Introducing @siriusai. AI retention model. We’ve proven that trust is more profitable than friction. Since August, we’ve helped companies like Promova generate millions in revenue. Comment Sirius to get an exclusive look at the platform ✨

We've analyzed thousands of cancel events this year. The reason users give is rarely the real reason. The survey at the cancel screen tells you what the customer WANTS to say. The behavioral data tells you what HAPPENED. ↳ "Too expensive" ...almost always means the customer stopped using the product weeks ago. The price would have been fine if the value was still showing up. ↳ "I have everything I need" ...usually means they never got to the second use case. The product worked once and then disappeared from their week. ↳ "Switched to a competitor" ...is the easiest answer to give and the least likely to be true. Most users who pick it stopped opening either tool weeks ago. The reason for most cancels lives in the engagement curve - 30 days before anyone clicks the cancel button. The customer was already gone. The survey only confirmed it. The companies that win retention stop building product around the answers on the cancel screen. They build it around the moment the user quietly stopped showing up. That moment is your real churn signal. The cancel page is the receipt.

Reminder: Renewal is not won or lost in week two. Obsess a little less about renewal date. By the time the invoice fires, the customer had already decided months ago. Retention is a memory of how the early days felt. Great onboarding = great retention

Save rate is the easiest metric to fake. Hand everyone a 50% discount and you smash your goal. A real save keeps the customer AND the margin.

Half your churn never chose to leave. A card expired. A bank declined. A renewal quietly failed. You are probably losing customers who still want you. Fix the billing before you blame the product.

Sirius is hosting a private evening in San Francisco on June 24. Leave with an Apple AirTag in your pocket. Sirius After Hours, June 24 in SF. Live product demo, board games, drinks, and custom merch. Seating is limited and by approval: luma.com/wv7e31n3

Flash discounts only teach customers one lesson: “If I wait long enough, the price will drop” You are actively training your buyers to hesitate.

Pouring acquisition into a leaky funnel is the most expensive way to stay the same size.

Logo retention tells you who stayed. Net revenue retention tells you whether staying was worth it.

Hot take from inside an AI retention company: The biggest unlock is an agent making judgment calls in real time. ↳ Save this one with a premium offer ↳ Pause that one for 60 days ↳ Let this one walk away clean Most teams treat every churner the same.

I look at retention dashboards for a living. Most measure how many customers stayed. The useful ones measure who stayed. A 95% retention rate of unprofitable users is a death sentence dressed as success.

For two decades the retention motion looked identical at every subscription company. ↳ The old motion a customer hits the cancel flow a CS rep gets pinged they open the account in a CRM that loads slowly they scan three months of activity they read the canned save script they offer a discount or a downgrade they get five minutes, maybe ten, to keep the revenue This model was already broken before AI showed up. ↳ Why the human save team loses The rep sees a fraction of the signals. They work from a script written for an average customer who has never existed. They are incentivized on save rate, which means they push discounts that destroy LTV. They handle one cancellation at a time while ten more are happening in parallel. The human save team is a horse-drawn carriage in a market that started shipping electric vehicles. ↳ What AI does at the cancel moment reads every signal in the subscriber's history in milliseconds compares the subscriber to ten thousand similar cohorts personalizes the offer to the actual cancellation reason runs in parallel across every cancellation event in your billing system learns from every outcome and improves week over week A human save rep stands no chance against this on speed, scope, or accuracy. ↳ The economics finish the argument A CS rep saving subscribers at fifteen dollars an hour costs more per save than an AI that runs on cents. The CS rep takes vacation, gets sick, leaves the company, requires training. The AI runs at three in the morning when most cancellations actually happen. ↳ The wrong asset to build When a CMO tells me they are scaling their retention team by hiring more humans, I tell them they are building the wrong asset. The save team of 2026 looks like one strategist and a system that runs autonomously underneath them. Sirius is one of the systems being built for that future. There will be others. The CMOs who staffed up before noticing the shift will spend the next two years quietly cutting the same hires.

Ad spend does NOT earn customers. It rents them. The lease ends the day a competitor outbids you. Retention is the only growth channel you actually own.

I've spent the last year staring at growth org charts. ↳ Marketing owns acquisition ↳ Product owns features ↳ Engineering owns reliability ↳ Retention sits in a shared folder That last line is the bug behind your churn rate.

The win-back email is a confession. “We knew you were unhappy enough to leave” “We waited until you were gone to do something” Retention is the discipline of speaking up earlier.

