Inventory Allocation and Replenishment
Inventory allocation is the decision that turns a bought quantity into a stock position at each door, channel or account; replenishment is the rule that refills that position as it sells. Where both sit in the plan-buy-allocate flow, the inputs, the methods stated as formulas, a fully worked five-door example from a 1,200-unit buy with a 15% holdback, the measures, and how allocation goes wrong.
Inventory allocation is the decision that turns a bought quantity into a stock position: which units of a style, in which sizes, go to which door, channel or wholesale account, and when. Replenishment is the rule that refills each of those positions as it sells, from a warehouse holdback or from a supplier, on a trigger and a review cadence that were set before the season opened. Together they are the point at which a plan that existed as a receipt budget and an assortment becomes units on a fixture in a particular store, and they are where a correct buy is converted into full-price sell-through or quietly wasted. This guide is the cluster's definitional page for the discipline: where allocation sits in the plan-buy-allocate flow, the four decisions that make it up, the inputs, the methods stated as formulas, the replenishment logic, a fully worked example that allocates one style across five doors and then replenishes it two weeks later, the measures, the ways it goes wrong, and a short pass across ten verticals.
Two sibling guides carry the depth this page points to rather than repeats. Allocation and replenishment by vertical follows the grain, the refill lead time and the exit through nine categories with a worked min/max example in two of them, and core and NOOS replenishment programs covers the standing program for items that are never meant to sell through. How modern apparel brands approach sizing and replenishment establishes the limit both of them rest on: replenishment cannot rebuild a size curve that was wrong at the buy. Readers who arrived looking for the software category, or for planning and allocation defined as a function, with the planner, the allocator and the buyer and the handoffs between them, should start at what is retail allocation software?, which defines the system these methods run in and returns here for the methods themselves.
What inventory allocation and replenishment are
The distinction that matters most is between a forecast and a response. Allocation is made once per receipt, on expected rates of sale, before a single unit of the style has sold in the doors it is going to. It is the best placement the history can support, and it will be wrong in detail at every door, because the history was built from different styles in different weeks. Replenishment runs on a cadence and responds to what actually sold, which is why a well-run allocation deliberately leaves some of the buy in the distribution center: the holdback is the part of the decision that is deferred until the doors have spoken.
Both sit downstream of the buy and upstream of the markdown. Upstream of them, the open-to-buy set the receipt budget, the assortment plan set the styles and depth, and the buy plan committed quantities by style-color and size to a vendor. Downstream, whatever allocation and replenishment leave in the wrong door at the end of the window becomes the residual that markdown and exit strategy have to clear. The allocation workflow on the platform side, allocation planning, describes the same position: pre-season allocation built from the confirmed buy and in-season reallocation when door-level sell-through diverges from it.
Where allocation sits in the plan-buy-allocate flow
The flow runs plan, buy, allocate, sell, replenish, and allocation is the hinge between the two halves. Everything before it is a quantity without a location; everything after it is a location with a quantity.
| Step | Decision | Grain | When |
|---|---|---|---|
| Financial plan and open-to-buy | How much receipt budget the season gets, by month and department | Department or class by month | Pre-season, months out |
| Assortment and line plan | Which styles, colors and price points, at what breadth and depth | Style-color | Pre-season |
| Buy plan and purchase order | Units by style-color and size, to a vendor, for a delivery window | Style-color by size | Pre-season, at the order deadline |
| Initial allocation | Which doors, channels and accounts receive which units of the receipt, and how much stays in the DC | Style-color by size, by door | After the PO, before the first receipt |
| Selling | The doors trade; door-level sales and on-hand become readable | Door by size by week | In season |
| Replenishment, reallocation, consolidation | Refill from the holdback or the vendor, move units between doors, pull back what cannot finish | Door by size | In season, on the review cadence |
| Markdown and exit | Clear the residual the previous steps left | Style-color by door | End of window |
Two consequences follow from the position. First, allocation inherits every error made above it. A size curve that was wrong at the buy arrives at the DC in the wrong proportions, and no door-level cleverness changes the units in the building. Second, allocation is the first step in the flow that produces a door-level record, and that record is what the season hindsight needs to tell a demand miss from a placement miss. A style that stocked out may have been under-bought, correctly bought and sent to the wrong doors, sent to the right doors on the wrong curve, or held in the DC while the doors broke. Sales history alone reads every one of those as an underbuy. Keeping the allocation decision, the curve it used, the holdback and the replenishment log next to the sell-through is what lets the hindsight write the right rule for next season.
The four allocation decisions
Allocation is spoken of as one decision. It is four, and they differ in their trigger, their inputs and their cost.
| Decision | Trigger | Main inputs | Cost of getting it wrong |
|---|---|---|---|
| Initial allocation | A receipt is due at the DC | Door grades and clusters, forecast rate of sale, presentation minimums, size curves, pack constraints | Stockouts at fast doors and residual at slow ones, from day one |
| Replenishment | On hand at a door-size position falls to a trigger, or a scheduled review finds a need | Trailing rate of sale, on hand, on order, lead time, target cover, holdback availability by size | Broken sizes at doors that were selling; holdback stranded in the DC |
| Reallocation and transfers | Door-level cover diverges: some doors well above the window, others below | Weeks of supply by door, remaining window, transfer cost, margin on the units moved | Freight and handling spent on units that arrive too late or recover less margin than they cost |
| Consolidation | A door cannot finish a style inside the window | Projected residual by door, display minimum, exit plan | A broken presentation at the losing door; a late consolidation that arrives after the markdown |
Initial allocation divides the first receipt between doors and the DC: the decision most planning effort goes into and the one with the least information, because it is made entirely on history. Replenishment sends units from the holdback, or from a supplier under a standing program, to door-size positions that have sold down; it is where the deferred part of the initial allocation is finally decided, on the style's own sales.
