One order, two warehouses: deciding where the stock lands

Two questions hide inside one reorder

A single-location store answers one question when a SKU runs low: how many. Add a second stocking site and the same row quietly carries a second question that looks like a detail and isn't — which building the units arrive at. The two run on different inputs, fail in different ways, and cost different amounts to get wrong. Merged into one motion, they collapse into the reflex answer: take the total, split it down the middle, move on.

A 50/50 split is not a decision but the absence of one — symmetry in the order standing in for an assumption of symmetry in the demand that is never stated, and so never checked.

Pulling the two apart is the whole method. The quantity question belongs to the business as a whole; the destination question belongs to the individual sites. They deserve separate evidence.

Plan the quantity on pooled demand, not two half-forecasts

The intuitive way to size a multi-location order is to forecast each site and add the forecasts together. It feels rigorous, and it systematically over-buys. The underlying pooling argument is worked through in Your Marketplace and Your Store Are Drawing From the Same Shelf, where the pool is a single physical pile serving two channels.

Physical locations are the harder case, because the pile is genuinely split. The rule that matters here is the square-root law: total system inventory scales roughly with the square root of the number of stocking locations (Acumen) — a planning heuristic, not a figure to trust to the unit, but reliable in direction. Average demand still pools cleanly: two sites selling a combined hundred a week need a hundred a week. It's the buffer that multiplies, and it multiplies with the count of buildings rather than with the demand.

So size the order against total need, once. Then treat the destination as a separate, later question.

Choose the receiving location from each site's own sell-through

Having pooled the quantity, the allocation question becomes concrete: which site currently has the least cover relative to its own rate of sale?

The working measure is days of cover per location — on-hand there, divided by what sells there — not units on hand. A site holding twice the stock can still be the more urgent one if it sells three times as fast. Local lead time matters too, and it genuinely differs by site: an inland warehouse fed from the same supplier as a port-adjacent one waits longer for the same pallet, so it needs topping up earlier (The Retail Exec). Replenishment cadence can differ too, with busier sites often suiting more frequent, smaller top-ups.

The standard default when several sites want stock at once is fair-share proration: allocate in the ratio of each location's forecast, or bring every location to the same days of supply (NetSuite). Both are defensible; both are decisions. An even split is neither.

Transfer cost is what makes the receiving choice asymmetric

A quantity error is self-correcting. Order too much and the next cycle's suggestion absorbs it, because whatever is on hand and inbound reduces what gets suggested next time. The mistake decays.

A destination error doesn't. Stock that landed at the wrong site stays there until someone pays to move it — freight, handling, and the in-transit days during which the units serve nobody. In Shopify, inventory is tracked per location, and moving it means creating a transfer from one location and receiving that transfer at the other (Shopify Help Center). Transfers are a legitimate tool, and sometimes faster and cheaper than a fresh supplier order, but they're for closing short-term gaps rather than standing in for replenishment (Doss).

That asymmetry carries a bias: when two sites look close, favour the one that can absorb the stock — the higher-velocity site, or the one with capacity to hold it — because a small surplus there is cheap and a shortfall at a thin site is not.

Where each decision lands in Assisty

The sequence maps onto the buying workflow, and the order of operations is the point: quantity first, destination at commit.

The Smart Buying Table sets the quantity. Every suggestion is computed from sales velocity, lead times, and incoming stock, then rounded to case sizes and capped at maximum stock levels — so "how much" is answered against real constraints before the destination comes up at all.

The destination is asked at order creation. When you create orders, Assisty splits your selection by supplier — exact item counts, exact order values — and asks which location receives the stock; starting from a supplier, the PO is pre-filled with suggested quantities, the supplier's tax and currency, and a delivery location. That placement is deliberate: it's the last thing you answer, after the quantity is settled.

For the per-location read that informs the choice, the AI Report Builder answers inventory, replenishment, and stock-availability questions by location, among other dimensions — which is where the per-location on-hand and sell-through numbers come from rather than from memory. The cover calculation is yours to make on top of them. And if the destination isn't obvious yet, the Purchase List holds the items, priced with each supplier's tax and currency, until you've finished reviewing and are ready to order. Once committed, the order continues past the draft: review line items, email the PO to the supplier, and receive against it when stock arrives.