How to Use Klaviyo's Predictive Analytics to Find Your Best Customers

Klaviyo is sitting on a goldmine of data about your customers. And most brands are completely ignoring it.

I'm talking about Klaviyo's predictive analytics — the built-in machine learning models that can tell you who's most likely to buy next, who's about to churn, when someone's next order will happen, and how much each customer will spend over their lifetime.

This isn't theoretical. We use these predictions across 150+ eCommerce brands at GOSH Digital, and they fundamentally change how we build segments, time campaigns, and allocate marketing spend. Brands that use predictive analytics properly see 25-40% higher revenue from their email channel versus brands that don't.

And the wildest part? This feature is included in every Klaviyo plan. You're already paying for it.

What Klaviyo Actually Predicts (And How Accurate It Is)

Klaviyo's predictive engine generates five core predictions for every customer profile:

Predicted Customer Lifetime Value (CLV): How much total revenue this customer will generate over their entire relationship with your brand. This is the big one. It tells you who your whales are — and who your one-and-done discount shoppers are.

Predicted Next Order Date: When Klaviyo thinks this customer will place their next order. Based on their purchase frequency patterns and recency.

Average Time Between Orders: The average number of days between purchases for this specific customer. Not a store-wide average — individualized.

Predicted Gender: Inferred from name and behavior data. About 85% accurate in our experience. Useful for product recommendations but shouldn't be your primary segmentation lever.

Churn Risk: This is the prediction most brands overlook. Klaviyo can flag customers whose behavior suggests they're about to stop buying from you. This is incredibly valuable for win-back campaigns.

How accurate are these predictions? In our experience across 150+ accounts, the CLV prediction is within 15-20% accuracy for customers with 2+ orders. For single-order customers, it's less reliable — it's basically extrapolating from one data point. The next order date prediction is accurate within a 7-10 day window about 60% of the time. Not perfect, but absolutely useful for timing campaigns.

The catch: You need at least 500 customers with 2+ orders for predictive analytics to activate. Smaller stores or newer stores won't have enough data yet.

Building Segments With Predicted CLV

This is where the real magic happens. Instead of treating all customers the same, you segment based on how valuable Klaviyo predicts they'll be.

The VIP Segment

Criteria: Predicted CLV in the top 10-15% of your customer base.

These are your best customers — the ones who'll spend the most over time. And most brands don't treat them any differently than someone who bought once with a 30% off coupon.

What to do with this segment:

• Early access to new products. Let VIPs see and buy new launches 24-48 hours before everyone else.
• Exclusive offers. Not bigger discounts — exclusive products, bundles, or experiences.
• Higher-touch communication. Personal thank-you emails. Birthday gifts. Handwritten note inserts in their orders.
• Reduced discount frequency. These people don't need 20% off to buy. Stop discounting to them. It's leaving margin on the table.

The Mid-Tier Segment

Criteria: Predicted CLV in the 30th-85th percentile.

This is your growth segment. These customers have potential but haven't fully committed. The goal: move them up to VIP.

What to do:

• Product education content. Help them discover more of your catalog.
• Cross-sell flows. Based on what they've bought, recommend complementary products.
• Moderate incentives. Occasional small discounts are fine here. 10% off or free shipping to nudge the next purchase.
• Loyalty program promotion. If you have a points program, this segment is the most likely to engage with it.

The At-Risk Segment

Criteria: Predicted CLV in the bottom 15% AND has purchased at least once.

These are one-and-done customers. Discount-driven buyers. People who grabbed a deal and probably won't come back on their own.

What to do:

• Lower your spend. Don't waste your best offers on people who aren't coming back.
• One good try. Send a compelling reason to come back — a new product, a different category, a limited-time bundle. If they don't bite, suppress them from campaigns.
• Learn from them. Why did they buy once and leave? Survey them. The answer often reveals a product-market fit issue.

Using Predicted Next Order Date for Campaign Timing

This is one of the most underused features in all of Klaviyo. You can time your sends to arrive right when a customer is most likely to buy.

How it works: Create a segment of customers whose predicted next order date is within the next 7 days. Send this segment a targeted campaign.

Think about what this means. Instead of blasting your entire list with a campaign and hoping some of them are ready to buy, you're targeting people who Klaviyo's models say are actually approaching a purchase window. The result: higher open rates, higher click rates, and dramatically higher conversion rates.

What we've seen: Campaigns sent to "predicted next order within 7 days" segments convert at 2-4x the rate of campaigns sent to the full active list.

How to build this in Klaviyo:

1. Go to Lists and Segments
2. Create a new segment
3. Choose "Predictive Analytics about someone" as the condition
4. Select "Predicted next order date" is "in the next 7 days"
5. Add a second condition: "Has placed order at least once"
6. Save the segment

Now use this segment for your regular campaigns. Or better yet — build a flow that triggers based on predicted next order date.