Reallocation and transfers move units between doors rather than from the DC, and carry a cost replenishment does not: a unit moved store to store is picked, packed, shipped and received twice, by store staff, and loses a week or more of selling time in transit. The in-season chase covers the related decision of reordering from the vendor instead. Consolidation is the reverse of allocation: pulling a style back from doors that cannot finish it into fewer doors, or into the e-commerce node, so that it sells at full price somewhere rather than at markdown everywhere, weighed against taking the markdown in place.
The inputs
Every allocation method is a way of combining the same handful of inputs. The quality of the output is set by the quality of these before any formula is chosen.
Door grades and clusters
A door grade ranks doors by volume for a category: A, B, C doors, or a numeric tier. A cluster groups doors that share a demand profile, which may cut across grades: two doors of the same volume can want different size curves, different color mixes and different depth. Store clustering for apparel covers how clusters are built; for allocation the working rule is that the grade sets how much and the cluster sets what mix. Grades built from total store revenue mislead at the category level, because a door that ranks high overall may rank low for the category being allocated. The door-level demand index is the usual way of expressing a door's demand for a category relative to the chain.
Sales history and rate of sale
The forecast rate of sale per door is the core input to fair-share and need-based methods. It is built from comparable styles in comparable weeks, adjusted for what is known to have changed at the door. The adjustment that is most often skipped is for availability: a door's history is demand filtered by what was in stock, so a door that broke its core sizes early last season recorded less than it would have sold, and allocating to that history under-serves it again. The in-stock rate is read beside the history before a rate is trusted.
Presentation minimums and fixture capacity
A presentation minimum is the smallest quantity that makes a credible floor set for the style at a door: enough units across sizes to fill the fixture arm or wall bay the style is planned on. Below it the style looks picked over on the day it lands. Fixture capacity is the ceiling: units beyond what the fixture and the back room will hold either crowd the presentation or sit in a stockroom where they are not sold. Both are set per door, or per fixture type, and both override a pure demand calculation, which is why they are applied after the fair share rather than built into it.
Size curves and pack constraints
A size curve is the share of a style's units in each size. The curve that matters for allocation is the door's or the cluster's, not the brand's, because a single curve sent everywhere produces broken core sizes in doors that skew one way and fringe residual in doors that skew the other. Size-curve allocation resolves a door's units into units by size. A pack constraint exists when the vendor ships pre-packed cartons in a fixed size ratio: a door's allocation then has to be a whole number of packs, or packs plus loose units, and the rounding is a real decision rather than an afterthought. Pre-pack vs singles covers the trade.
Lead times
Three lead times govern the decisions. The vendor lead time sets whether a chase is possible inside the window at all. The DC-to-door lead time sets how far ahead a replenishment has to be placed to land before a position breaks. The review period, the gap between one replenishment review and the next, adds to the DC-to-door lead time in every cover calculation, because a position has to survive until the next review and then the shipment. Lead time covers the definitions.
DC holdback
The DC holdback is the share of the receipt kept in the distribution center at the initial allocation. Its purpose is to defer part of the placement decision until the doors have traded, and its size is a judgement set by five things: how long the selling window is, how quickly door-level sales become readable, how long the DC-to-door lead time is, whether a vendor chase is possible inside the window, and how costly it is to move units between doors later. A short window with no chase argues for a smaller holdback, because units in the DC in week four of a six-week window are units that will not sell at full price; a long window with expensive transfers argues for a larger one. There is no correct percentage independent of those five facts, and the worked example below uses 15% as an illustration, not a recommendation. The holdback is also, as the example shows, a residual by size: whatever the buy curve supplied that the door curves did not take.
The allocation methods
Each method below is stated as a formula where one exists, with a short illustration. The fully worked example in the next section resolves them together on one style.
Fair share, or proportional, allocation
Door units = units to allocate × (door demand ÷ total demand across the doors)
Demand is a forecast weekly rate of sale, or a season demand figure, built from comparable history. Fair share is the baseline that every other method adjusts: it puts units where the demand is, in proportion, and nothing else. In the worked example, 1,020 units are allocated across five doors whose planned weekly rates sum to 180, so a door planned at 60 a week receives 1,020 × (60 ÷ 180) = 340 units. The fair-share allocation formula page carries the calculator and the common overrides.
Need-based allocation
Need = (target weeks of cover × weekly rate of sale) − on hand − on order, floored at zero.
Need-based allocation sends each door what it needs to reach a target position rather than a proportional share of what is available. It is the natural method for a second or later receipt, and for replenishment, because it reads the door's current on hand. On a first receipt, with on hand at zero, it collapses into cover-based fair share: each door gets target weeks × its rate. When the sum of needs exceeds the units available, the shortfall is scaled back in proportion to need, or by priority, which is where the method meets fair share again.
