. 🚦 Adoption: Mix of % of licenses with active usage and % of features used. 🚦 ROI: Recency of a verified outcome from a customer, as logged in Gainsight. All of these are blended using our AI health scoring feature. Every year, our data magician Mark Peterson backtests the above signals versus actual renewal data. Importantly, he looks as those values 6 months prior to renewal (so as not to have leakage in the model). We then set our annual Gross Retention plan using the DEAR scores, with adjustments for large clients with known statuses (e.g., known churns). A few times, I personally decided to override the statistical score with my judgment to set plan. Each time, we missed plan 🤦🏽♂️. The DEAR score-based prediction has always been highly correlated to renewal rates almost every year. The periods where it missed were ones where there was a significant change in the “macro” economy. How do you predict your GRR? How accurate is your approach?

🚨 Stop guessing why customers churn. Start predicting and preventing it—with AI. Retention isn’t just a KPI. It’s a competitive moat—if you know how to build it. I’ve seen firsthand how retention turns from reactive to predictive when you fuse advanced data science with sharp business strategy. 🚀 5-Step AI/ML Retention Playbook 🔍 1. Integrate CLV-Powered Data Architecture 🔗 Unify transactional, behavioral, and sentiment data. 📉 Double down on features driving lifetime value erosion. 💼 Value Prop: Aligns spend with long-term profitability. 🤖 2. Build Explainable Churn Models 🌳 Use SHAP values with gradient-boosted trees. 🧪 Validate with causal inference, not just correlations. 💡 Value Prop: Creates defensible IP through interpretable AI. 🎯 3. Dynamic Risk Segmentation ⚡ Score users in real-time across engagement, fit, and payment health. 🚨 Trigger interventions at 85%+ confidence. 📊 Value Prop: Reduces CAC payback by 22%. 💡 4. Prescriptive Retention Engines 🧠 Reinforcement learning > static rule sets. 🎁 Test personalized win-backs based on elasticity modeling. 📈 Value Prop: +400bps lift from hyper-targeted nudges. 🔄 5. Closed-Loop Analytics Flywheel ♻️ Let intervention results train your models. 💰 Measure marginal ROI per dollar across segments. ⚙️ Value Prop: Retention becomes a growth engine, not just a metric. 💬 Want to put this playbook into action? Let’s connect—I'm always up for a deep dive into AI-driven growth. 👇 What’s one unexpected retention tactic that worked wonders in your org? #AI #MachineLearning #CustomerRetention #CTOInsights #SaaS #GrowthStrategy #GenerativeAI #PredictiveAnalytics #Leadership #DigitalTransformation #ProductStrategy #DataScience #BusinessGrowth #RetentionStrategy #B2BTech #TechLeadership #MLops #CustomerSuccess

