Retail & e-commerce Usage scenario

Know your customers and send the right WhatsApp message to the right people

A shop in Conakry gathers its customers in a base fed by the till and WhatsApp, sorts them by recency and frequency, and sends targeted messages with their consent.

Describe your project See the solution

Project overview

This scenario illustrates what we build; it is not a client reference.

This scenario shows how a shop in Conakry can gather its customers in a single base, fed by the till and by WhatsApp, then send targeted messages instead of one broadcast to everyone. Each customer has her purchase history, her open credit and a group: regular, occasional, not seen for three months. A dashboard shows the best customers and those who are drifting away.

Client challenges & requirements

A ready-to-wear and accessories shop runs two stores in Conakry. The manager knows her customers by first name and phone number. About 1,200 customers have gone through the till, 300 of them regulars, but nobody knows exactly which ones.

The information is scattered: the till receipt, hundreds of WhatsApp conversations, and a credit book where instalment payments are written down. When a new collection arrives, the manager sends the same WhatsApp broadcast to her whole list.

Customers who only buy shoes get the dresses, those who bought the day before get the same photo again, and some end up blocking the number. Customers who have not come back for months go unnoticed, and overdue credit is only chased when someone rereads the book.

Main objectives

Keep one record per customer, with her number, her purchases and her credit, filled in automatically by the till

Sort customers into simple groups by last visit, purchase frequency and basket size

Send targeted WhatsApp or SMS messages, with their consent, to a specific group rather than the whole list

See each week the best customers, those drifting away and the credit to chase

Project solution

We start with the customer base

A Laravel back office with a PostgreSQL database receives each sale from the till with the customer's phone number. If the number exists, the purchase is added to her record; otherwise, a record is created and the sales assistant fills in the first name. Orders placed on WhatsApp are entered into the same record. The credit book also goes into the back office: amount due, instalments, payments received.
For contact management, we use Twenty CRM or a custom module, depending on what the team prefers to work with. Then the groups. The rule is written in plain words: if a customer has bought at least three times in the last six months, she is a regular; if she has bought once or twice, she is occasional; if she has bought nothing for three months, she is due a reminder. Average basket and categories bought complete the sorting.
This is the recency, frequency, monetary method, without the jargon. Messages go out through the WhatsApp Business API or by SMS, only to customers who have given their consent.
Example
A new arrival of shoes goes to the 180 customers who have already bought shoes, not to all 1,200.
A reminder goes to customers not seen for three months, with a different message for former regulars. A credit reminder goes out a few days before the due date. A Metabase dashboard shows the best customers of the month, the number of customers in each group and overdue credit. Product recommendations will come later, only if the purchase history is rich enough to justify it.

Indicators to track

Share of sales linked to a customer record
Number of customers per group each month
Customers due a reminder who came back to the store after a message
Replies and purchases after each targeted campaign
Credit paid on time
Customers who ask to stop receiving messages

Tech stack

Back-office et base clientes: Laravel Livewire PostgreSQL
Gestion des contacts: Twenty CRM Module sur mesure
Messaging: WhatsApp Business API Passerelle SMS
Tableau de bord: Metabase
Exploitation: Sauvegardes quotidiennes Contrat de maintenance

Frequently asked questions

No. The base fills up as sales come in: each till transaction with a phone number creates or completes a record. The most regular customers show up on their own as sales accumulate. If the manager already keeps a list in a file or on her phone, we import it at the start so nothing begins from scratch.

The sales assistant asks for the customer's consent at the till or in the WhatsApp conversation, and the answer is noted on the record. Those who have not given consent receive nothing. Each message includes a simple way to say stop, and the request is applied immediately. Targeted, less frequent messages are better received than a broadcast to everyone.

Because with a few hundred customers and a catalogue that changes every season, the history is too thin for a model to suggest better than the sales assistant does. Groups by recency, frequency and basket are enough to target messages. If after several seasons the history is rich enough, we add simple recommendations, for example the category bought most often.