Top-down by cluster or grade, then door
Grade units = units to allocate × (grade demand ÷ total demand), then door units = grade units × (door demand ÷ grade demand).
Allocating first to a grade or cluster and then to the doors inside it gives the planner a place to apply judgement at the cluster level, where the history is deep, before the arithmetic reaches the door, where it is thin. When the door rates are consistent with the grade rates, the result is identical to door-level fair share; the worked example shows this. The method earns its place when the cluster-level split is deliberately changed from the proportional one, for example to protect a new cluster that has no history, and the door split inside each cluster is then left to the formula.
Size-curve allocation
Size units = door units × size share for the door's curve, with the shares summing to 100% and the units rounded to whole numbers that still sum to the door total.
The rounding is not trivial. Multiplying 333 units by a five-size curve produces five decimals, and rounding each to the nearest whole number can sum to 332 or 334. The standard fix is the largest-remainder rule: take the whole-number part of each size, and give the remaining units, one each, to the sizes with the largest fractional parts. The worked example applies it at every door.
Pack-constrained rounding
Packs = door need ÷ pack size, rounded by the rule the business chose, and the door's units become packs × pack size, plus any loose units allowed.
With a pre-pack of twelve in a 1-3-4-3-1 ratio across XS to XL, a door needing 57 units receives either four packs, 48 units, or five packs, 60 units. Rounding up protects the fast doors and spends the holdback faster; rounding down protects the holdback and under-serves the fast doors on the day the floor set opens. The ratio inside the pack is also a size curve the door did not choose, which is why pack-constrained programs usually carry a loose-unit top-up for the sizes the pack under-supplies.
Holdback and chase
Holdback units = receipt × holdback share, then units to doors = receipt − holdback. The chase is the decision to reorder from the vendor rather than release holdback: it is possible only when the vendor lead time plus the DC-to-door lead time lands the units with enough window left to sell them.
In the worked example, 1,200 × 15% = 180 units are held and 1,020 go to the doors. Whether the holdback is released by need-based replenishment, by a second proportional allocation, or by a pre-set schedule is a separate decision; the example uses need-based replenishment because that is what a holdback is for.
Replenishment logic
Replenishment logic is a trigger, a quantity and a cadence. The trigger says when to act, the quantity says how much, and the cadence says how often the trigger is checked. The three families below differ mainly in how the quantity is set.
Min/max
Minimum = (lead time in weeks × weekly rate) + safety stock; maximum = minimum + one order multiple. When sellable on hand plus on order falls to or below the minimum, order up to the maximum.
Min/max is the standing rule for items with stable demand and a reliable refill: core basics, a replenished accessories color, a hard-goods consumable. Safety stock and the replenishment trigger formula pages carry the two components with calculators, and the by-vertical guide works the same rule for an accessories core refilled in weeks and a furniture core refilled by the container in months.
Order-up-to
Order quantity = order-up-to level − (on hand + on order), checked at every review.
An order-up-to rule drops the fixed minimum and simply tops the position back to a target at each review. The target is set as cover: lead time plus review period plus safety, expressed in weeks of the trailing rate. It suits a scheduled weekly review better than a continuous trigger, and it is the rule the worked example uses, with a four-week target.
Forecast-driven
Order quantity = forecast demand over (lead time + review period) + safety stock − (on hand + on order).
Forecast-driven replenishment replaces the trailing rate with a forward forecast, which matters at demand inflections: the run-up to a peak, the decline after it, a planned promotion. In a stable mid-season week it produces the same answer as the trailing rate. The forward weeks of supply formula is the cover measure that pairs with it.
The trigger and the review cadence
Whichever quantity rule is used, the trigger is checked on a cadence, and the cadence has to be at least as fast as the shortest lead time in the rule. A weekly review on a position that can break in four days is a rule that will be late once a month. The trigger also counts the right things: sellable stock only, excluding damages and display units; on order, so that a long lead time does not double-order; and it has a stop date. The last replenishment that can land and sell is placed at the exit date minus the lead time, and after that date the rule is switched off, or it keeps filling a position the calendar is about to close.
NOOS and core programs
A never-out-of-stock (NOOS) item is run on a standing replenishment rule rather than bought once a season, measured by in-stock rate rather than sell-through, and funded from the open-to-buy before the seasonal breadth is set. The rule above applies; the discipline around it, qualification, review cadence, re-setting the trigger and the OTB conflict, is the subject of core and NOOS replenishment programs, with the short form at never out of stock.
A fully worked example: five doors, two grades, one style
The figures below are illustrative. They were chosen to divide cleanly, they are not benchmarks or targets, and they are not drawn from any brand. Every number is recomputed in the text so that it can be checked.
The set-up
A women's crewneck sweater in one color, five sizes XS to XL, bought for a six-week full-price window with no vendor chase possible inside it. The buy is 1,200 units, and the team sets a 15% DC holdback, so 1,200 × 0.15 = 180 units stay in the DC and 1,200 − 180 = 1,020 units go to the doors.
Five doors in two grades. The planned weekly rate of sale for this style, built from comparable sweaters last autumn, is 60 at door A1, 48 at A2, 30 at B1, 24 at B2 and 18 at B3, a total of 60 + 48 + 30 + 24 + 18 = 180 units a week. Over the six-week window that is 6 × 180 = 1,080 units of planned demand against a 1,200-unit buy, a planned full-price sell-through of 1,080 ÷ 1,200 = 90%, with 120 units expected to reach the markdown.
