In 2023 we chose a handheld system for picking orders in our yards. The picker scans the bay and the product, and the handheld refuses anything that does not match the order. Wrong items on deliveries cost us a return trip, a credit note and sometimes a customer, so the board was interested.
We piloted it at our Harrogate branch from January to March. Its manager volunteered, and his yard was the tidiest in the company. Wrong items fell from two point one percent of order lines to under point seven. The business case for the other thirty seven branches assumed a reduction of about two thirds, and the board approved it on that basis.
A year after the rollout, wrong items across the company had fallen from two point four percent to one point seven, less than a third. Our finance director asked, reasonably, where the rest had gone.
Most of it had never been the system's. Before go live, the Harrogate manager had relabelled every bay and reorganised his yard so that similar products were no longer side by side. When we looked at his figures for the previous autumn, wrong items had already started falling while the relabelling was under way. During the pilot one of my analysts was on site three days a week and fixed problems the same day, while the rollout branches had two days of support at go live. And January to March is the quietest quarter in a builders' merchant's year, so the pilot never met a busy Saturday in May.
We had not lied to the board. We had measured a new system together with a tidy yard, a keen manager, an engineer on call and a quiet season, and given all the credit to the system.
The remedy was mostly the relabelling, which we then did at every branch, and wrong items came down to about one point one percent. Pilots now run at two branches that we choose, one typical and one difficult, never a volunteer, and for at least one busy month. A comparison branch without the change is measured over the same weeks, everything else that changes at the pilot branch is written down, and the business case states the level of support it assumes.
A pilot tests a system together with the conditions you gave it. Promise the board only what will survive ordinary conditions.
– Serguey Shinder