Case studies / Trading
Case studyAutomated order intake for a trading company
A trading company receiving hundreds of emailed orders a week automated extraction and validation into its ERP, with an exception queue that keeps staff in control.
Overview
A food ingredients trading company importing and distributing dried fruit, nuts, seeds and spices to bakeries, food manufacturers and wholesalers. A small customer service team handles orders for a large and loyal customer base.
Scale. About 600 orders every week, arriving by email in every format imaginable, from structured purchase orders to a line in the body of a message.
The challenge
Every morning, the customer service team opened an inbox full of orders. Some arrived as PDF purchase orders from customers' systems, some as spreadsheets, some as scanned forms and many as a few lines typed into an email by a busy bakery owner.
Orders were re-keyed from emails into the ERP by hand. Each one meant finding the customer, matching product descriptions to item codes, checking prices and quantities and entering delivery details. On a busy Monday, the backlog stretched into the afternoon.
Re-keying caused delays. Customers who ordered early in the morning sometimes waited hours for confirmation, and urgent orders could sit behind routine ones in the inbox.
Re-keying also caused errors. Wrong items, mistyped quantities and missed delivery instructions reached the warehouse, leading to incorrect deliveries, credit notes and frustrated customers.
The team was reluctant to hand order entry to a machine entirely. They knew their customers, and they had seen what happened when a supplier's automated system confidently processed an order it had misunderstood.
Peaks made the problem worse. Before holiday seasons and at the start of each week, order volumes rose sharply, and the team relied on overtime and temporary help who did not know the product codes or the customers.
The solution
Orders arriving in the orders inbox are picked up automatically. Structured attachments are read directly, and free-text emails and scanned forms are processed by extraction tuned to the formats each customer uses.
Validation rules check every extracted order. The customer must be recognised, each product must map to an item code, quantities and prices must fall within expected ranges and requested delivery dates must be feasible.
Orders that pass validation are created in the ERP and confirmed to the customer automatically. Orders that fail any rule, or where the extraction is uncertain, go to an exception queue with the original email alongside the extracted data.
In the exception queue, customer service staff review, correct and approve orders in a single screen. Their corrections, such as a new product description a customer has started using, can be saved as mappings so the same issue does not recur.
Stock is checked at the point of validation, so customers are told straight away if an item is short. Reports show volumes, automation rates, exceptions by reason and processing times.
The engagement
- Four months
- Three specialists
- Analysis of order formats, design and build of order extraction, validation rules and an exception queue, integration with email, ERP, customer systems and stock, a pilot with key customers and extension to all customers.
Integrations
- ERP
- Customer systems
- Stock
- Reporting
How we delivered it
- 01
Analyse order formats
Three weeks
We analysed several months of orders, grouping them by format and customer and identifying the information that had to be found in each: customer, products, quantities, prices, dates and delivery instructions.
- 02
Build extraction and validation
Seven weeks
We built extraction for each family of formats and validation rules that check every order against customer records, product mappings, prices and stock. Anything uncertain is routed to an exception queue.
- 03
Pilot with key customers
Three weeks
We piloted with a group of key customers covering the main order formats. Customer service reviewed every automated order during the pilot and the rules were tuned on real traffic.
- 04
Extend to all customers
Four weeks
The service was extended to all customers in stages, with new formats and product mappings added as they appeared.
Results
- Less manual order entry
- Order confirmation within minutes
- Fewer errors reaching the warehouse
- Staff in control of every uncertain order
Manual entry has fallen. Most orders are now created automatically, and customer service spends its time on exceptions, customer queries and proactive calls rather than typing.
Order confirmation is faster. Customers who order in the morning receive confirmation within minutes, and urgent orders are no longer stuck behind routine ones.
Errors have fallen. Validation stops orders with unknown items, unusual quantities or wrong prices before they reach the warehouse, and incorrect deliveries and credit notes have dropped.
The exception queue kept control with staff. The team trusts the service because it only processes what it is sure of and hands everything else to people who know the customers.
Peaks are easier to manage. Automated orders absorb most of the extra volume before holiday seasons, and the team no longer relies on temporary help to clear the inbox. New customers can be onboarded by adding their formats and product mappings rather than training more staff.
It handles the easy orders and hands us the tricky ones. That is exactly the split we wanted.
What made the difference
An exception queue kept control with staff.