The presentation minimum is 120 units per door: one fixture arm per size at opening depth. The DC-to-door lead time is one week, and replenishment is reviewed weekly.
Step 1: top-down by grade, then door
The A grade's planned rate is 60 + 48 = 108 a week and the B grade's is 30 + 24 + 18 = 72, out of 180. The A grade's share is 108 ÷ 180 = 60%, so it receives 1,020 × 0.60 = 612 units; the B grade's share is 72 ÷ 180 = 40%, so it receives 1,020 × 0.40 = 408 units. Check: 612 + 408 = 1,020.
Inside the A grade, A1 takes 612 × (60 ÷ 108) = 340 and A2 takes 612 × (48 ÷ 108) = 272; 340 + 272 = 612. Inside the B grade, B1 takes 408 × (30 ÷ 72) = 170, B2 takes 408 × (24 ÷ 72) = 136 and B3 takes 408 × (18 ÷ 72) = 102; 170 + 136 + 102 = 408.
Door-level fair share gives the same answer, because the door rates are consistent with the grade rates: 1,020 ÷ 180 = 5.667 units per unit of weekly rate, and 60 × 5.667 = 340, 48 × 5.667 = 272, 30 × 5.667 = 170, 24 × 5.667 = 136, 18 × 5.667 = 102. The top-down route would only differ if the planner had deliberately moved units between the grades.
Step 2: the presentation minimum
B3's fair share of 102 is below the 120-unit presentation minimum, so B3 is raised to 120, a shortfall of 120 − 102 = 18 units that has to come from the other four doors. The holdback is not touched: it exists for replenishment, and spending it on the floor set defeats it.
The 18 units are taken from the four doors above the minimum in proportion to their planned rates, which sum to 60 + 48 + 30 + 24 = 162. A1 gives 18 × (60 ÷ 162) = 6.67, A2 gives 18 × (48 ÷ 162) = 5.33, B1 gives 18 × (30 ÷ 162) = 3.33 and B2 gives 18 × (24 ÷ 162) = 2.67. The whole-number parts are 6, 5, 3 and 2, which sum to 16; the two remaining units go to the two largest fractional parts, A1 (0.67) and B2 (0.67), so the reductions are 7, 5, 3 and 3, summing to 18.
| Door | Grade | Planned rate / week | Fair share | Presentation-minimum adjustment | Allocated | Cover at plan rate |
|---|---|---|---|---|---|---|
| A1 | A | 60 | 340 | −7 | 333 | 5.55 weeks |
| A2 | A | 48 | 272 | −5 | 267 | 5.56 weeks |
| B1 | B | 30 | 170 | −3 | 167 | 5.57 weeks |
| B2 | B | 24 | 136 | −3 | 133 | 5.54 weeks |
| B3 | B | 18 | 102 | +18 | 120 | 6.67 weeks |
| Doors | 180 | 1,020 | 0 | 1,020 | ||
| DC holdback | 180 | |||||
| Buy | 1,200 |
Check: 333 + 267 = 600; 600 + 167 = 767; 767 + 133 = 900; 900 + 120 = 1,020. The adjustments sum to −7 − 5 − 3 − 3 + 18 = 0. Four doors open at about five and a half weeks of cover at the plan rate; B3 opens at 6.67 weeks, because the presentation minimum, not its rate, set its number. That extra cover at B3 is the price of a credible floor set, and it is the first place the team expects residual.
Step 3: size curves by grade
The two grades carry different size curves, read from comparable sweaters at those doors. The A-grade curve is XS 10%, S 25%, M 30%, L 23%, XL 12%, which sums to 100%. The B-grade curve is XS 6%, S 20%, M 30%, L 26%, XL 18%, which also sums to 100%. The B doors skew larger. Each door's allocation is multiplied by its grade's curve and rounded by the largest-remainder rule.
Door A1, 333 units on the A curve: XS 333 × 0.10 = 33.3, S 333 × 0.25 = 83.25, M 333 × 0.30 = 99.9, L 333 × 0.23 = 76.59, XL 333 × 0.12 = 39.96. Whole-number parts 33, 83, 99, 76, 39 sum to 330; the three remaining units go to the largest fractions, XL (0.96), M (0.90) and L (0.59), giving 33, 83, 100, 77, 40, which sum to 333.
Door A2, 267 units on the A curve: 26.7, 66.75, 80.1, 61.41, 32.04. Whole parts 26, 66, 80, 61, 32 sum to 265; the two remaining units go to S (0.75) and XS (0.70), giving 27, 67, 80, 61, 32, which sum to 267.
Door B1, 167 units on the B curve: XS 167 × 0.06 = 10.02, S 167 × 0.20 = 33.4, M 167 × 0.30 = 50.1, L 167 × 0.26 = 43.42, XL 167 × 0.18 = 30.06. Whole parts 10, 33, 50, 43, 30 sum to 166; the one remaining unit goes to L (0.42), giving 10, 33, 50, 44, 30, which sum to 167.
