Single-factor models have one flaw: they don't work

A retailer running about 1,000 orders a day opens a second warehouse to shorten the last mile. Shipping gets cheaper – orders now travel a shorter distance to part of the customer base. That's the number that goes into the business case. What doesn't: the same inventory, split across two sites, needs its own safety stock at each one, because no single site can serve every order and demand at one site alone is harder to predict than pooled demand at one. That drives up storage, inbound freight, staffing and the systems needed to keep stock in sync – costs that never appear on the delivery line, because they were never modeled on it.

The practical conclusion isn't 'model every factor' – the obvious cost drivers already get modeled, because they're what the business case is built to justify. What gets missed is the cost with no obvious owner: the second safety stock, the extra freight lane, the sync layer nobody assigned a line to. Those are the numbers that decide whether the plan pays off – and a single-factor model was never built to go looking for them.