Subscription Churn Prediction and Winback Strategies
How to identify at-risk customers before they cancel and bring them back with automated retention flows
Arpit Mehar
Subscription Churn Prediction and Winback Strategies
Most subscription brands wait for churn to show up in their dashboards. By then, it's already happened. The brands growing faster than their markets are different. They predict churn weeks in advance, then intercept it before the customer leaves.
Prediction doesn't require a data science team. It requires knowing what to look for. This post walks you through a practical system: three observable signals that identify at-risk customers, how to build a cohort in Klaviyo without extra tools, and three proven flows to bring them back.
Why subscription brands miss the churn signal
A subscription customer's behaviour changes before they cancel. They order less frequently. Their cart values drop. They stop opening emails. Engagement collapses before the cancellation email arrives.
Most brands notice this retroactively. They run a report in week two after the churn happens, see the pattern, and think about fixing it next quarter. Meanwhile, hundreds of dollars in predictable revenue have already walked out the door.
The fix is stupidly simple. Spot the signals early, segment those customers separately, and hit them with targeted retention flows before they feel compelled to cancel.
According to Bain & Company research, repeat customers spend 67% more than first-time buyers on average. Every cancellation you prevent has outsized leverage on your lifetime value.
The three signals that predict churn
You're looking for a pattern, not a single red flag. Think of this as a risk score. The more signals a customer shows, the more likely they are to churn within the next 30 days.
Declining order frequency — A customer who used to order every 30 days suddenly orders every 45 days. Then 60 days pass with no purchase. This is the strongest signal. In Klaviyo, you're checking the time since last order. For subscription brands, define your baseline order window based on your vertical: a beauty box might expect orders every 30 days, a wellness supplement brand every 14 days, a food delivery service every 7 days. When a customer drifts 20-30% beyond their expected window without ordering, flag them.
Declining cart values — A customer's average order value slides from $45 to $28 to $12. They're still technically buying, but they're buying the minimum and pulling back. This is softer than declining frequency, but it signals waning commitment. Track average cart value over the last 3-5 orders and look for a 25% drop over that window.
Engagement collapse — A customer who opened 60% of your emails now opens 10%. They haven't unsubscribed, but they've mentally checked out. In Klaviyo, this is a drop in 30-day open rate. A customer who went from 40% open rate to 5% in the last month is primed to churn.
These signals work across verticals. A wellness brand uses them to predict cancellations on supplement subscriptions. A beauty brand uses them to predict downgrades or pauses on their tier membership. A food box brand uses them to predict skipped orders and eventual cancellations.
The magic is combination. A customer showing one signal might just be busy. A customer showing all three (declining frequency, dropping spend, radio silence) is shopping for an exit.
Building your first churn cohort
You don't need Segment, mParticle, or a data warehouse to identify these customers. Klaviyo's list logic is enough.
Here's the simplest approach: build a static segment in Klaviyo that updates daily, pulling customers who meet your churn criteria. You check it every morning. You push them into a winback flow.
The segment logic (plain English first):
- Time since last order is greater than [your baseline window × 1.3]
- AND average order value (last 3 orders) is 25% below their historical average
- AND email engagement (opens in last 30 days) is below 15%
In Klaviyo's segment builder:
- Go to Segments > Create Segment.
- Add a filter: "Time since last order" > [X days]. For a 30-day cycle, set this to 39+ days.
- Add a filter: "Average order value" < [X amount]. If your typical customer orders $40, set this to $30.
- Add a filter: "Email engagement" or "Email opens (30 days)" < 15% (or your threshold, adjust based on your baseline open rate).
- Combine with AND logic. Save and check the count.
Start conservative. If your segment pulls in 5% of your total subscriber base, that's a healthy size. If it's 25%, your thresholds are too loose. Tighten them. If it's under 1%, you might be missing genuine churn risks. Loosen the window.
Don't include customers in a churn cohort if they've already received your winback flow in the last 60 days. Nothing tanks retention rates faster than hitting the same customer with "come back!" emails three weeks apart. Add an exclusion filter: "Has not received [Winback Flow] in the last 60 days."
Once your segment is stable, push customers into a flow (not a campaign). Flows trigger automatically every day. A customer enters the churn segment, they hit the flow within 24 hours. A customer shows improvement and exits the segment, they don't trigger again.
The three-email winback sequence
The best winback flows use three emails, timed 0/3/7 days apart. Each email has one job.
Email 1 (Day 0): Acknowledge the silence, not the churn
The customer hasn't cancelled yet. Don't assume they're leaving. Instead, acknowledge the pattern without calling it out:
"We noticed you haven't had [product] in a while."
That's it. No panic, no discount code thrown at them. The goal is to remind them the product exists and spark a thought: "Oh right, I do like that."
