A national public-health medicines agency · Southern Africa
Redesigning a national medicines network
$14M
saved annually, a 40% reduction
2,010
health facilities served
39 → 47
depots in the recommended network
61%
reduction in annual kilometres
65%
reduction in overnight stops
The problem
A national public-health medicines agency needed to know whether its distribution network was the right shape. It was moving medicines to 2,010 health facilities across three regions from a network of 39 depots, and the question was not how to drive those routes better, but whether the network itself was costing more than it needed to.
What we did
We built a routable model of the country’s road network and combined it with an activity-based cost model that separated depot storage, transport and overnight costs so they could be traded off against one another rather than bundled together. We then evaluated fifteen distinct network configurations — five network snapshots against three buffer-day policies — across four vehicle classes, so that the client could see the shape of the trade-off rather than a single recommended answer.
What we found
The most valuable finding was structural. Depot storage accounts for 78% of total network spend, against 20% for transport and 2% for overnight stops. Buffer-day policy and depot siting therefore drive cost far more than route optimisation or fleet composition — the opposite of where optimisation effort is conventionally directed.
Acting on that, expanding the depot network from 39 to 47 depots reduces annual network cost by USD 14 million, a 40% reduction, while cutting annual kilometres by 61% and overnight stops by 65%.
We also quantified what resilience costs. Routing that remains viable for the 441 to 480 facilities exposed to seasonal flooding carries a premium of USD 8.2 million, or 13.4% — a figure the client can now weigh deliberately rather than discover during a flood. Running the fleet in-house rather than outsourcing it differed by only 2.5%, which reframed that decision as one about control rather than cost.
What we delivered
A full network study with all supporting analysis, delivered in both of the client’s working languages, together with the scenario model that produced it.