Door B2, 133 units on the B curve: 7.98, 26.6, 39.9, 34.58, 23.94. Whole parts 7, 26, 39, 34, 23 sum to 129; the four remaining units go to XS (0.98), XL (0.94), M (0.90) and S (0.60), giving 8, 27, 40, 34, 24, which sum to 133.
Door B3, 120 units on the B curve: 7.2, 24, 36, 31.2, 21.6. Whole parts 7, 24, 36, 31, 21 sum to 119; the one remaining unit goes to XL (0.60), giving 7, 24, 36, 31, 22, which sum to 120.
| Door | XS | S | M | L | XL | Total |
|---|---|---|---|---|---|---|
| A1 | 33 | 83 | 100 | 77 | 40 | 333 |
| A2 | 27 | 67 | 80 | 61 | 32 | 267 |
| B1 | 10 | 33 | 50 | 44 | 30 | 167 |
| B2 | 8 | 27 | 40 | 34 | 24 | 133 |
| B3 | 7 | 24 | 36 | 31 | 22 | 120 |
| Allocated to doors | 85 | 234 | 306 | 247 | 148 | 1,020 |
| Bought (buy curve) | 108 | 288 | 360 | 288 | 156 | 1,200 |
| DC holdback (residual) | 23 | 54 | 54 | 41 | 8 | 180 |
Column checks: XS 33 + 27 + 10 + 8 + 7 = 85; S 83 + 67 + 33 + 27 + 24 = 234; M 100 + 80 + 50 + 40 + 36 = 306; L 77 + 61 + 44 + 34 + 31 = 247; XL 40 + 32 + 30 + 24 + 22 = 148. Row check: 85 + 234 + 306 + 247 + 148 = 1,020.
Step 4: the holdback by size
The buy was placed months earlier on a blended brand curve of XS 9%, S 24%, M 30%, L 24%, XL 13%, which gives 108, 288, 360, 288 and 156 units, summing to 1,200. The holdback is what the door curves left behind: XS 108 − 85 = 23, S 288 − 234 = 54, M 360 − 306 = 54, L 288 − 247 = 41, XL 156 − 148 = 8. Check: 23 + 54 + 54 + 41 + 8 = 180.
The holdback is 15% of the buy in total and about 5% of the buy's XL: 8 ÷ 156 = 5.1%. The doors, read on their own curves, wanted more XL and less XS than the brand curve supplied, and the holdback has absorbed the difference. This is worth seeing before the season opens, because it says what the holdback can and cannot do two weeks from now.
Step 5: two weeks of selling
The window opens and the doors trade for two weeks. A1 sells 136 units, A2 88, B1 72, B2 40 and B3 28, a total of 136 + 88 + 72 + 40 + 28 = 364 units. On hand by door is the allocation less sales: A1 333 − 136 = 197, A2 267 − 88 = 179, B1 167 − 72 = 95, B2 133 − 40 = 93, B3 120 − 28 = 92, summing to 197 + 179 + 95 + 93 + 92 = 656. Check: 1,020 − 364 = 656.
| Door | Allocated | Sold, weeks 1–2 | Trailing rate / week | Plan rate / week | On hand | Sell-through | Weeks of supply |
|---|---|---|---|---|---|---|---|
| A1 | 333 | 136 | 68 | 60 | 197 | 40.8% | 2.90 |
| A2 | 267 | 88 | 44 | 48 | 179 | 33.0% | 4.07 |
| B1 | 167 | 72 | 36 | 30 | 95 | 43.1% | 2.64 |
| B2 | 133 | 40 | 20 | 24 | 93 | 30.1% | 4.65 |
| B3 | 120 | 28 | 14 | 18 | 92 | 23.3% | 6.57 |
| Doors | 1,020 | 364 | 182 | 180 | 656 | 35.7% | 3.60 |
Sell-through is units sold divided by units received: 136 ÷ 333 = 40.8%, 88 ÷ 267 = 33.0%, 72 ÷ 167 = 43.1%, 40 ÷ 133 = 30.1%, 28 ÷ 120 = 23.3%, and 364 ÷ 1,020 = 35.7% in total. Weeks of supply is on hand divided by the trailing weekly rate: 197 ÷ 68 = 2.90, 179 ÷ 44 = 4.07, 95 ÷ 36 = 2.64, 93 ÷ 20 = 4.65, 92 ÷ 14 = 6.57, and 656 ÷ 182 = 3.60 in total.
The chain total says the style is on plan: 182 a week against 180 planned, 3.60 weeks of cover with four weeks of window left. The doors say something else. B1, a B-grade door, has the highest sell-through in the chain at 43.1%, and is the first door that will break; B3 is carrying 6.57 weeks of cover into a four-week window. The grade was right about A1 and wrong about B1 for this style.
Step 6: the replenishment order
The team's replenishment rule for this drop is order-up-to with a four-week target cover at the trailing rate: four weeks is the whole of the remaining window, the DC-to-door lead time is one week, and with no vendor chase this is the last review at which units can be placed and still sell for most of what is left. Need = (4 × trailing weekly rate) − on hand − on order, floored at zero; nothing is on order.
At door level: A1 needs 4 × 68 = 272 − 197 = 75; A2 needs 4 × 44 = 176 − 179 = −3, so 0; B1 needs 4 × 36 = 144 − 95 = 49; B2 needs 4 × 20 = 80 − 93 = −13, so 0; B3 needs 4 × 14 = 56 − 92 = −36, so 0. Door-level need is 75 + 49 = 124 units, comfortably inside a 180-unit holdback.
