How E-Commerce Marketers Use Klaviyo AI Predictive Analytics to Segment Audiences by Churn Risk and Expected Lifetime Value

Klaviyo AI's predictive analytics transforms transactional e-commerce data into forward-looking behavioral models, allowing marketers to move beyond static, backward-looking email lists. Instead of segmenting customers by arbitrary calendar dates—such as "purchased in the last 60 days"—merchants can segment by dynamic machine learning metrics, specifically churn risk and predicted customer lifetime value (Predicted CLV). By pairing the probability that a customer will defect with their anticipated future spend, brands can protect profit margins, deploy high-touch win-back flows for VIPs, automate replenishment triggers, and suppress unengaged contacts to safeguard domain deliverability. This guide outlines how Klaviyo calculates these predictive metrics, how to construct precision segments in the Klaviyo segment builder, and how to execute targeted retention strategies across key customer lifecycle stages.

Understanding Klaviyo's Predictive Analytics Engine

Klaviyo’s predictive engine analyzes historical order history, buying frequency, and store-wide cohort behavior to forecast individual customer actions. Rather than relying on simple rule-based heuristics, the underlying machine learning models adapt continuously to each customer’s unique purchase rhythm.

Core Predictive Metrics Explained

When predictive analytics is enabled on an account, Klaviyo generates several proprietary metrics on individual customer profiles under the Metrics and insights tab:

• Historic Customer Lifetime Value (Historic CLV): The actual cumulative monetary amount an individual customer has spent across all completed orders to date.
• Predicted Customer Lifetime Value (Predicted CLV): An AI-driven forecast estimating how much additional revenue the customer is expected to generate for the business over a rolling future window (typically the next 365 days).
• Total Customer Lifetime Value (Total CLV): The mathematical sum of Historic CLV and Predicted CLV, representing the comprehensive financial value of the relationship.
• Churn Risk Prediction: A calculated probability score ranging from 0.0 (0%) to 1.0 (100%) indicating how likely a customer is to stop purchasing from the brand. Klaviyo categorizes this score into Low, Medium, and High risk tiers based on individual order cadence and storewide patterns.
• Expected Date of Next Order (EDNO): A dynamically calculated calendar date forecasting when a repeat buyer is statistically most likely to place their next order.
• Average Time Between Orders: The average interval (in days) between consecutive orders for that customer, used as a baseline to detect when a customer is drifting away.

Account Eligibility and Data Prerequisites

Klaviyo's predictive models require a baseline volume of clean historical data to achieve statistical confidence. To populate predictive metrics, an account must satisfy the following thresholds:

If an account meets these store-wide criteria but an individual profile lacks predictions, it indicates that the individual does not yet have enough order history or interaction data for the model to generate a reliable forecast.

How to Build Predictive Segments in Klaviyo

Segmenting by predictive analytics allows e-commerce brands to create dynamic audiences that update automatically every time a customer makes a purchase, clicks an email, or elapses past their expected order window.

Step-by-Step Segment Configuration

To create a predictive segment in Klaviyo:

1. Log in to the Klaviyo dashboard and navigate to Audience > Lists & Segments.
2. Click Create List / Segment in the upper-right corner and select Segment.
3. Enter a clear, descriptive name (for example, Predictive VIPs - High CLV & Low Churn).
4. In the Definition dropdown, choose Predictive analytics about someone.
5. Select the specific predictive metric you want to evaluate:
   • To target by future spend: Choose Predicted Customer Lifetime Value, select an operator (such as is greater than or is between), and enter the target monetary amount.
   • To target by lapse likelihood: Choose Churn Risk Prediction, select an operator (such as equals for categorical tiers or is greater than for numerical values), and specify High, Medium, or Low (or a decimal threshold like 0.70).
6. Use AND or OR connector logic to layer additional conditions, such as email subscription status (Properties about someone > $consent contains email), purchase categories, or SMS engagement.
7. Click Create Segment to trigger Klaviyo’s background calculation.

Segment Rule Architecture:
[Predictive analytics about someone] -> [Predicted Customer Lifetime Value] [is greater than] [$250]
AND
[Predictive analytics about someone] -> [Churn Risk Prediction] [equals] [High]
AND
[If someone can or cannot receive marketing] -> [can receive] [email marketing]

The Predictive Segmentation Matrix: 4 Strategic Customer Archetypes

E-commerce marketers achieve the highest return on marketing spend by cross-referencing Predicted CLV against Churn Risk. Plotting these two dimensions produces four distinct customer cohorts, each requiring a tailored communication cadence and promotional strategy.

                  HIGH PREDICTED CLV
                          │
         At-Risk VIPs     │    Active VIPs &
    (High-Touch Win-Back) │   Brand Champions
                          │   (Loyalty & Access)
──────────────────────────┼──────────────────────────
                          │
       Lapsed / Sunset    │   Steady Re-orderers
      (Low-Touch Sunset)  │   (Cross-sell & Cadence)
                          │
                  LOW PREDICTED CLV
           HIGH CHURN ───────────── LOW CHURN

1. Active VIPs (High Predicted CLV + Low Churn Risk)

These customers represent the economic backbone of an e-commerce brand. They purchase frequently, have large basket sizes, and are actively engaged with marketing campaigns.

• Segment Logic:
  • Predicted Customer Lifetime Value is greater than [Top 10% AOV/CLV Threshold]
  • AND Churn Risk Prediction equals Low
• Marketing Objectives: Maximize brand affinity, encourage organic advocacy, and increase retention without discounting.
• Recommended Playbook:
  • Early Access: Provide exclusive 24-to-48-hour early shopping windows for new product collections and holiday catalog drops.
  • VIP Concierge: Route inquiries from this segment to senior customer support representatives.
  • Non-Discount Perks: Offer free priority shipping, surprise gifts with purchase (GWP), or invitations to private loyalty tiers rather than percentage-off coupons that erode margins.