𝗘𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 𝘆𝗼𝘂'𝗿𝗲 𝘁𝗿𝗮𝗰𝗸𝗶𝗻𝗴 𝗶𝘀 𝘁𝗼𝗼 𝗹𝗮𝘁𝗲. 𝗔𝗻𝗱 𝗶𝘁'𝘀 𝘄𝗵𝘆 𝘆𝗼𝘂𝗿 𝗿𝗲𝘁𝗲𝗻𝘁𝗶𝗼𝗻 𝗶𝘀𝗻'𝘁 𝗶𝗺𝗽𝗿𝗼𝘃𝗶𝗻𝗴. In Customer Success, most teams rely on lagging indicators:  • Logins dropped  • Customer ghosted for 30+ days  • Customer says, “𝘞𝘦’𝘳𝘦 𝘦𝘷𝘢𝘭𝘶𝘢𝘵𝘪𝘯𝘨 𝘰𝘵𝘩𝘦𝘳 𝘷𝘦𝘯𝘥𝘰𝘳𝘴…” These are all accurate signs of churn. But by the time they show up? 𝗧𝗵𝗲 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗵𝗮𝘀 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗳𝗮𝗶𝗹𝗲𝗱. 𝗬𝗼𝘂’𝗿𝗲 𝗻𝗼𝘄 𝗶𝗻 𝘀𝗮𝘃𝗲 𝗺𝗼𝗱𝗲 — Low odds. High effort. And emotionally draining for your team. Here’s the fix: 𝗦𝘁𝗮𝗿𝘁 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗼𝘄𝗻 𝙡𝙚𝙖𝙙𝙞𝙣𝙜 𝗶𝗻𝗱𝗶𝗰𝗮𝘁𝗼𝗿𝘀. Few companies ever create them because the truth is: Leading indicators don’t naturally exist in your business. You have to 𝗰𝗿𝗲𝗮𝘁𝗲 𝘁𝗵𝗲𝗺. So what 𝘪𝘴 a leading indicator? It’s a signal that appears before the customer knows they’re in trouble (that you know is associated with churn risk) And while you still have the time (and relationship strength) to do something about it. 𝗛𝗲𝗿𝗲’𝘀 𝗵𝗼𝘄 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱 𝘁𝗵𝗲𝗺: 𝟭. 𝗦𝘁𝘂𝗱𝘆 𝘆𝗼𝘂𝗿 𝗯𝗲𝘀𝘁 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀: Pick 2–3 accounts with clear, measurable success.    𝟮. 𝗜𝗱𝗲𝗻𝘁𝗶𝗳𝘆 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲𝘆 𝗱𝗶𝗱 𝗲𝗮𝗿𝗹𝘆: What were the key processes, decisions, or actions that led to their results?    𝟯. 𝗔𝘀𝘀𝗶𝗴𝗻 𝘁𝗶𝗺𝗶𝗻𝗴: When should every new customer do the same things? (7 days in? 30 days in?)    𝟰. 𝗕𝘂𝗶𝗹𝗱 𝗰𝗵𝗲𝗰𝗸𝗽𝗼𝗶𝗻𝘁𝘀: Use your product, CRM, and QBRs to confirm: Did it happen? Or is a risk forming? 𝗘𝘅𝗮𝗺𝗽𝗹𝗲 𝗳𝗿𝗼𝗺 𝗼𝗻𝗲 𝗼𝗳 𝗺𝘆 𝗰𝗹𝗶𝗲𝗻𝘁𝘀 (𝗲𝗺𝗮𝗶𝗹 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝗦𝗮𝗮𝗦): Leading indicators of future churn:  • No single owner of email strategy assigned - by Day 7  • No sales email sent - by Day 7  • Marketing calendar isn’t filled out 6 weeks ahead by Day 90 If any of these are missing — no, the customer won’t churn tomorrow. But they 𝘸𝘪𝘭𝘭 eventually. And if you wait, you’ll be stuck in a low success rate “save mode”. When you act on leading indicators:  • You prevent problems while they’re still small  • You keep your team proactive, not panicked  • You shift retention from firefighting to forecasting 𝗪𝗵𝗮𝘁 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗹𝗲𝗮𝗱𝗶𝗻𝗴 𝗶𝗻𝗱𝗶𝗰𝗮𝘁𝗼𝗿𝘀 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗿𝗶𝘀𝗸? #customersuccess

