Week one
Till connected to six of the systems already running in the estate: POS, labor, stock, drive-thru data, brand reporting and the cameras. Then it looked for the first thing worth fixing. It was labor: the forecast underneath it, and the rota built on top of it.
The way labor was run before
Labor was run to a weekly percentage target, with each restaurant manager deciding how to spend it across the week. It's how a lot of multi-site operators work. It's a reasonable system, and it hides a specific problem: a weekly percentage says nothing about when.
A restaurant can hit its number exactly and still be overstaffed through a quiet afternoon and short-handed at peak. The target is met. The sales are lost anyway.
What Till did
Two steps, in order.
First, a better forecast. The forecast supplied by the brand was running 18% off actuals. Till built its own and got to 6%, a third of the error. Everything downstream depends on this. Allocate labor against a forecast that's wrong at shift level and you've only distributed the error more precisely.
Then labor by shift, not by week. Instead of a weekly percentage split at each manager's discretion, Till allocated the right people to the right place and the right shift: staffed up to capture peak trade, hours taken out of genuinely quiet ones. Above-restaurant leaders received the proposed labor changes and approved them before they reached restaurants. Nothing was imposed. Till proposes; a person decides.
Then the part that makes the difference. Till tracked whether the approved schedule was actually implemented, and followed the knock-on effects through transactions per hour, revenue and guest feedback.
What happened next
The restaurants implemented the changes. Over the following couple of weeks Till and ops leadership refined the allocation together, and the weekly labor saving settled at $27,000.