30% of your churn is not a customer decision. It's a card on file that quietly expired. Most subscription brands assume churn is a story of disappointment. The product failed The pricing slipped A competitor moved faster The data tells a different story. A real share of every churn report is involuntary. Cards expire without warning. Banks decline international charges as fraud. Renewal attempts get blocked by the same security systems that are supposed to protect the customer. The customer never opened a cancellation flow because they never tried to leave. The system just...removed them. The brand counts that loss the same way it counts a deliberate cancel. This is the part that should bother every CFO in subscription: The customer was still happy. They wanted to keep paying. The infrastructure between them and the renewal collapsed and nobody told anyone. The recovery here is the easiest revenue in subscription ↳ Retry the charge on a schedule that matches the issuer's reset window ↳ Email the customer before the third decline, before they assume the service is over ↳ Surface the payment update inside the product, where they actually live ↳ Use card network data to predict expirations 60 days in advance None of those require a discount or a save offer (the customer already wants to stay.) The reason this problem persists is organizational. Billing lives under finance. Retention lives under growth. Neither team treats the renewal as a product surface. Stripe sits on the largest churn dataset in software history and almost no founder has ever read theirs. The companies that fix this recover meaningful retention from inside the customer base they already won.

Counterintuitive truth I've learned from analyzing 100s of cancel flows: Making your cancel button easier makes retention climb. ↳ Friction breeds resentment ↳ Resentment compounds across cohorts ↳ Quiet rage churns louder than visible rage Stop hiding the door.

Churn gets decided in week one. The cancel screen is the autopsy. Subscription brands obsess over the cancel click. They pour money into all the usual suspects. offer engines exit surveys win-back flows save scripts All of it fires after the decision was already made. By day seven your subscriber has built a habit or written you off. ↳ Most churn is mislabeled More than half the people churning this quarter checked out in week one. You never saw it happen. You saw them cancel three months later and called it a retention problem. It was an onboarding problem you mislabeled. ↳ Pull the week-one log Look at any subscriber who churned in month three. Pull their activity log from week one. The dropoff started on day three or four. Always one of these things: a feature they failed to find a value moment that never happened a confusing email that taught them the product was harder than promised The damage was done before they ever opened the billing dashboard. ↳ The strongest signal lives upstream At Sirius we run cohort analysis across the full lifecycle. The strongest retention signal has nothing to do with price or the offer at cancellation. It is the depth of activation in days one through seven. Subscribers who hit multiple activation milestones in week one churn at a fraction of the rate of those who never get there. ↳ Move the budget upstream The real retention spend belongs upstream. onboarding sequencing activation milestones second-session triggers the moment the product earns its slot in a crowded subscription stack Save attempts at the cancel page are damage control. The retention work that compounds happens before the customer ever considers leaving.

Cancel or stay is a false choice. Pause is the often-skipped option. When a subscriber hits your cancel flow your product gives them two doors. Keep paying or walk away. Real life has a third option and you stripped it out of your UX. ↳ The actual reasons people cancel a subscriber goes on vacation a subscriber gets laid off a subscriber finishes the season of the show they signed up for None of these people want to cancel. They want a break. Force them into the binary and they pick cancel. Cancel is the only exit you offered from a billing cycle that has stopped working for them. You lose the customer entirely instead of losing them for ninety days. ↳ The win-back economics are brutal Reacquiring a churned subscriber costs five to seven times what retaining one does. Roughly half of them never come back at any price. ↳ What pause unlocks A pause option flips the equation. Twenty to forty percent of would-be cancellations convert to pause when the option is offered well. The subscriber returns automatically when the pause ends. And everything stays in place. ↳ The reason most brands skipped it Their billing stack treats every subscription as a binary object, active or canceled, with no state in the middle. Stripe, Chargebee, and Recurly all support pause natively in 2026. The blocker is product priority. Product priority is set by whoever owns retention. If retention sits inside growth marketing, pause never makes the roadmap. If retention sits with the founder, pause shipped last quarter ↳ The fastest lift we deliver At Sirius the first integration we run with every new customer audits their pause capability. In most cases the feature exists in the billing layer and was never exposed at the cancel screen. The fastest retention lift we deliver is turning that switch on.

If a retention vendor takes payment upfront… no matter their save rate… YOU are carrying their risk. We built Sirius on the opposite model. Pay after the save. You've already lost customer. We only get paid when we recover him.