--- title: Spot customers at risk of churning description: Kite finds the warning signs that came before your past cancellations, checks every paying customer against them each week, and drafts the message most likely to keep each one. slug: /use-cases/churn-risk-detection page_type: use-case lead_prompt: How do I spot customers at risk of churning with AI? target_prompts: - Can an AI marketing agent spot customers at risk of churning? - How do I predict which customers are about to cancel? - What are the early warning signs of customer churn? - How do I find accounts that have stopped using our product? - How can I reduce churn by catching at-risk customers early? - predictive analytics for marketing - marketing next best action entity: Kite category: AI marketing agent author: Devon Wells and Kite date_published: '2026-11-12' date_modified: '2026-11-12' related_integrations: - /integrations/stripe.md - /integrations/posthog.md - /integrations/amplitude.md - /integrations/mixpanel.md - /integrations/supabase.md --- # Spot customers at risk of churning ## Key Takeaways - Most customers who cancel showed warning signs weeks earlier. They logged in less, stopped using the feature they bought for, removed teammates, downgraded, or let a card payment fail. - The warning signs that matter differ for every product, so the best way to find yours is to look at what your past cancellations did before they left. Kite does this with your Stripe billing history and your product data in PostHog, Amplitude, Mixpanel, or Supabase. - Kite checks every paying customer against those warning signs on a schedule, ranks the ones slipping by how much revenue is at stake, and explains what changed for each account. - For each at-risk customer, Kite picks the next step that fits the reason they're slipping and drafts the message, whether that's a personal check-in, a guide to a feature they never set up, or a reminder to update a failed card. - A few weeks later, Kite checks who stayed and who left, and uses the results to sharpen the warning signs. ## Why do customers churn without warning? Customers rarely leave without warning. The warning just shows up in places nobody is watching. Churn, also called customer attrition, is the share of your customers who cancel over a period. If you start a month with 100 subscribers and 5 cancel, your monthly churn is 5%. For most small subscription businesses, the cancellation email is the first time anyone notices, and by then the decision is already made. The signs usually appear weeks earlier, spread across two systems. Your product data shows people logging in less often, skipping the feature that made them buy, or removing teammates from their account. Your billing data shows downgrades, a subscription set to cancel at the end of the period, or a card that keeps declining. Each sign is easy to see on its own. What's hard is checking every customer against all of them, every week, and then doing something useful for each one. That's what people mean by predictive analytics in marketing: using what customers did in the past to estimate what they'll do next. Large companies build statistical models to score every customer's risk. A small team can get most of the value without building a model, by finding the handful of behaviors that came before its own cancellations and watching for them. The work usually means someone has to: - work out which behaviors actually come before a cancellation in your product - pull billing and usage data together for every customer on a regular schedule - decide which slipping customers matter most and why each one is slipping - write a message that fits that reason, rather than one generic "we miss you" email - check later whether the outreach kept anyone The work is mostly careful reading of data and careful writing, and the results are easy to measure because a customer either renews or doesn't. That makes it a good job for an AI agent. Kite is an AI marketer. It researches your business, builds a growth strategy, and ships the work that brings you new customers. Here's how it finds the customers you're about to lose and helps you keep them. ## How Kite spots customers at risk of churning Kite starts with the customers you've already lost. It reads your Stripe history to find who canceled and when, then reads your product data to see what those customers did in the weeks before they left. It compares them with the customers who stayed and proposes a short list of warning signs, with the numbers behind each one. Once you agree on the warning signs, Kite checks your current customers against them on a schedule you choose. It ranks the customers who are slipping by how much revenue they bring in and how many signs they show, and it explains what changed for each account. It then decides the most useful next step for each customer and drafts the message. After you approve, Kite sends the messages and keeps the list current. A few weeks later, it checks which of the customers it flagged renewed, recovered, or canceled, and it uses those results to adjust the warning signs. ### What are the early warning signs of customer churn? The early warning signs of churn are changes in how a customer uses and pays for your product, measured against their own normal behavior. A customer who used to log in daily and now logs in once a week is a stronger warning than a customer who has always logged in once a week. Common signs include: - **Falling activity.** Fewer logins, shorter sessions, or fewer of the actions that make up your product's core job, such as reports created or orders processed. - **A dropped key feature.** The customer stopped using the feature that most of your long-term customers rely on, or never set it up at all. - **A shrinking team.** Teammates were removed, invites stopped, or only one person in a company still logs in. - **Billing changes.** The customer downgraded, removed seats, switched off annual billing, or set their subscription to cancel at the end of the period. - **Failed payments.