I was tired of guessing, and being wrong. Here's how I'm using AI to build customer health scores. As someone who's used Customer Success software for over 10 years and works with companies to design their health scores, I can tell you, this has always been a challenge. Most folks were working off assumptions, copying what others had done, or over-engineering scores thinking more inputs meant more accuracy. We’ve all seen it: ✅ Green customers churn ❌ Red customers renew And every time, we scratch our heads and ask ourselves, what are we getting wrong? This doesn't make sense. AI can give us the answer. It allows us to look at everything ... who our customers are, how they behave, what they need, and what they actually do. And from that, we can build truly intelligent profiles of health. No more guessing. Here’s a 5-step process that I used to redefine health: 1️⃣ Redefine your segments Move beyond spend-based segmentation. Segment by journey stage, product use case, or engagement pattern to get more meaningful insights. 2️⃣ Enrich your data Pull together all available data, product usage, support interactions, sentiment signals, firmographics, and demographics. The richer the picture, the better the model. 3️⃣ Label your historical outcomes Identify which customers renewed, expanded, or churned over the past 12–24 months. These become your training labels. 4️⃣ Run AI modeling Use AI to analyze patterns across your segments and outcomes. Prompt it to define health indicators tied to success and risk. 5️⃣ Operationalize in real time Build the model into your workflow. Let it learn and adapt as new data comes in so your health score always reflects what’s actually happening, not what you assumed. The goal isn’t to be perfect. The goal is to be accurate enough to act with confidence. Bonus: Loop in your CS teams to validate and pressure test the output. They’ll help refine the model and drive adoption. What’s powering your health score today ... insights or assumptions?

💡📈 I have BAD news for those of you who slept through Statistics 101: REGRESSION MATTERS! 📈💡 As SaaS companies strive for best-in-class customer net revenue retention (NRR), they must unpack what I call the "sticky drivers" of NRR: what customer attributes and traits yield the stickiest retention curves? DO. THE. MATH. The metrics in the example summary below are not exhaustive, but offer a sense of the types of attributes to consider. For Company X (which services ecommerce customers), a customer's ARR was not a significant predictor of NRR, but many other qualitative factors were: leveraging integrations, attending training sessions, using multiple products, etc. More importantly though, the analysis considered customer and product performance metrics in addition to just binary qualitative traits, which proved very revealing. The % change in the client’s monthly site visits is an intuitive customer health metric: when the customer’s volume declines, so does their net retention rate. The % change in monthly ARR/visits is a proxy for customer value: the lower the multiple, the stickier the customer. In my experience, the most meaningful attributes are VERY unique to any one specific business, so figuring out which variables to assess makes for a great cross-functional working session within your company. But here's my hot tip: when brainstorming that list of potential sticky drivers, be sure to include a price/value metric in your analysis, as that metric is very often a leading indicator of churn risk. Read more about NRR-specific analysis here: https://lnkd.in/geTuUa7a

Why does churn so often surprise us? Because we’re measuring health wrong - “Healthy” customers suddenly churning We’re not seeing the whole picture. - Not getting ahead of it in time We look at: - Number of support tickets - Product usage metrics Without looking at: - HOW they use the product - Executive engagement - Champion turnover - Payment history - Etc. Waiting until 30 days before renewal - To find out they’re churning - When it’s too late to save As an AE at Salesforce, I got to see how this machine can work when running on all 8 cylinders. I could see everything outlined above and then some, throughout the year, to not only get ahead of the renewal but also to expand the account. I could see how they were using or not using Salesforce and infer what might be happening. I could get a holistic view many B2B SaaS companies don’t have. I could form a point of view. More importantly, I could score my accounts and plan accordingly. Creating a Customer Health Score is one of the best ways to improve renewals, and a prerequisite to having a well oiled renewal process. But, the best health scores aren’t built in a silo. They’re reverse-engineered from churn. - What patterns showed in usage, behavior, payment, or sentiment? - What did churned accounts look like 90 days before they left? - What data did you already have — but didn’t act on? Start there and then iterate. - Just like a Lead Score - It's not set it and forget it - It requires lots of fine tuning A good CHS doesn’t just tell you who’s healthy today. - It tells you who’s about to churn tomorrow. In tomorrow’s newsletter we’ll breaks this down in more detail — including how to weigh metrics, validate the score, and make it actionable across GTM. Get the high-res image below and a thorough breakdown in tomorrow’s 📰 𝙍𝙚𝙫𝙊𝙥𝙨 𝙒𝙚𝙚𝙠𝙡𝙮 📰 Subscribe to get it here: https://bit.ly/49RCm0h ✌️ Which signals have been most predictive of churn for you? What would you add to this? 🤔