But need is calculated by size, because a door does not sell "units"; it sells sizes, and a door total nets a surplus in one size against a shortage in another. Door A1 sold its 136 units as XS 10, S 30, M 46, L 32, XL 18, which sum to 136. Its on hand by size is XS 33 − 10 = 23, S 83 − 30 = 53, M 100 − 46 = 54, L 77 − 32 = 45, XL 40 − 18 = 22, summing to 197. Its trailing weekly rate by size is sales divided by two: XS 5, S 15, M 23, L 16, XL 9, summing to 68.
| A1 by size | Allocated | Sold | On hand | Rate / week | Target (4 × rate) | Need |
|---|---|---|---|---|---|---|
| XS | 33 | 10 | 23 | 5 | 20 | 0 (surplus 3) |
| S | 83 | 30 | 53 | 15 | 60 | 7 |
| M | 100 | 46 | 54 | 23 | 92 | 38 |
| L | 77 | 32 | 45 | 16 | 64 | 19 |
| XL | 40 | 18 | 22 | 9 | 36 | 14 |
| A1 | 333 | 136 | 197 | 68 | 272 | 78 |
A1 needs 75 units at door level and 78 units by size: 7 + 38 + 19 + 14 = 78. The three-unit difference is the XS surplus, which the door total silently netted against the M shortage. A door-level replenishment of 75 units on the door's curve would have sent XS the door does not need and under-shipped the M it does.
Door B1 sold its 72 as XS 4, S 14, M 24, L 20, XL 10, which sum to 72. Its on hand by size is XS 10 − 4 = 6, S 33 − 14 = 19, M 50 − 24 = 26, L 44 − 20 = 24, XL 30 − 10 = 20, summing to 95. Its trailing rate by size is XS 2, S 7, M 12, L 10, XL 5, summing to 36. Targets at four weeks are 8, 28, 48, 40 and 20, summing to 144. Need is XS 8 − 6 = 2, S 28 − 19 = 9, M 48 − 26 = 22, L 40 − 24 = 16, XL 20 − 20 = 0, summing to 49, the same as the door-level figure, because B1 has no size in surplus.
The request to the DC is the sum by size: XS 0 + 2 = 2, S 7 + 9 = 16, M 38 + 22 = 60, L 19 + 16 = 35, XL 14 + 0 = 14, a total of 2 + 16 + 60 + 35 + 14 = 127 units. Check: 78 + 49 = 127.
Step 7: filling from the holdback
| Size | Requested | In holdback | Shipped | Short | Holdback after |
|---|---|---|---|---|---|
| XS | 2 | 23 | 2 | 0 | 21 |
| S | 16 | 54 | 16 | 0 | 38 |
| M | 60 | 54 | 54 | 6 | 0 |
| L | 35 | 41 | 35 | 0 | 6 |
| XL | 14 | 8 | 8 | 6 | 0 |
| Total | 127 | 180 | 115 | 12 | 65 |
Checks: shipped 2 + 16 + 54 + 35 + 8 = 115; short 6 + 6 = 12, and 127 − 115 = 12; holdback after 21 + 38 + 0 + 6 + 0 = 65, and 180 − 115 = 65. The fill rate on the replenishment is 115 ÷ 127 = 90.6%.
The 54 units of M are split between the two requesting doors in proportion to need: A1 receives 54 × (38 ÷ 60) = 34.2, rounded to 34, and B1 receives 54 × (22 ÷ 60) = 19.8, rounded to 20; 34 + 20 = 54. The 8 units of XL all go to A1, the only door that asked. A1's shipment is S 7 + M 34 + L 19 + XL 8 = 68 units, and B1's is XS 2 + S 9 + M 20 + L 16 = 47 units; 68 + 47 = 115. After receipt A1 holds XS 23, S 60, M 88, L 64, XL 30, a total of 197 + 68 = 265, with M at 88 ÷ 23 = 3.8 weeks and XL at 30 ÷ 9 = 3.3 weeks of cover against the four-week target.
The holdback was 180 units against a request of 127 and still fell 12 units short, because it was 15% of the buy in total and about 5% of the buy's XL. If the trailing rates hold, those 12 units, 6 M and 6 XL, are the lost sales the holdback could not prevent, and they are concentrated in the two sizes the A-grade doors were selling fastest. What remains in the DC, 21 XS and 38 S and 6 L, is 65 units that only an XS or S shortage can use; the week-three review will be about consolidation and transfers, not replenishment.
The transfer question is already visible in the week-two table. B3 is carrying 92 units at 14 a week into a four-week window, a projected residual of 92 − 56 = 36 units; B2 projects 93 − 80 = 13 and A2 projects 179 − 176 = 3, a total of 52 units of expected residual at three doors while A1 and B1 are 12 units short. Moving a dozen units of M and XL from B3 to A1 is arithmetically obvious, and it is worth doing only if the margin on the units moved exceeds the freight and handling to move them, and only if they land with weeks left to sell. That is a decision for the week-three review, with B3's sales by size in front of it.
