You're staring at three numbers that could tell you who's about to leave, who's worth keeping, and exactly when to nudge someone back. Instead, they sit there like dashboard decoration while you blast the same win-back flow to everyone who hasn't ordered in 60 days—the serial returner and the one-time buyer get identical treatment.
The predictions exist. The data's already being calculated. You're just not using it to change who gets what message, and when.
What Klaviyo's Predictive Analytics Actually Tell You
Klaviyo's predictive analytics give you three numbers per profile: predicted customer lifetime value (CLV), churn risk, and predicted date of next order. All three live in the customer profile under the Predictions tab, and all three update continuously as behavior changes.
Predicted CLV estimates total revenue a customer will generate over their lifetime with your brand. It's forward-looking, not historical. Someone who spent $500 last year might have a predicted CLV of $2,000 if their ordering pattern suggests they'll keep coming back.
Churn risk scores the likelihood a customer will stop buying, expressed as low, medium, or high. Klaviyo looks at purchase frequency, recency, and engagement patterns to flag customers drifting away before they're fully gone.
Predicted date of next order estimates when an active customer is likely to purchase again based on their historical cadence. If someone orders every 45 days, Klaviyo projects forward and marks the calendar.
These aren't guesses. Klaviyo's machine learning models run on your actual transaction and engagement data. But the predictions are only as good as the data you're feeding them.
The Data Klaviyo Needs Before the Numbers Are Trustworthy
Klaviyo's predictive models need volume and consistency before they're actionable. The platform won't surface predictions until you've hit minimum thresholds, and even when the numbers appear, they're not all equally reliable.
You need at least 500 profiles with purchase history and 180 days of transaction data before Klaviyo starts generating predictions. Below that, the models don't have enough signal. In most accounts I work in, predictions start appearing around month six if you're pushing decent volume.
But "appearing" and "trustworthy" aren't the same thing. A predicted CLV calculated on six months of data is directionally useful. One calculated on 18 months of repeat behavior across multiple cohorts is something you can segment on and build flows around.
The predicted date of next order is the first to stabilize because it only requires individual purchase cadence. If someone orders every 30 days like clockwork, Klaviyo spots that pattern fast. Churn risk and CLV need more time because they're comparative—they model each customer against your broader population.
If your predictions feel off, check two things: (1) are your Placed Order events structured consistently, and (2) do you have meaningful repeat purchase rates? If most customers only buy once, Klaviyo has nothing to model. The predictions work best for businesses with natural repurchase cycles—subscriptions, consumables, DTC brands with product depth.
Flow One: Win-Back Timed to Predicted Next Order Date
Most win-back flows trigger on a fixed delay: 30 days since last purchase, 60 days, 90 days. Everyone gets the same clock. That works if your customer base is homogeneous. It doesn't work when one segment reorders weekly and another buys twice a year.
Klaviyo's predicted date of next order lets you time the nudge to the individual, not the average.
Here's how to build it. Create a segment of customers where predicted date of next order is in the past—meaning they're overdue based on their own behavior—and churn risk is medium or high. These are people who usually reorder by now and haven't.
Trigger a flow when someone enters that segment. The first email isn't a discount. It's a product reminder tied to what they bought last time. "Running low on X?" or "Time to restock Y" if you're a consumable brand. "Here's what's new since your last order" if you're apparel or gear.
If they don't open or click within 48 hours, the second email offers help: a product recommendation quiz, a link to best-sellers, or a direct reply option. The third email, three days later, is where you introduce an incentive if you're going to use one.
The entire flow is 5–7 days start to finish, because you're catching them right when they're supposed to be thinking about you anyway. A traditional 60-day win-back gives them two months to forget you exist. This one interrupts at the moment their internal timer goes off.
I see this flow generate meaningfully more reactivation than calendar-based win-backs, because the timing is individualized. You're not guessing. You're following the pattern Klaviyo already identified.
Flow Two: VIP Track for High Predicted CLV
You're treating a customer who's going to spend $5,000 with you the same way you treat someone who'll spend $200. Same emails. Same offers. Same level of attention. That's a retention leak.
