Dispatch software that replaced eleven spreadsheets

Atlas planned 340 daily deliveries across four depots in a set of linked spreadsheets. It worked, in the sense that goods arrived, but it depended on two dispatchers who held the exceptions in their heads. When either was on holiday, planning took twice as long and mistakes tripled.

Two previous attempts to buy off-the-shelf transport management software had failed. Both assumed a depot model Atlas does not use, and neither could handle the shared-load arrangements that make their margins work.

We spent three weeks sitting with dispatchers, documenting the rules nobody had written down: which customers tolerate a two-hour window, which drivers can take which vehicle classes, and how shared loads are split for invoicing.

The platform is deliberately unambitious technically: a Go service, PostgreSQL, and a dense React interface built for keyboard use. No route optimisation algorithm, because the dispatchers’ judgement beat every optimiser we tested on their real constraints. Instead the software removes clerical work and makes conflicts visible.

We shipped a thin slice to one depot in week nine and let it run alongside the spreadsheets for a month before extending it.