Allocation vs. Replenishment: Getting the Right Stock to the Right Store
Allocation decides where the first units go. Replenishment keeps them there. Confusing the two produces stores drowning in size 8 while three miles away it is sold out.
On this page
Two decisions govern where inventory sits in a store network, and they fail in different ways.
Allocation distributes a fixed quantity — a first receipt, a limited drop — across locations before demand signal exists. Replenishment tops locations back up against observed demand and a service target.
When teams run allocation with replenishment logic, new products get pushed toward stores with high historical volume in other products. When they run replenishment with allocation logic, mature products get equal-share treatment and the top stores starve.
Allocation without history
The hard part of allocation is that the item is new. There is no demand history for the thing you are allocating, so you have to borrow signal:
- Attribute similarity. Find comparable items by category, price band, brand, colour family, and seasonality profile; use their store-level demand shape.
- Store demand profile. Some stores over-index on premium price bands, some on entry price; the profile transfers across items even when the item does not.
- Space and presentation minimums. A store that cannot merchandise fewer than 6 units should not be allocated 2.
- Size and variant curves. These can vary by location and change over time, so validate them at the level where the allocation will be used.
A useful sanity check: the allocation should reproduce known local skews. If your beachfront store historically over-indexes on the largest sizes and the allocation sends it a national average curve, the model is not using store profile.
Replenishment once signal exists
As sell-through accumulates, the item's own store-level history becomes another useful signal, and the job becomes:
- Forecast demand per store per period at the replenishment horizon.
- Size buffers from forecast error and lead-time variability.
- Order to cover, respecting case packs, minimums and shelf capacity.
The transition between the two regimes should be gradual — a weighted blend that shifts from comparable-item signal to own-item signal as observations accumulate — not a hard switch at week 4.
Rebalancing: the third decision nobody owns
Even with good allocation and good replenishment, stock ends up in the wrong place. Weather diverges, a local event shifts demand, one store merchandises better than another.
Inter-store transfers are one possible correction. A candidate rule set might be:
| Condition | Action |
|---|---|
| Source has > X weeks cover and destination < Y weeks | Candidate transfer |
| Transfer cost < expected margin recovered | Approve |
| Item within N weeks of markdown | Prioritise — value decays fast |
| Both locations short | Do not transfer; escalate to buy |
The economics need to be calculated rather than assumed. Compare handling and freight with the expected margin recovered, while accounting for forecast uncertainty at both locations.
Measure the three separately
Aggregate availability hides which of the three decisions is failing. Track:
- Allocation quality — sell-through variance across stores in the first 4 weeks
- Replenishment quality — forecast bias and in-stock rate at the SKU-store level
- Rebalancing quality — units transferred vs. units eventually marked down
If early sell-through dispersion is wide, investigate allocation inputs before assuming replenishment tuning is the only fix.
Sources and further reading
These references support the technical concepts discussed above. Examples and recommendations in the article remain editorial interpretation.