This email should be short. One paragraph, one ask. Link directly to their most-purchased item or category. No complicated navigation. No secondary CTAs. Make it trivial to re-order.
A wellness brand might say: "Your last order of Energy Greens was [date]. Ready to stock up?" A beauty brand: "Your usual Vitamin C serum is ready to ship." A food brand: "We've got fresh coffee beans this week. Your usual roast is in stock."
The tone is helpful, not accusatory. You're doing them a favour by reminding them.
Email 2 (Day 3): Add a gentle incentive
Three days have passed. If they didn't re-order from email one, it's time to sweeten it. Not a massive discount. A small incentive tied to their usual purchase.
"15% off your next order of [product]" or "Free shipping on orders over $40" or "Reorder your last box and get $5 off."
The incentive should expire in 7 days. This creates mild urgency without being pushy. The customer knows the offer has an edge, so they don't sit on it forever.
The copy moves slightly: "We miss having you as a regular." Humanise it. You're not a brand sending retention emails. You're someone at a company who genuinely wants them back.
Email 3 (Day 7): Last touch, different angle
It's been a week. This is the last email in the sequence. If they haven't re-ordered by now, it's time to ask a different question: why?
Maybe they found a substitute. Maybe they paused subscriptions temporarily. Maybe they can't afford it right now. This email asks why, not demands they come back.
"What changed? Let us know what we can do better."
A link to a one-question survey: "What's stopping you from reordering?" Options: "Found something cheaper," "Went out of budget," "Using a substitute," "Just not interested right now," "Shipping costs too much," "Something else, tell us."
The survey drives two things: data and segmentation. You now know why this cohort is at risk. Customers who say "too expensive" go into one flow, customers who say "found a substitute" go into a competitive flow, customers who say "just not interested" go into a different one.
Even if they don't click the survey, the email itself shows you care. It's the difference between feeling abandoned and feeling like someone's trying to understand you.
SMS amplifies all three emails. Text the same customer on day zero with a short message: "Hey! Your [product] is back in stock." SMS open rates are 98% vs email's 20-30%, so you're hitting them twice with the message via the most reliable channel. Add the SMS sequences right alongside the email flow in Klaviyo. Text day 0, day 3, day 7, same cadence, different channel. Learn more about SMS marketing strategy.
Automation over campaigns
The winback sequence works because it's automatic, not because of the copy.
A campaign is a one-off send. You manually segment, decide to send, hit go, and that's it. The next day you do it again. This is workable, but it's management overhead, and you'll miss people.
A flow is trigger-based and repeats forever. Every single day, new customers enter your churn segment. Every single day, they fire through the winback flow. You build it once, ship it, and it works indefinitely.
More importantly, flows let you branch. After email three, you can branch based on their response:
- Reopened an email? Send them to a replenishment reminder flow instead.
- Clicked the product link? Send them to a upsell flow.
- Completed the survey and said "too expensive"? Move them to a discount pool.
- Completed the survey and said "found a substitute"? Move them to a product education flow about the differences.
This branching is impossible in a campaign send. It's the leverage of automation.
Measuring success
Track three numbers:
-
Conversion rate (reorders divided by flow entrants). A good winback flow converts 8-15% of at-risk customers back to a purchase. If you're converting 5%, your copy or incentive is weak. If you're converting 25%, your segment is pulling in customers who were never at risk.
-
Segment size (how many customers enter daily). Should be 2-5% of your subscriber base. If it's creeping above 10%, your churn is accelerating. Focus on root causes, not just winback. If it's near zero, your thresholds are too tight.
-
Revenue per entrant (total revenue from the flow divided by entrants). A winback flow should generate $15-40 in revenue per customer who enters, depending on your AOV. If you're at $5, your incentive might be too generous or your segment is the wrong people.
A wellness brand running a 10% conversion rate on 400 daily entrants generates $60,000 in monthly revenue from customers who were already lost. That's a 20x return on the engineering time to build the flow.
The first wave
Start with email. SMS amplifies it, but email alone is enough to test the concept. Build your segment, launch your three-email flow, and measure for 60 days. Once you have data, layer SMS. Once SMS is working, you can get fancy with branches and survey-driven segmentation.
Most brands waste cycles optimising email copy when their segment criteria are wrong. Fix the segment first. Get the right people in the flow. Copy optimisation compounds on that foundation.
Ready to ship a churn prevention system? Book a call with our retention team to see how we work with subscription brands.
Keep reading.
All posts
Win-Back Triage: When to Fight for a Lapsed Customer
Arpit Mehar · Oct 28, 2026
What to Do With First-Time BFCM Buyers Before They Go Cold
Shrestha Ghosal · Jun 16, 2026
5 Signs Your Post-Purchase Flow Is Leaving Repeat Revenue on the Table
Shrestha Ghosal · Jun 10, 2026