Case · Wholesale distribution · Wholesale distributor of cleaning products
Four out of ten orders arrived after 6pm. Now they load on their own.
The order comes in over WhatsApp, in whatever format it arrives. The system checks it against the current price list, confirms it with the customer and loads it into Tango. It flags only what doesn't add up for review.
The context
Rosario · 14 people · 380 active customers across corner stores, self-service shops and kiosks · two WhatsApp numbers, one per salesperson · a price list in a PDF that changed every two weeks · one admin person loaded orders into Tango from 8 to 10:30am.
Integrated withWhatsApp Business · price list in Excel · Tango.
The capabilityWhatsApp service and order taking
01
What we found
The 9:40pm order waited until 9am, and by then there were twelve just like it. The typical error wasn't a typo: it was an old price list, because the customer ordered at the previous cycle's price and nobody noticed until the invoice. The errors showed up at dispatch: 19 a month, counted in credit notes.
02
What we built
The message comes in as text, voice or a photo of a handwritten list. The system checks it against the current price list, builds the order, sends the customer the details back with the total and delivery date, and loads it into Tango. It flags three things for review: an out-of-stock product, a customer with an overdue account, and an unusual quantity (more than three times what that customer usually orders). The admin finds the loaded orders in the morning and only looks at the flagged ones, which came to 12%.
How it works
Comes in
Goes out
WhatsApp, text, voice or photo
The current price list
The system
Interprets, builds, loads
Interprets
Builds
Loads
Against this cycle's price list, not last cycle's.
The order loaded into Tango
Confirmation to the customer with total and date
Out of stock, an overdue account or an odd quantity: it gets flagged, and the admin looks at it.
03
What happened
Order confirmation to the customer went from the next morning (14 hours on average) to a median of 3 minutes. Errors caught at dispatch dropped from 19 to 4 a month, same three months against the year before. The admin stopped spending two and a half hours a day loading orders; now she handles the running account, which nobody used to.
