How to Segment by Predicted CLV in Klaviyo (UK Guide)

Most UK stores treat every customer the same: the same discount, the same cadence, the same offers. That was survivable when acquisition was cheap. It isn't any more. Klaviyo's predictive analytics gives you a way out — it reads your existing order and engagement data and tells you which customers are worth spending on, which are about to lapse, and roughly when each one will buy again.

This guide walks through how to build predicted Customer Lifetime Value (CLV) and churn-risk segments, message them differently, and wire them into automated flows so the returns compound. No new tooling required — if your store already sends through Klaviyo, the data is either there or a few thresholds away.

Klaviyo splits lifetime value into two figures. Historic CLV is the total a customer has actually spent to date — a backward-looking fact. Predicted CLV uses AI to estimate how much a customer will spend, and how many purchases they'll make, over the next 365 days. It's the forward-looking number, and it's the one that matters for segmentation.

The prediction is built from three inputs: purchase frequency, average order value and the time between orders. The underlying models retrain weekly, so forecasts update automatically as new purchases and interactions arrive — there's no manual tuning to do. Every standard predictive insight uses that 365-day window; brands on Klaviyo Marketing Analytics can shorten or lengthen the CLV calculation window to suit their buying cycle.

Historic CLV tells you what a customer was worth. Predicted CLV tells you what they'll be worth — and that's what you budget against.

Predictive fields don't appear on demand. Klaviyo needs enough history to model behaviour reliably, and it enforces three conditions — all of which must be met before the fields show up:

• At least 500 customers who have placed an order. This counts real purchasers, not active profiles. A big newsletter list won't get you there on its own.
• At least 180 days of order history, with orders in the last 30 days. The store needs to be trading, not dormant.
• Some customers with 3 or more orders. Without repeat buyers, the model has nothing to learn frequency from.

For newer or lower-volume UK brands, this is the honest catch: you simply won't see predictions until you cross these thresholds. If you're close, focus on turning first-time buyers into second-time buyers — that both unlocks the data and improves the accuracy of the forecasts once they arrive.

Once your store qualifies, the data sits on each customer profile under Profiles → Metrics & Insights. Klaviyo visualises Historic CLV as a blue bar and Predicted CLV as a green bar, sat alongside a handful of related metrics:

• Average order value. The typical spend per order for that customer.
• Average time between orders. Your natural replenishment window, per person.
• Expected date of next order. The single most useful field for timing winbacks and replenishment.
• Churn-risk score. The probability that the customer won't buy again — more on this below.

You read these on individual profiles, but you act on them at scale by building segments that filter on the predictive fields.

The core play is simple: split your customers by predicted spend and message the extremes differently. A common approach is a high-CLV segment of the top ~20% of predicted spenders and a low-CLV segment of the bottom ~20%.

High predicted-CLV customers

These are the people who justify your best offers. Push them higher-value products, new arrivals, VIP and loyalty invitations, and review requests. The Willow Tree Boutique, for example, targeted customers with a predicted CLV over £400 (or an AOV over £120) with campaigns highlighting their pricier apparel — putting premium product in front of the people most likely to buy it, rather than blanket-discounting to everyone.

Low predicted-CLV customers

Lower-CLV customers respond better to tried-and-true best sellers and the occasional sale item — not your premium range. A well-worn tactic from Klaviyo's own help docs is to build a segment of customers predicted to spend no more than a small threshold and hit them with a discount-led winback.

Critically, layer engagement conditions into that segment to protect deliverability — for example, is in the main newsletter list AND has opened an email in the last 90 days. Sending discounts to low-value, disengaged addresses is how you damage your sender reputation. Only send to people still paying attention.

Churn Risk Prediction is the probability that a customer will not purchase again, expressed between 0 and 1. A score of 0.45 means a 45% churn risk; 0.9 means a 90% chance they won't buy again. Klaviyo commonly buckets it as:

• Low — under roughly 33% probability of not purchasing.
• Medium — 33% to 66%.
• High — over roughly 66%.

Churn risk falls when a customer places an order and rises the longer they stay inactive. One caveat worth flagging: Klaviyo itself notes churn risk can misfire for stores with few repeat buyers, because first-time purchasers often score as high-churn immediately after buying. If that describes your store, lean on Expected Date of Next Order as a more reliable trigger for winback timing.

The most valuable segments combine metrics rather than relying on one. The standout combination is high predicted CLV plus high churn risk: valuable customers on the verge of lapsing. Call them at-risk champions or sleeping VIPs — either way, they're the group that justifies premium retention effort.

Because they're worth far more than the average lapsing customer, you can afford to spend on keeping them:

• Early access to new ranges before the general list.
• Exclusive discounts reserved for top spenders, not blanket sale codes.
• Personal outreach — a genuine message from a founder or account manager, especially for higher-ticket brands.

A blanket 10% off is wasted on a champion and unnecessary for a bargain hunter. The point of combining metrics is to spend where it changes the outcome.

Segments alone are useful. Segments wired into automated flows are where the compounding returns sit. Klaviyo's own data shows automated flows generate far more revenue per recipient than one-off campaigns — commonly cited as roughly 30x. Three flows do most of the heavy lifting:

• Predictive replenishment. Triggered by Expected Date of Next Order, so the reminder lands when the customer is actually running low — not on a fixed time delay that ignores individual buying cycles.
• VIP upsell. Aimed at high predicted-CLV customers, pushing premium product, new arrivals and loyalty perks.
• Predictive winback. Aimed at high churn-risk customers, with the sleeping-VIP subset getting the premium offer and the low-value subset getting a lighter, engagement-gated discount.

For UK Shopify, BigCommerce, Magento and WooCommerce stores, this predictive data syncs automatically once the integration is connected. That makes it a low-effort, high-leverage upgrade over the static time-delay flows most brands still run.

Predictive fields aren't locked inside Klaviyo. You can export them via CSV or API and push them to Facebook and Google as the source for lookalike or similar audiences. Feeding your platform your highest predicted-CLV customers — rather than all buyers indiscriminately — tells the algorithm to find more people like your best ones, which sharpens paid targeting and lifts acquisition efficiency. Retention data, in other words, makes your acquisition spend work harder too.

The economics have shifted. Rising Meta and Google ad costs mean brands can no longer afford to treat every customer the same, and retention has become the cheaper lever. McKinsey data indicates companies that excel at personalisation generate around 40% more revenue from those activities than their peers — and predictive segmentation is personalisation with a clear payback attached.

The work isn't complicated. Clear the data thresholds, build a top-20% and bottom-20% CLV segment, layer in churn risk to find your sleeping VIPs, and wire the whole thing into replenishment, upsell and winback flows. The models retrain weekly on their own, so once it's set up it keeps improving without you touching it. The brands pulling ahead aren't spending more — they're spending it on the right customers.