Use cases
six jobs, one decision engine

Six jobs Purands runs

Each page below covers one retention job: what sets it off, what it reads from the customer profile, what gets sent, and how the moment and channel are chosen.

• Abandoned cart recovery: a filled cart with no order, followed up once, on the channel that shopper reads.
• Replenishment reminders: a predicted run-out date, modelled per customer and per product.
• Win-back campaigns: a lapse judged against that customer's own rhythm, not a ninety day rule.
• Back-in-stock alerts: a sold-out item returning, told to the waiting customers most likely to buy now.
• Review requests: a delivered order, asked about once the product has had time to be used.
• Order updates: confirmation through delivery, two-way, on whichever channel that customer reads.

What every one of them decides

A retention tool usually asks one question: has this trigger fired? Purands asks four, per customer, every time.

• Whether to send at all. Silence is a valid decision, and often the right one.
• Which channel. Chosen from where that customer actually replies, not from a channel plan.
• When. Inside that customer's own peak engagement window, not at a fixed offset.
• What, and at what price. Including whether a discount is needed, or would just cost margin.

Those four answers come from one profile per customer, updated as orders, browsing and conversations arrive.

One engine, not six tools

These are not separate campaigns competing for the same customer. One engine decides the next action for each person, then prevents fatigue across every channel.

That means a replenishment reminder can lose to a delivery update, and a win-back can be held back because the customer has already heard enough this week.

The full list of skills is longer than these six. See Agentic Skills for what else runs, and Profiling for the data each decision reads.

Why they are written as use cases

Most brands do not arrive looking for an engine. They arrive with a job: carts are leaking, or repeat orders have stalled.

So each page starts from the job and works back. It names the trigger, the data read, an example message per channel, and what setting it up involves.

The numbers on each page come from brands running that skill, and are reported as ranges rather than guarantees. Your catalogue and list will differ.

Where they run

Every use case can run on WhatsApp, email, SMS, mobile push and the web assistant.

Which one a given customer gets is decided per message, from where that customer actually replies.