Building a Churn Prevention Flow

Churn risk is where predictive analytics moves from "nice to have" to "revenue saver." Here's the exact flow we build for clients:

Trigger: Profile enters the segment "High churn risk" (Klaviyo automatically scores this).

Email 1 (Day 0): The Check-In

Subject: "Everything okay, [first name]?"

This email is personal. Not salesy. Just a genuine check-in: "We noticed it's been a while since your last order. Everything okay? We want to make sure you're still enjoying [product category]."

Include a feedback survey link. Sometimes customers churn because of a bad experience, and they'll tell you if you ask.

Email 2 (Day 3): The Value Reminder

Subject: "Here's what you've been missing"

Show them what's new since their last purchase. New products, bestsellers, content — whatever gives them a reason to come back. No discount yet.

Email 3 (Day 7): The Incentive

Subject: "We want you back — here's 15% off"

Now you bring the offer. But note the structure — you tried non-discount approaches first. This prevents you from training your customer base to churn intentionally to trigger a discount.

Email 4 (Day 14): Last Chance

Subject: "Last chance: your 15% off expires tomorrow"

Urgency plus the offer. If they don't convert here, move them to a lower-engagement segment and reduce email frequency. Don't keep hammering a dead horse.

Results across our client base: This churn prevention flow recovers 8-12% of at-risk customers. On a base of 5,000 at-risk customers with an AOV of $65, that's $26,000-$39,000 in recovered revenue.

Advanced Tactic: Predicted CLV for Ad Spend Allocation

Here's something most email agencies won't tell you because it crosses into paid media territory. We do both at GOSH Digital, so I'll share the playbook.

Export your top-CLV customer segment from Klaviyo. Upload it to Meta as a Custom Audience. Build a Lookalike Audience from that segment.

Now you're telling Meta: "Find me more people who look like my most valuable customers." Not just any customers — your best ones.

Compare this to what most brands do: they upload their full customer list and build lookalikes off everyone, including one-time discount buyers who'll never come back. Your ad spend is trying to find more of the people you don't want.

The impact: Lookalike audiences built from high-CLV segments typically produce 30-50% higher ROAS than lookalikes from general customer lists. We've seen this consistently across multiple verticals.

Combining Predictions for Power Segments

The real sophistication comes from combining multiple predictions:

"About to Buy" VIPs: Predicted next order date within 7 days AND predicted CLV in top 15%. These people are your absolute highest-priority targets. Any campaign you send to this group should be your A-game content.

"Slipping VIPs": Predicted CLV in top 15% AND high churn risk. This is urgent. A high-value customer who's about to leave deserves a personal touch — maybe even a direct outreach from your team.

"Rising Stars": First order placed in last 60 days AND predicted CLV in top 30%. These are new customers who Klaviyo thinks will become valuable. Invest in the relationship early. Welcome flow with extra care, early access offers, VIP fast-tracking.

"Bargain Seekers": Predicted CLV in bottom 20% AND last purchase used a discount code. These people are only buying on deals. Reduce their promotion frequency and see if they'll buy at full price. If not, stop wasting margin on them.

Setting Up Predictive Analytics (Step by Step)

If you haven't checked whether predictive analytics is active in your account:

1. Go to Klaviyo Analytics section
2. Look for "Predictive Analytics" in the left nav
3. If it's there, it's already running. If not, you may not have enough order data yet.

For predictive analytics to work well:

• You need 500+ customers with 2 or more orders
• Your Shopify (or eCommerce platform) integration needs to be syncing order data correctly
• Give it 2-3 weeks after activation to calibrate

Once it's running, the predictions update automatically as new data flows in. You don't need to maintain them manually.

Common Mistakes We See

Mistake 1: Ignoring predictions for new customers. Predictions are less accurate for single-order customers, but they're not useless. Use them directionally, not absolutely.

Mistake 2: Not refreshing segments. Predictive segments should be dynamic (real-time recalculating), not static lists. In Klaviyo, segments auto-update, but make sure you're using segment conditions, not saved lists.

Mistake 3: Over-segmenting. Five segments based on predicted CLV is usually enough. Don't create 20 micro-segments that you can't meaningfully differentiate in content or strategy.

Mistake 4: Using predictions without action. Knowing who your VIPs are is useless if you still send them the same emails as everyone else. The insight has to change the behavior.

The Bottom Line

Klaviyo's predictive analytics isn't a flashy gimmick. It's a fundamentally better way to understand and communicate with your customers. The brands in our portfolio that use it seriously — segmenting based on CLV, timing campaigns to predicted purchase windows, building churn prevention flows — consistently outperform brands that don't.

You're already paying for these features. You already have the data flowing in. The only thing missing is the strategy to use them.

---

Mark Cijo is the founder of GOSH Digital, a Klaviyo Gold Partner agency that's driven over $70M in revenue for 150+ eCommerce brands. If you want help building predictive-powered email strategies, book a free strategy call.