Forklifts beeped and a route board blinked like a grumpy Christmas tree. I had drivers lined up, each facing the same choice at every checkpoint, left or right. I needed a mountain of finished route slips fast, not the best one, just ones that followed the same rulebook.
The usual trick was to have office clerks work out which routes should appear most, then check the pile. That’s fine when the board is small. When the choices stack up, the list of possible routes grows so wildly that the clerks can’t keep up, even with extra hands.
So we changed the floor, not the routes. We laid the lanes out like a big grid and fitted little doors between neighbouring lanes. Open a door for a quick handoff, shut it again, and most lanes stay calm while many handoffs happen at once.
Fast can still be wrong, so we needed a simple score. We’d grab a small handful of slips and ask, for these exact routes, do they look like the rulebook’s favourites or like random scribbles. A perfect run would score high, a messy one low, and we landed in between.
When the full board was too big to check, we made easier shifts on purpose. We kept some doors shut to split the grid into smaller chunks the clerks could still track, and tried a half-open version that stayed just about trackable. The scores lined up with the hard cases we could still verify.
With the toughest setup, the floor rattled on and spat out an enormous pile of slips in minutes. A clerk team wouldn’t be slow from laziness, they’d be slow because every extra checkpoint blows up the number of routes to consider, and every open door tangles the lanes together.
At closing time I noticed the mistakes weren’t one big mystery, they were lots of small local slips, like a sticky door here or a late handoff there. When we added up those little error rates, the score behaved the way we expected on the cases we could check. Takeaway, a huge, tightly coordinated floor can run and still be measured, even when full counting is out of reach.