Five tiers, one trigger

When a large California retailer moved 9 million profiles to Klaviyo during peak Q4, the migration was the visible job. The quieter one was deciding how the platform would see each customer once they arrived. A customer who arrives without order history looks like a first-time buyer and gets the first-order offer, and at 9 million profiles that mistake would have been sent to everyone at once.

So the tiers came first, before the flows. I built them on two numbers every store can produce, order count and lifetime value, in the same way I had built a five-tier model at an earlier brand before Klaviyo had RFM at all. The data behind the thresholds came from the cohort study in the first assignment: 24 monthly cohorts, about 1.5 million first-time buyers, followed for two years.

Why the default split misfiles the base

The finding that shaped everything: about 64% of first-time buyers never placed a second order (share of first-time buyers, two-year cohorts). The first order was 70% of a customer's two-year value, and the steepest fall in repeat buying came between month zero and month two.

Klaviyo's default RFM scoring splits customers into roughly equal thirds on each of recency, frequency and value. With 64% of the base on one order, the middle frequency third is made largely of one-time buyers, and a flow that treats "medium frequency" as a repeat customer sends the wrong message to most of them. The scoring has to follow the shape of the base, not divide it evenly. In the settings I recommended, the lowest frequency score covers the bottom 70% of customers, which is the one-order group, and the top score goes to the top 10%, people with four or more orders.

The five tiers

• No orders. Subscribers who have never bought. The welcome and browse series are theirs, and the first-order offer belongs here and nowhere else.
• One order. The 64%. The post-purchase series runs 14 to 21 days from the order, and the next 60 days are the window where a second order is most likely.
• Two to four orders. Repeat buyers. They have proven the product; the job is cadence and range.
• Five or more, under $840 lifetime value. Loyal, lower basket.
• Five or more, over $840. The top tier, where the $840 line separates frequent small orders from the customers who carry the margin.

Five tiers is enough to make every flow behave differently and few enough that a person can hold them in their head, which is the test. A model nobody can explain at a whiteboard does not survive the first campaign brief.

One trigger, five messages

Every flow splits by tier at the top, so the same trigger sends a different message depending on where the person is. The abandoned-cart flow is the clearest example.

• No orders: the 15% first-order offer, at 72 hours. The person is deciding whether to trust the brand at all, and the offer is the answer.
• One order: 10% off the cart, at 24 hours. They have bought once and the next few weeks are when a second order is most likely, so the offer comes sooner and is smaller.
• Two or more orders: free shipping or a perk, never a discount. A discount teaches the people who least need one to wait for it.

The same split runs through post-purchase, winback and loyalty. A winback to a one-time buyer asks whether the first order was right for them. A winback to a five-order customer asks what changed. The trigger is the same event; the tier decides the sentence.

Once the retailer was live, the core flows, welcome, browse and cart abandonment, post-purchase, winback and loyalty, were rebuilt inside 30 days, and the tier split went into each of them as it was rebuilt. I have not measured the lift of the tier split against a flat flow at this brand, so this piece is the model and the reasons, not a result.

Do this Monday

1. Count your customers by order count: none, one, two to four, five or more. Use the store's orders, not the email platform's profile counts.
2. Divide. If the one-order group is more than half of everyone who has bought, the default equal-thirds RFM will misfile most of it.
3. Set the frequency scores to the shape: the lowest score covers the one-order group, the top score covers your top tenth by order count.
4. Pick the value line between your two top tiers from lifetime value. At the retailer it was $840; yours is where the margin starts to concentrate.
5. Open your cart flow and write down what each tier gets. If the answer is the same offer for everyone, the tiers exist in the data and not in the program.

In one line: build the tiers on order count and lifetime value before the flows, score RFM to the shape of your base, and let the same trigger send five different messages.