The measures
Allocation and replenishment are measured at the door-and-size grain, because that is where the failures happen and where chain averages hide them. The worked example illustrates each.
| Measure | Formula | In the worked example | What it tells you |
|---|---|---|---|
| In-stock rate | Door-size positions with sellable stock ÷ planned positions | 25 of 25 positions in stock at the end of week two, with A1's M at 54 ÷ 23 = 2.3 weeks from empty | Whether the customer could buy their size, before sales history records the miss |
| Fill rate | Units shipped ÷ units requested | 115 ÷ 127 = 90.6% on the replenishment | Whether the DC could answer what the doors asked for |
| Weeks of supply by door | On hand ÷ trailing weekly sales | From 2.64 at B1 to 6.57 at B3 | Where cover is short of the window and where it exceeds it |
| Sell-through by door | Units sold ÷ units received | From 43.1% at B1 to 23.3% at B3 after two weeks | Which doors the allocation under- and over-served |
| Lost sales | Estimated units a position would have sold while empty, from its rate before the stockout | 12 units if the trailing rates hold and nothing else arrives | The cost of the gap that sales history never shows |
In-stock rate is read at the position, a door and a size in a week, rather than at the style, because a style can read as in stock at every door while its core sizes are empty; the in-stock rate entry covers why it is read before any sales history is trusted. Fill rate is the DC's measure: the fill rate formula page covers the wholesale version, where a shortfall against an account's order becomes a chargeback, and for store replenishment the same arithmetic says whether the holdback was the right size and the right shape.
Weeks of supply by door is read by door and preferably by size, never as a chain average: the example's 3.60 weeks said the style was fine, and B1 at 2.64 and B3 at 6.57 said where the units were wrong. The weeks of supply formula page covers the trailing window and when to switch to the forward version. Sell-through by door is what the allocation is judged against at the hindsight, and a door's sell-through rate is only comparable to another's when both were in stock. Lost sales have to be estimated, because a position with no stock records no sales whatever its demand; the lost sales entry covers the methods and their limits, and the door-level demand index turns a chain rate into a door's.
How allocation goes wrong
Each of these produces an allocation that looks reasonable on the sheet.
Allocating to history that was itself out of stock
Last season's door history is demand filtered by last season's availability. A door that broke its M and L in week three recorded a smaller M and L share than it wanted, and an allocation built on that history sends it less M and L again, breaks it earlier, and records a still smaller share. The error compounds season over season and reads, in the hindsight, as a door that "does not sell larger sizes". The fix is to read in-stock beside sales before a rate or a curve is trusted, and to estimate the censored demand rather than accept the recorded figure.
Averaging size curves across grades
The example's A and B doors wanted different curves: 12% XL at the A doors and 18% at the B doors. A single chain curve of 15% sent everywhere over-supplies XL at the A doors and under-supplies it at the B doors, and both read as a size problem rather than an allocation problem. The same error at the buy is worse, because it fixes the units in the building: the example's brand curve supplied 156 XL where the door curves wanted 148 plus a holdback that could answer the fastest doors, and the holdback ended up with 8. Size and pack optimization covers building curves at the grain the history supports.
Ignoring presentation minimums
A pure fair share sent B3 102 units, which on a 120-unit floor set is a fixture with gaps on the day it opens. A floor set that opens thin sells thin, because the customer reads a picked-over fixture as a style that is on its way out. The reverse error exists too: a presentation minimum applied in-season, as a floor the replenishment keeps topping up, turns a slow door into a residual machine. The presentation minimum is an opening constraint; the in-season floor is a lower display minimum below which the door is consolidated, not replenished.
Chasing with the wrong size mix
When a style breaks, the instinct is to chase units. The example shows what a chase actually needs: 6 M and 6 XL at specific doors, not 12 units of the style. A chase placed on the brand curve, or on the door's original allocation curve, arrives with XS and S the doors cannot use and less M and XL than they asked for, and because the chase lands late in the window the surplus has nowhere to go. A chase is sized from the size-level need at the doors that will receive it, and it is only placed when the vendor lead time plus the DC-to-door lead time lands it with weeks left to sell. The in-season chase covers the decision in full.
Transfers that cost more than the margin they recover
A store-to-store transfer is picked, packed, shipped and received twice, by store staff, and the units are not selling anywhere while they travel. For a low-ticket style, or a transfer of a handful of units, the freight and handling can exceed the gross margin the units will earn at the receiving door, and a transfer that lands with a week of window left recovers almost nothing. The test is the margin on the units moved, net of the cost to move them, against what the same units would recover at markdown in place. Transfers are a correction to a wrong allocation, and the better investment is usually the next allocation.
Reading the chain average
Every table in the worked example had a chain total that said the style was fine: 182 a week against 180 planned, 3.60 weeks of cover, 35.7% sell-through. The doors underneath it ranged from 2.64 to 6.57 weeks of cover and from 23.3% to 43.1% sell-through. An allocation review that reads the total has nothing to act on; one that reads by door and size has a transfer, a consolidation and a chase to decide.
Allocation and replenishment across ten verticals
The methods above are the same in every category; what changes is the grain the product sells at, the lead time on the refill and what forces the exit. The paragraphs below are the short form; allocation and replenishment by vertical carries the depth, with the invariant table and a worked min/max example in two categories.