High predicted CLV customers should enter a different communication track the moment Klaviyo identifies them. Not just a "VIP" label—a structurally different experience.
Build a segment: predicted CLV is greater than a threshold you set based on your own data. Look at your top decile of customers by actual lifetime spend, take the entry point for that group, and use it as your floor. If your top 10% have spent $1,500 or more, set predicted CLV at $1,500 or higher.
Trigger a flow when someone enters that segment. Email one is a direct acknowledgment: "You're one of our best customers. Here's what that means." Then tell them what they're getting—early access to new releases, a dedicated reply address, free shipping on everything, first look at restocks, whatever you can operationally deliver.
Email two, three days later, delivers on the first promise. If you said early access, give them a product drop 48 hours before everyone else. If you said dedicated support, introduce them to a real person.
Ongoing: these profiles should be in a segment that excludes them from discount-heavy campaigns. You're not training high-value customers to wait for 20% off. They get access and experience instead. They also get different win-back treatment—higher touch, faster response, actual phone calls if you have the team for it.
The frame I use across the programs I run: predicted CLV is forward-looking, so you're not rewarding past spend. You're investing in future spend. Treat a $2,000 predicted CLV customer like they've already given you $2,000, and they're more likely to.
Flow Three: Save Flow for Rising Churn Risk
Churn risk updates dynamically. Someone can go from low to medium to high as their engagement drops and their purchase cadence stretches. Most accounts I audit do nothing with that signal until churn risk hits high and the customer's already half out the door.
You want to intervene earlier, when they move from low to medium. That's the moment where behavior is shifting but the relationship isn't broken yet.
Build a segment: churn risk equals medium, and predicted CLV is above your median. You're targeting customers who are worth keeping and starting to drift. Don't waste this flow on low-value one-time buyers; they're not worth the manual effort.
Trigger a flow when someone enters the segment. Email one is diagnostic, not promotional. "We noticed it's been a while—what's going on?" Give them three reply options: (1) "I'm good, just taking a break," (2) "I'm not finding what I need," (3) "I'm spending elsewhere." Use a tool like Typeform or a simple "reply with 1, 2, or 3" structure.
The responses feed conditional splits. If they say they're taking a break, tag them churn risk / taking a break and suppress promotional emails but keep sending content. If they're not finding what they need, route them to a product specialist or a quiz. If they're spending elsewhere, that's a pricing, product, or positioning problem—tag it and analyze later.
Email two, if they don't respond, goes to content. Send your best editorial piece, a founder note, a behind-the-scenes story, or a customer spotlight—something with no ask. You're reminding them why they liked you in the first place.
Email three offers a low-friction return path: "Here's what we've added since you were last here." Showcase new products, new content, or new features. If you're going to use an incentive, this is the email, but frame it as a welcome-back gift, not a desperate discount.
The entire flow runs over 10–14 days. If they reengage—open, click, or purchase—they exit and churn risk usually drops back to low. If they don't, they stay in the segment and eventually graduate to your high-churn save sequence.
Why Most Accounts Leave This on the Table
Klaviyo hands you behavioral predictions most platforms charge extra for or don't offer at all. The data's already there. The models are already running. You're already paying for it.
But using predictions requires segmentation discipline and a willingness to treat customers differently based on what they're likely to do, not just what they've already done. That's a harder mental model than "everyone who hasn't purchased in 60 days gets the same email."
The accounts I work in that actually use predictive analytics—win-backs timed to next order date, VIP tracks for high CLV, early interventions for rising churn risk—retain more customers and generate more repeat revenue than the ones that don't. Not because the predictions are magic. Because they let you stop guessing and start responding to the signal Klaviyo's already giving you.
If you're running Klaviyo and you've never opened the Predictions tab on a customer profile, you're leaving retention revenue on the table. If you want a second set of eyes on how predictive analytics could fit into your flows, I'll audit your Klaviyo account for free: getner.ai/audit.