Network TCO Simulator S1

How many DCs · transport vs facility fixed cost vs inventory · square-root law applied

The question this answers is “how many DCs”. Look only at freight and more sites always wins. Facility fixed cost and inventory (which scales with √n) move the other way, and they routinely reverse the conclusion. Never set hub count from a routing study alone. Inventory parameters must come from finance, not from logistics.

Outbound volume (DC → customer)

Annual shipped weight540 M lb

Outbound percent full70%

Measured at 70% in the v1 optimizer. Raise it to see what better consolidation is worth.

Demand distribution

Inbound replenishment (source → DC)

Inbound percent full95%

⚠️ Inbound cost rises with more sites because replenishment fragments. Leave it out and adding DCs looks better than it is. This is the most commonly omitted term in real network studies.

Search range & candidates

Search up to N DCs8

Teitz-Bart swap improvement (escapes greedy local optima)

Evaluate a specific hub set

Upload the hub export from v3 and the TCO gap against the optimal set is computed.

Freight rates (shared v1/v3 engine)

Base Rate + FSC$2.90 /mi

Minimum Charge$400

Regional circuity (East 1.15 / Central 1.22 / West 1.30)

Import assumption registry (from S2)

Upload the S2 registry and the combined rate and annual volume are set from actuals.

Facility cost

Fixed cost per DC per year$3.0 M

Lease, fixed labour, equipment depreciation and overhead. Volume-driven cost is separated into handling below.

Handling$1.50 /cwt

Purely volume driven, so it shifts the whole curve without changing where the optimum sits.

Inventory cost ⚠️ needs finance sign-off

Annual COGS$360 M

Inventory turns per year12.0

Safety stock share40%

Carrying rate20%

Cost of capital, obsolescence, insurance and shrinkage. Logistics cannot set this number alone — it has to come from finance.

Safety stock scales with √n (square root law) Cycle stock scales with √n (independent EOQ replenishment)

SS(n) = SS(1) × √n, assuming low demand correlation and a common service policy across sites. Wide variation between product families can move this ±20%.
Cycle stock grows the same way when each site replenishes on its own EOQ. Uncheck it if replenishment is centralised.

Service benefit (optional)

Model the revenue effect of better next-day coverage

Base annual revenue$500 M

Revenue lift per point of next-day coverage0.05%

Gross margin30%

⚠️ β is an unverifiable assumption. It is off by default. Switch it on and a single guess can decide the answer — always check the sensitivity tab before trusting it.

Optimal number of DCs

Optimal sites

p-median search

1-day / 2-day coverage

population weighted

vs current 3-DC

TCO per cwt shipped

unit cost to serve

TCO curve by number of DCs

Cost composition

TCO detail by DC count ($M)

n	Out	In	Fixed	Handling	Inventory	Benefit	TCO	1-day

Incremental analysis (n → n+1)

Step	Freight saved	Fixed added	Inventory added	Net	Breakeven fixed	Required β*

Breakeven fixed is the ceiling on annual cost per DC at which that step still pays. If a real lease quote comes in below it, the site is justified. Current setting:
Required β* is how much revenue must rise per point of next-day coverage to justify a step that cost alone does not. It converts a network debate into a specific question for the commercial team.

Tornado — TCO change at the optimum

How far TCO moves when each parameter is shifted ±25%. The longest bars are the values to verify with real data first.

Stability of the optimum — where the answer changes

Parameter	−25%	Base	+25%	Verdict

Parameters that never move the optimum do not need precise estimates. Concentrate verification effort on the ones that do.

Two-way: DC fixed cost × carrying rate

Each cell is the optimal DC count for that combination. A wide plateau of the same number means the conclusion is robust.

Legend