Latest / The Retention Marketing Podcast with Fexingo: Email, SMS, and Loyalty for Repeat Customers / How One Brand Uses SMS to Predict When Customers Will Churn
Transcript
- Lucas: So last week I was looking at a churn report from a DTC brand I've been following called Pique Life. They sell tea—like, premium loose-leaf and matcha—and they've got a pretty loyal subscriber base. But what caught my eye is how they're using SMS to predict when a customer is about to churn, before it actually happens. Luna: Predict churn proactively? That's different from the usual reactive win-back campaigns we talk about. How do they pull that off? Lucas: Well, they built a simple model based on purchase frequency. They know their average repurchase cycle is about 45 days—that's the mean time between orders for their typical subscriber. And they also know the standard deviation is around 12 days. So after about 57 days—that's mean plus one standard deviation—they flag the customer as 'at risk'. Luna: Right, so they're basically looking at each customer's own cadence, not just the overall average. Lucas: Exactly. Then they send an SMS. Not an email—specifically an SMS—with a personalized discount. The message says something like, 'Hey, it's been a while since your last order. Here's 20 percent off your next subscription box.' And they track the response. The result was a 28 percent reduction in churn within one quarter. Luna: 28 percent is significant. But why SMS instead of email? Isn't email cheaper and easier to scale? Lucas: Great question. Their data showed that for this particular intervention—flagging a potential churner—SMS had a 98 percent open rate within 3 hours, versus email's 22 percent within the same window. And the conversion rate from SMS was over 12 percent, while email was under 2 percent. For a high-stakes moment like 'the customer is about to leave forever,' SMS is worth the cost. Luna: So the timing is really tight. They're hitting that window when the customer hasn't ordered in a while but hasn't fully decided to cancel yet. Lucas: Exactly. And they don't just send one message. If the customer doesn't respond within 7 days, they send a second SMS with a slightly higher discount, like 25 percent. If there's still no response after 14 days, they switch to email because at that point the customer is likely already gone—the SMS channel feels too pushy. Luna: They're balancing persistence with respect. That's smart. Did they have a data science team building this model? Lucas: No, and that's the part I think is really useful for our listeners. The CMO told me they used a basic spreadsheet with a few formulas. They calculated each customer's average time between purchases, flagged anyone beyond their personal mean plus one standard deviation, and then exported that list to their SMS platform. It's not a machine learning model—it's just descriptive statistics. Luna: So a brand with just a few thousand customers could replicate this today. Lucas: Absolutely. And the key metric they focused on wasn't the discount redemption rate, but the re-engagement rate—the percentage of at-risk customers who placed an order within 30 days of the SMS. That's the metric that shows you're actually preventing churn, not just shifting a purchase forward. Luna: That's a good distinction. A lot of brands celebrate a win-back sale, but if that customer would have bought anyway in a week, you didn't really save them. Lucas: Precisely. And Pique Life tracked this for six months. They saw that 73 percent of customers who responded to the SMS and placed an order went on to place at least one more order in the next three months, compared to only 31 percent of customers who didn't get the SMS at all. So the intervention actually changed long-term behavior. Luna: That's a huge lift. What about the customers who got the SMS and didn't buy? Did they churn faster? Lucas: Interesting question. They actually found that the SMS had no negative effect. The churn rate for the SMS group that didn't buy was the same as the control group that received no message. So there's no downside to sending the reminder. And for the ones who did buy, the lifetime value was significantly higher. Luna: So the risk is basically zero. Why don't more brands do this? Lucas: I think two reasons. One, they don't track the repurchase cycle at the individual customer level. They look at aggregate metrics like 'average order frequency' and miss the outliers. Two, they default to email because it's cheaper, but for time-sensitive interventions like predicting churn, SMS is more effective. Luna: And there's also the fear of being seen as pushy, I imagine. Nobody wants to get a text from a brand saying 'Hey, you haven't bought from us in a while.' Lucas: Right. But Pique Life found that the customers who actually appreciated the reminder were the ones who genuinely forgot to reorder. Their subscription customers had auto-renew, but the one-time buyers—which is where most churn happens—often just lost track. The SMS felt helpful, not creepy. Luna: They framed it as service, not sales. That's the key. Lucas: Exactly. And they A/B tested the messaging. The version that said 'We noticed you haven't reordered' underperformed the version that said 'Your favorite matcha might be running low.' The latter felt like a useful tip, not a guilt trip. Luna: That reminds me of something we saw with a skincare brand last year. They sent a 'you're almost out' SMS based on the typical usage rate of their product. But the timing was off because people use skincare at different paces. Lucas: Yeah, that's the weakness of a one-size-fits-all model. Pique Life's approach is better because it's personalized to each customer's actual purchase history. And they update the model every week, so the threshold adjusts as the customer's behavior changes. Luna: So if a customer starts ordering more frequently—say, they become a heavy user—the system adapts and doesn't flag them too early. Lucas: Exactly. It's a living model. And that's something a small brand can manage manually with a weekly spreadsheet export. The CMO told me they spend about 30 minutes a week on it, just reviewing the flagged customers and approving the SMS copy. Luna: Thirty minutes a week to reduce churn by 28 percent. That's an incredible return on time. Lucas: Yeah, it really is. And we've been talking about this approach quite a bit on the show lately—using data to anticipate customer behavior rather than just react to it. Luna: If these marketing conversations have sparked something you've actually used or a tactic you've tried, we'd love for you to consider supporting the show. It keeps us ad-free and independent. You can find us at buy me a coffee dot com slash fexingo. Lucas: Thanks for that, Luna. And honestly, knowing that people apply these ideas makes it worth the effort. So back to Pique Life—one more detail I found fascinating. They also track which products are most likely to be reordered. For their matcha, the repurchase cycle is 38 days, but for their herbal teas it's 52 days. So they tailor the SMS timing by product, not just by customer. Luna: Oh, that's a nice layer of personalization. So the SMS for a matcha buyer comes earlier than for a herbal tea buyer, even if both customers have the same overall ordering frequency. Lucas: Exactly. And that level of granularity is what drives the 28 percent number. They're not just sending a generic 'come back' message—they're sending a relevant, product-specific nudge at the exact moment the customer is most likely to need a refill. Luna: So for a brand listening today, what's the first step? Should they start with the spreadsheet model, or is there a tool that can do this automatically? Lucas: I'd start with the spreadsheet. It forces you to understand your own data and what the metrics mean. Once you've validated the approach for a few months, then you can look into tools like Klaviyo or Postscript that have built-in predictive models. But the key is to define your 'at-risk' threshold carefully. Don't just guess—calculate it from your own data. Luna: And test the messaging. What works for a tea brand might not work for a pet supply brand. Lucas: Right. The tone, the offer, the timing—all of that needs to be tuned. But the core insight—predict churn using purchase frequency, intervene via SMS—is repeatable across pretty much any DTC brand with a recurring revenue model. Luna: It's a great example of using simple data to deliver a high-impact customer experience. And it doesn't require a huge budget or a team of data scientists. Lucas: Exactly. Just a spreadsheet, an SMS platform, and a willingness to think about retention from the customer's perspective. Luna: I think that's something we can all take away.