** A card declined and the invoice is past due. This kind of churn is often accidental, and it's the easiest to save. Kite finds out which of these signs matter for your product by looking at your own history. With Stripe connected, it lists the subscriptions that ended over the past year and the dates they ended. With PostHog, Amplitude, Mixpanel, or Supabase connected, it lines up each canceled customer's activity in the eight weeks before they left with the activity of similar customers who stayed. If your canceled customers almost all stopped using a particular feature a month before canceling, while your loyal customers kept using it, that drop becomes one of your warning signs. Kite writes the result as a short list you can read in plain sentences. One example is "a paying account whose weekly active users fell by half compared with the previous month, or whose owner hasn't created a project in 21 days." It shows how many of your past cancellations matched each sign and how many loyal customers matched it too, so you can see which signs are reliable and which would flag too many happy customers. Once you agree, Kite saves the warning signs to your team's knowledge base. ### How do you find accounts that have stopped using your product? Kite finds the accounts that have stopped using your product by checking every paying customer against your warning signs on a regular schedule. A weekly check suits most small teams, and Kite can run it every weekday if your customers move quickly. On each run, Kite pulls your active subscriptions from Stripe, along with each customer's plan, monthly revenue, renewal date, and any recent downgrade or failed payment. It then reads each account's recent activity. In PostHog, it can query your events directly and save the at-risk accounts as a cohort, a saved group of users that updates on its own. In Amplitude, it reads each user's activity and your retention data, and can save the flagged users as a cohort your team can chart. In Mixpanel, it runs queries on your events and builds the same kind of cohort. In Supabase, it writes its own queries against the tables where your product records what customers do. Kite delivers each run's results to a Slack channel, by email from your team's Kite address, or as a conversation in the Kite web app. If a single event should never wait for the next check, such as a customer setting their subscription to cancel, Kite can react to it as it happens. It can react to your PostHog events as they arrive and check your Stripe account for that kind of change as often as once an hour, and it tells your team right away. ### Which at-risk customers should you focus on first? Focus first on the customers who bring in the most revenue and show the clearest decline, because a few saves there are worth more than many saves among small accounts. Kite ranks every flagged customer on both, so you're not working through a list in alphabetical order. For each one, Kite explains what changed in plain language and separates the facts it found from the conclusions it drew. A typical weekly list might read like this: - **Highest priority.** A design agency on your team plan, and one of your ten largest accounts. Its five active users fell to one over the past three weeks, nobody has opened the client-sharing feature since early October, and the renewal is in 18 days. - **Medium priority.** A solo customer on your starter plan whose card has declined twice. Their usage is steady, so this looks like an expired card rather than a decision to leave. - **Watch.** A customer who downgraded from the team plan to the starter plan last week. Their activity is normal for the smaller plan, so Kite notes the downgrade and keeps checking. Kite can also research the company behind a business account with its built-in sources. A customer that just announced layoffs or a merger is at risk for reasons your product data won't show, and knowing that changes what you say to them. ### What should you do when a customer is at risk of churning? The right next step depends on why the customer is slipping, so Kite picks a different action for each reason rather than sending everyone the same email. Marketers call this next-best-action marketing: deciding, one customer at a time, what to send next. Large companies run it on enterprise decisioning software. For a small team, it mostly comes down to matching the message to the reason. Kite drafts the message that fits each case: - **A customer who never set up a key feature** gets a short guide to that feature, written around what they already use. - **A team that's shrinking** gets a personal check-in from the founder or account owner asking what changed, with an offer to help. - **A failed payment** gets a plain reminder with a link to update the card. Kite takes that link from Stripe, and it opens the Stripe page where the customer can pay the overdue invoice with a new card. - **A high-value customer who has gone quiet** gets a personal email offering a call, and Kite tells your team that this one deserves attention. - **A customer who set their subscription to cancel** gets a short note asking what would have made them stay. If your team has agreed to a save offer, such as a discount on the next three months, Kite can include it and create the coupon in Stripe once you approve. Kite writes each message in your brand voice, using what it knows about your product, your customers, and what this particular customer did. It sends from your team's Kite address or a connected email account, and only makes offers your team has agreed to. Kite shows you each message before it goes out. Approve it with the button in the conversation or with a reply, and Kite sends it. ### Did the outreach keep anyone? Kite judges the work by whether the customers it flagged stayed, because that's the only result that matters. A few weeks after the messages go out, it checks Stripe for who renewed, who recovered from a failed payment, and who canceled anyway. It also reads your product data to see whose activity came back. Kite then compares the flagged customers who got a message with similar customers who didn't, so it can tell whether the outreach made the difference. It also checks how accurate the warning signs were. If customers who dropped the key feature canceled at a high rate but customers who logged in less often usually recovered on their own, Kite proposes giving more weight to the first sign. It records what it learned in your team's knowledge base, so the next round of checks and messages starts from those results. Kite can also publish a dashboard that shows your churn rate, the customers currently at risk, and the revenue your outreach saved, and refreshes it each time you open it. ## What you get - **A written list of warning signs.