2. At-Risk VIPs (High Predicted CLV + High Churn Risk)

This is the most critical cohort for e-commerce marketers. These customers historically spent significant sums and have high modeled upside, but their purchase interval has extended far beyond normal expectations. If they churn, the business loses substantial future revenue.

• Segment Logic:
  • Predicted Customer Lifetime Value is greater than [Top 20% Threshold]
  • AND Churn Risk Prediction equals High (or numeric churn score > 0.70)
• Marketing Objectives: Intervene proactively before the customer completely disengages, diagnosing friction and incentivizing a return visit.
• Recommended Playbook:
  • High-Incentive Win-Back: Offer a high-value, margin-supported incentive (e.g., "$30 off your next order over $100" or a free full-sized product) rather than a weak 10% discount.
  • Founder or Personal Outreach: Send plain-text, unbranded emails signed by the store founder or head of product asking for honest feedback regarding their recent experience.
  • Paid Social Retargeting: Sync this segment dynamically to Meta Ads or Google Customer Match to present customized re-engagement creative directly in social feeds.

3. Steady Replenishers (Low-to-Medium Predicted CLV + Low Churn Risk)

These buyers exhibit consistent, predictable purchasing behaviors but maintain modest average order values. They have a healthy relationship with the brand and low immediate defection risk.

• Segment Logic:
  • Predicted Customer Lifetime Value is between [Store Median] and [VIP Floor]
  • AND Churn Risk Prediction equals Low
• Marketing Objectives: Expand basket size through cross-selling and migrate repeat buyers into recurring subscriptions.
• Recommended Playbook:
  • Predictive Cross-Selling: Recommend complementary products based on items frequently bought together by customers with similar browsing patterns.
  • Bundle Incentives: Offer tiered promotions (e.g., "Buy 2, Get 1 at 30% off") to increase average units per transaction.
  • Subscription Enrollment: Highlight the convenience and savings of automated replenishment programs for consumable goods.

4. Lapsing Low-Value Buyers (Low Predicted CLV + High Churn Risk)

These profiles typically represent one-off discount hunters who purchased once during a holiday flash sale and have shown no subsequent brand loyalty.

• Segment Logic:
  • Predicted Customer Lifetime Value is less than [Store Median]
  • AND Churn Risk Prediction equals High
  • AND Historic Customer Lifetime Value is less than [1x AOV]
• Marketing Objectives: Minimize sending costs, preserve sender reputation, and attempt automated, low-cost reactivation.
• Recommended Playbook:
  • Automated Last-Chance Re-Engagement: Trigger a single, aggressive "Last chance before we say goodbye" flow.
  • Deliverability Protection: Exclude this segment from all daily and weekly marketing newsletters. Continuing to blast high-churn, low-engagement profiles severely depresses open rates and damages inbox placement with Gmail and Yahoo.
  • List Hygiene / Sunset: If the profile fails to engage after a 60-day cool-down window, automatically suppress or archive the contact to reduce Klaviyo subscription tier costs.

Executing Automated Predictive Workflows and Flows

Beyond standalone campaign segmentation, Klaviyo’s predictive data functions as a branching mechanism within automated flows. Marketers can use conditional splits based on predictive analytics to personalize customer journeys at scale.

Predictive Win-Back Flow with Value-Based Branching

In standard e-commerce setups, a win-back flow triggers when a set number of days passes since the last order. By introducing a predictive conditional split, brands ensure that valuable customers receive appropriate incentives while low-value buyers do not receive costly discounts:

[Trigger: Placed Order]
       │
[Time Delay: 60 Days]
       │
[Conditional Split: Predicted CLV > $200?]
      ├── YES (High-Value Cohort)
      │     ├── Email 1: Personal letter from founder + exclusive $25 credit
      │     └── SMS Reminder (if consented): VIP access perk
      │
      └── NO (Standard Cohort)
            ├── Email 1: Catalog highlight of trending items
            └── Email 2 (3 days later): 10% discount code

Dynamic Replenishment Using Expected Next Order Date

For consumable, beauty, wellness, and beverage brands, replenishment timing is essential. Rather than assuming every customer finishes a product in exactly 30 days, marketers can leverage the Expected Date of Next Order (EDNO):

1. Create a flow triggered by the Placed Order event.
2. Add a Date Delay configured to pause until a specific date property.
3. Set the date delay to trigger 5 to 7 days before the profile's Expected Date of Next Order.
4. Dispatch a restock reminder email showcasing the specific product previously purchased, equipped with a 1-click reorder link.

Best Practices and Strategic Guardrails

To maximize ROI and prevent operational errors when using predictive segmentation, e-commerce teams should adhere to the following operational principles:

• Calibrate Thresholds to Store Metrics: Generic dollar thresholds borrowed from other brands can distort segmentation. Calculate your store’s specific median AOV, top 10% customer spend, and purchase intervals from Klaviyo’s Analytics > Benchmarks before defining VIP and low-value cutoff amounts.
• Use Dynamic Segments Over Static Lists: Never export predictive data into static lists for recurring campaigns. Customer churn scores and predicted spend update continuously; dynamic segments ensure customers automatically enter and exit cohorts as their behaviors change.
• Implement Fallback Logic for New Contacts: Newly acquired subscribers and first-time buyers will not possess predictive scores until sufficient event data is collected. Ensure flows include default fallback branches so new customers receive standard onboarding messages without interruption.
• Align Cross-Channel Touchpoints: When a customer enters the At-Risk VIP segment, synchronize that segment directly to your SMS marketing list and paid advertising audiences (such as Meta Custom Audiences and Google Customer Match) to create a cohesive, multi-touch recovery experience.