Until 2022 we had a training copy of our order system. It cost about forty thousand pounds a year to keep, it was rarely refreshed, and when we moved to new servers we let it go. From then on our trainers taught new customer service staff in the live system, on an account called ZZ Training Customer, and cancelled the orders at the end of each session.
A session has eight to twelve trainees. Each places five or six practice orders of twenty to forty lines, following a script, and the script uses the same products every time: about thirty of our commonest lines, such as bottled water, copier paper, cleaning chemicals and disposable gloves. We run about three sessions a month.
Two years later our head buyer asked why we kept overstocking exactly those products. We were carrying about eighty five thousand pounds more of them than we were selling, and the overstock was worst at the depot next to the training room.
Our demand forecasting reads order lines at the moment they are entered. A cancellation in our order system is recorded on the order header, and the forecasting extract had never looked at headers, because cancelled orders used to be rare. So every training order counted as genuine demand, for the products in the script, at the trainees' depot. Replenishment did the rest.
It was not only the forecast. In the sales analysis, ZZ Training Customer was our ninth most active account by number of orders. It had been sent our marketing emails, at a mailbox nobody owned. Twice a trainee's order had not been cancelled, and it was picked, loaded and driven to the delivery address on the account, which was our own head office.
We have restored a training environment, refreshed from live each month with customer details masked, and it costs less than the stock we wrote down. Training accounts that must exist in live now carry a test flag, and everything that reads orders, including forecasting, sales reporting and marketing, excludes it. The forecasting extract now works from demand after cancellations.
A record created for practice is real to every system that reads it, and most of those systems were never told it was pretend.
– Serguey Shinder
Top comments (0)