** The changes in usage and billing that came before your past cancellations, with the numbers behind each one, saved to your team's knowledge base. - **A regular at-risk list.** Every paying customer who matches a warning sign, ranked by revenue and strength of decline, with a plain explanation of what changed. - **A recommended next step for each customer.** The action that fits the reason they're slipping, from a feature guide to a personal call. - **Drafted messages, ready to approve.** An email for each at-risk customer in your brand voice, plus card-update links or approved save offers where they fit. - **Quick alerts for urgent changes.** A message to your team when a valuable customer sets their subscription to cancel or a payment fails. - **A results check.** Who stayed, who left, how much revenue the outreach kept, and the changes Kite recommends to the warning signs. ## How to get started Tell Kite what you want in plain language. You might say, "Find the customers who are likely to cancel this month and draft a message to each of them." You can work with Kite in the Kite web app or in Slack. Connect Stripe first, because it holds who pays you, how much, and who has already canceled. Then connect the place your product records what customers do, such as PostHog, Amplitude, Mixpanel, or Supabase. Kite can find billing-side risks such as failed payments and downgrades with Stripe alone, but it needs your product data to see who has stopped using the product. Kite proposes the warning signs from your history, a schedule for checking them, and a sample of the messages it would send. You adjust or approve them, and Kite starts checking. If your product doesn't record an activity Kite needs, it tells you which event to add and what details it should carry, and someone on your team adds it to your product's code. Kite asks before it sends outreach or creates a discount. Once you trust its judgment on who is at risk and what to say, you can let it send more of these messages on its own. ## Integrations used to spot customers at risk of churning Spotting churn risk takes two kinds of data: what customers pay you and what they do in your product. Kite reads both through the tools you already use, then uses its own email and research to act on what it finds. ### Read billing and cancellations - [Stripe + Kite](/integrations/stripe.md): Kite reads your subscriptions, invoices, and cancellations to learn what came before past churn and to spot downgrades, failed payments, and subscriptions set to cancel. It can also pull the link to each overdue invoice's Stripe payment page so customers can pay with a new card, and create coupons for save offers you approve. ### Read what customers do in your product - [PostHog + Kite](/integrations/posthog.md): Kite queries your events to find falling activity, saves at-risk accounts as a cohort that updates on its own, and reacts to important events as they happen. - [Amplitude + Kite](/integrations/amplitude.md): Kite reads each user's activity and your retention data, and saves the customers it flagged as a cohort your team can chart. - [Mixpanel + Kite](/integrations/mixpanel.md): Kite runs queries on your events to compare customers who left with customers who stayed, and builds a cohort from your warning signs. - [Supabase + Kite](/integrations/supabase.md): Kite queries your product database directly to measure each account's activity, team size, and use of key features. ## Common questions ### What does a 20% churn rate mean? A 20% churn rate means one in five of your customers canceled during the period you're measuring. If that's a monthly rate, it's very high, because you'd need to replace most of your customer base every year just to stay the same size. If it's an annual rate, it's common for small businesses selling to other small businesses. Kite calculates your churn from your Stripe data, so you can see your rate by month, by plan, and by when customers signed up. ### Do I need a data science team or a churn model to predict churn? No. A statistical model can rank risk more finely, but most small teams get the bulk of the value from a few clear warning signs drawn from their own history. Kite finds those signs, checks every customer against them, and explains each flag in plain language, so you can see why a customer is on the list and disagree if you know something Kite doesn't. As your history grows, Kite keeps testing which signs actually predicted cancellations and adjusts them. ### What's the difference between voluntary and involuntary churn? Voluntary churn is when a customer decides to cancel. Involuntary churn is when a subscription ends because a payment failed, usually from an expired or replaced card, even though the customer wanted to stay. Involuntary churn can be a large share of cancellations for small subscription businesses, and it's the easiest to fix, because a timely reminder with a link to update the card often solves it. Kite treats the two separately, since a customer with a declined card needs a payment reminder, not a "we miss you" email. ### How early can you spot a customer who is going to cancel? It depends on your billing cycle and how often customers use your product. For a product used every day, a drop in activity often shows up several weeks before a cancellation. For a product used once a month, the signs take longer to appear. When Kite compares your past cancellations with the customers who stayed, it also measures how far ahead each warning sign appeared, so you know how much time you have to reach out. ### Can Kite stop a customer from canceling? Kite can't make the decision for a customer, but it can make sure you reach them while there's still time to change their mind, with a message that addresses why they're slipping. It finds the customers at risk, explains what changed, drafts the right outreach, and measures how many it kept. Some customers will leave for reasons no message can fix, such as closing their business, and Kite's results check shows you which reasons your outreach can actually address.