Back in 2008, Dell was using,
in the US market,
a so-called GeoManufacturing strategy,
consisting of assembling
identical desktop computers
in different manufacturing plants,
which were assigned
different geographic markets.
This created the benefits of manufacturing
closer to the end consumer
and therefore reducing outbound
shipping costs from the factories.
This also spread the disruption risks
across multiple locations
with identical capabilities.
However, this strategy also increased
component inventories by spreading
buffer stocks of identical components
across multiple locations,
instead of pooling them together.
From an operational standpoint,
it also created a need
to manage four different
inbound streams of material,
coming mostly from Asia,
taking into account
the merge center in Reno, Nevada
to where some monitors were shipped directly
before being merged
with computers manufactured in Austin
so that the western market customers
would receive their monitors
and desktops at the same time.
For large parts such as chassis and monitors,
suppliers in Asia were responsible
sufficient levels of inventory
in the four different locations.
To do that, they were provided
updated demand forecasts
on a weekly basis, covering
a rolling horizon of three months,
based on which these suppliers
were making initial destination
and shipment quantity decisions
for every part.
Because ocean shipping,
with a lead time of about one month,
was used for these heavy and bulky items,
the different inventory levels
and actual demand realization
across the four main locations
would often change a lot
while containers were in transit.
Dell initially corrected these imbalances
by using transfers of component inventories
between its different sites
and could do that
either with a contracted milk run,
operating on a pre-set schedule,
or with more expensive special shipments.
To further reduce these trans-shipments
and handling costs,
Dell had also developed the capability
to change the final destination
and ground transportation mode
of any container up until two days
before it reached the Port
of Long Beach near Los Angeles.
It called that rerouting a diversion
with a ground transportation expedite
which relied on fast,
but expensive, truck shipping from LA
to the final destination
instead of the default rail option.
Because the volume of monitors
shipped by Dell into the US
at that time was often more than
300,000 monitors per week,
its supply chain presented
a very challenging flow optimization problem,
namely, based on the latest
Dell employees had to decide
on the best final destination
and ground transportation
for hundreds of containers at a time
along with the transfer shipment origins,
destinations, and quantities
for both the milk runs
and special trucks on a daily basis.