3.5 Days to 18 Hours: What Order Automation Actually Delivered

The situation

A B2B wholesale distributor taking orders by email, phone, web, EDI and WhatsApp. Four roles' worth of manual order processing. Stock checked against a spreadsheet that was accurate yesterday. Around forty “where is my order” calls a day.

Order-to-delivery averaged 3.5 days against a 24-hour industry benchmark, and £2,000–£3,000 a week was going out in returns and credit notes traceable to manual handling.

Where the time actually went

Following one order with a stopwatch found roughly eight working hours of waiting across three handoffs, none of which involved touching a product. The warehouse did its part in nine minutes.

An order arrived at 4pm, was read the next morning, checked against a spreadsheet, queried by phone about a substitution, typed into the order system at midday, and made the following day's dispatch.

What we built

1. Intake from email, web, EDI and WhatsApp into one normalised order record
2. Extraction for free-text and PDF orders, with customer shorthand learned per account
3. Real-time stock checking, so substitutions are raised at ordering rather than the next morning
4. Smart dispatch routing by postcode and product class
5. Invoicing from the same record, so it cannot disagree with the delivery note
6. Automatic customer tracking notifications

The results

What made it work

Two things. First, the error line was bigger than the labour line — staff time was worth about £30,000 a year, returns and credits caused by manual entry considerably more. The business case rested on the second number.

Second, parallel running for three weeks before cutover. It found the customers with two delivery addresses, the product codes that changed in 2019, and the supplier who invoices in a different unit. All of those would otherwise have been found by a customer.

Frequently asked questions