Use case

Win-back campaigns
a reason to return, not a blanket discount

Lapsed customers already know your product. Purands finds who is drifting, decides who is worth reaching, and writes to each one about what they actually bought.

The problem

A customer who bought three times stops. Nothing visible has gone wrong, and the churn stays invisible until the quarter closes.

The usual reply is a discount to everyone who has not ordered in ninety days. It pays customers who were coming back anyway.

What triggers it, and what it reads

The trigger is a lapse against that customer's own pattern, not a calendar date. A weekly buyer is lapsed long before a twice-a-year one.

Churn risk, scored daily against their own purchase rhythm
What they bought, and how they talked about it
Lifetime value, so the offer fits the relationship
Channel affinity and past engagement
Conversation history, so the agent opens where you stopped

Brands running Purands report roughly three times the win-back rate on lapsed customers.

What gets sent

Examples only, written for this page. Real messages are generated per customer from their own profile and your catalogue.

WhatsApp: Hi Ellie, it has been a while since your last Cold Brew order. Same one again, or shall I show you what is new?
Email: Subject references the last order by name. The body leads with what changed since, not with a percentage.
SMS: Reserved for customers who read SMS and nothing else, kept short and infrequent.
Push: Deep-links to the product they bought, not to a generic sale page.
Web assistant: Greets a returning lapsed customer with their history already in context.

How the agent picks the moment and the channel

Timing comes from the customer's own rhythm. The agent reaches out when the risk score rises, not on a fixed sixty or ninety day mark.

The message references the last order and the preferences on file, rather than a generic promotion the customer has no reason to open.

Discounts go to customers predicted to need one. A customer likely to return unprompted is not handed margin to do it.

If the customer replies, the conversation continues on that channel. The agent answers, re-orders and applies codes without a handover.

Setting it up

Connect the store so the full order history syncs, including the orders that predate Purands. Churn scoring needs that history to judge a lapse.

Confirm opt-ins per channel. A lapsed customer is exactly the person whose consent should be checked before the first message.

The other retention jobs

Abandoned cart recovery: one decision per shopper, not one schedule.
Replenishment reminders: before they run out, not after.
Back-in-stock alerts: the moment it lands, to the people still waiting.
Review requests: asked at the moment the answer is yes.
Order updates: the messages customers actually open.

All six run on one engine: Agentic Skills, reading Profiling.

FAQ