Ask any Salla or Zid merchant what their support team does all day and you will hear the same four words before you finish the question: فين طلبي؟ — where is my order. It arrives at 9am, at 2pm, and at 11pm. It arrives from customers whose parcel is already out for delivery. It arrives twice from the same person on the same day.
This is not a Gulf problem in origin — WISMO dominates support queues everywhere. But two things make it heavier here, and both are worth understanding before you throw a chatbot at it.
The Tracking Page Did Not Fix It
Every store has one. Almost nobody uses it. The customer is already inside WhatsApp — they typed to you the way they type to their brother. Asking them to open an email, find an order number, and paste it into a web form is asking them to leave the channel they chose in order to do your work for you.
Expecting the tracking page to absorb WISMO is expecting the customer to behave like your database. They will not. So the message lands in your inbox anyway, and someone has to open the order, read the carrier scan, and type a sentence.
Cash on Delivery Raises the Stakes
In a market where a large share of orders are paid on delivery, the customer has not spent money yet. They are not chasing a parcel they already bought; they are deciding whether they still want it. A slow or vague answer is not just an annoyance — it is a reason to refuse the package at the door and hand you the return shipping cost.
That reframes the whole question. WISMO is not a support-cost problem. It is a delivery-acceptance problem that shows up in the support queue.
Why Generic Chatbots Make It Worse
The usual answer is a bot that replies "Thank you for contacting us! Your request is important to us." in formal Arabic. The customer reads that as being handled, not answered. They repeat the question, louder, and now you have two messages instead of one plus a customer who has decided your store is careless.
A useful reply has three properties: it is in the dialect the customer wrote in, it names the actual state of the actual order, and it arrives in seconds. Two out of three is not a partial win — it is still a human opening a dashboard.
What Answering It Properly Requires
Three things, none of them optional:
Live order data. The reply has to come from the order record — the order number, the carrier scan, the delivery window — not from a script that says "usually 3 to 5 days". A guess that turns out wrong costs more than no answer.
Real dialect. A customer who writes "فين طلبي" and receives a Modern Standard Arabic paragraph knows exactly what happened. Matching how they wrote is the difference between a reply and a form letter.
A hard stop. When the order is genuinely late, or lost, or the customer is angry, the right move is not a better sentence. It is a human, immediately, with the thread and the order already in front of them.
The Part Worth Measuring
Do not measure how many messages a bot deflected. Measure how many customers stopped asking the second time, how many parcels were accepted at the door, and how many conversations reached you with enough context that you did not have to ask "which order?" first.
Those are the numbers that move. Deflection rate is the number vendors quote because it always looks good.