Apparel
The grain is the style-color by size, to a door or a cluster, and the first decision is whose size curve a door receives. The holdback refills in days from the DC and the chase, where held fabric allows it, in weeks, so the window usually admits one meaningful replenishment and at most one chase. The size curve is the one thing replenishment cannot repair if the buy got it wrong. The by-vertical guide covers the apparel model beside the other eight.
Footwear
The grain is the size run per model-color, allocated as pairs per door by size and width fit, because a run with its core sizes missing is not sellable at all. Replenishment is run repair from the DC, and the vendor refill is tied to the model year: a carryover model replenishes on a rule across seasons, while a model facing changeover is allocated once and run down. The by-vertical guide covers runs, widths and breakage.
Accessories and bags
A replenished core, the black belt and the everyday tote, runs on min/max from a supplier on a leather-and-hardware lead time measured in weeks, while the hero color and the seasonal shape are allocated once and exited with the collection. Attach rate to the apparel assortment sets which doors carry depth. The by-vertical guide covers the core-under-fashion structure.
Home and furniture
The grain is the container to a DC and the floor sample to a door, with long lead times covering the factory slot and ocean transit, so the minimum carries months of cover and is correct. Container minimums set the order multiple, floor sets are allocated as samples rather than stock, dealer prebooks commit the quantity before the season with no refill behind them, and landed cost makes the maximum a capital decision. The by-vertical guide works the furniture min/max example in full.
Outdoor
Product runs on a model year, with the season's quantity committed at the dealer prebook and an at-once pool held back for in-season orders, so most of the allocation is made before a unit ships and the holdback is the only lever. The weather window sets the stop date, and carryover into the next model year is a markdown decision made at allocation time. The by-vertical guide covers prebook versus at-once.
Sporting goods
Seasonality runs by sport, so the same door is allocated against a different calendar for each category, and the prebook sets the quantity for the sport's season with the at-once pool behind it. Hard-goods consumables, balls, grips and tape, run on domestic min/max all year, a replenished floor under a seasonal allocation. The by-vertical guide covers counter-seasonal categories and the hard-goods rule.
Health and beauty
The grain is the shade, and a shade range is allocated as a complete ladder per door, because a range with its ends missing reads as broken. Replenishment is bounded by batch shelf life and the retailer's required-remaining-life window, so the maximum is capped by the dating ceiling regardless of the order multiple, and gondola resets fix the dates at which a planogram changes. Testers are not sellable stock. The by-vertical guide covers the ladder and the dating ceiling.
Toys and games
Demand concentrates in the gifting peak and the licensed window sets a hard exit, so allocation is case multiples to a planogram facing ahead of the retailer reset, with replenishment from a domestic DC inside the peak and nothing from the factory. A licensed item left after the window is a write-off, not a markdown. The by-vertical guide covers the peak and the reset.
Baby and juvenile
Hard goods replenish on a long tail driven by the registry, so a car seat is allocated by production lot and replenished for years on a certified configuration, with recall traceability requiring the lot to be known by door. Apparel in the category follows the apparel model with a size curve by age band. The by-vertical guide covers registry-driven replenishment and lot traceability.
Jewelry and watches
The grain is the individual piece, refilled one-for-one from the bench or made to order, with depth per door set by capital and security rather than by rate. The precious metal cost base changes the refill cost between one order and the next, memo rotates pieces between doors instead of transferring owned stock, and watches run on a model year with changeover decided at allocation. The by-vertical guide covers memo and the capital review.
Where RetailNorthstar fits
RetailNorthstar is a merchandise planning platform: OTB, assortment and line planning, buy planning, allocation, sizing, PO and WIP tracking and analytics on a shared data model. As the allocation planning page describes it, pre-season allocation is built from the confirmed buy and OTB rather than in a separate file, allocation recommendations by door are generated from historical sell-through patterns, store capacity and style-level performance data, and in-season reallocation recommendations are surfaced when door-level sell-through signals an imbalance. Recommendations are a starting point: planners review and adjust by exception before the allocation is finalized.
The judgements in this guide remain the planning team's: the holdback share, the presentation minimums, which curve a door receives, the target cover and the review cadence, and whether a transfer is worth its cost. OTB planning holds the receipt budget that the buy, and therefore the allocation, is sized against.
Related resources
- Allocation and Replenishment by Vertical — the grain, the refill and the exit through nine categories, with a worked min/max example
- Core and NOOS Replenishment Programs — the standing program for items that never sell through
- How Modern Apparel Brands Approach Sizing & Replenishment — why replenishment cannot rebuild a wrong size curve
- Store Clustering for Apparel and Localized Assortment Planning — the grades and clusters allocation reads from
- The In-Season Chase — reordering from the vendor rather than moving existing units
- What Is Open-to-Buy Planning — the receipt budget upstream of the buy
- How to Hindsight a Season — where the allocation record is finally read
- Fair-Share Allocation, Size Curve Allocation, Door-Level Demand, Replenishment Trigger, Safety Stock, Weeks of Supply and Fill Rate formulas — the arithmetic, with calculators
- Allocation Optimization, Replenishment, Never Out of Stock, In-Stock Rate and Lost Sales — the terms
See how RetailNorthstar builds pre-season allocation from the confirmed buy and surfaces in-season reallocation recommendations from door-level sell-through, in one connected workflow.
Start Free Trial →No credit card. No commitments.
Common questions
What is inventory allocation?
Inventory allocation is the decision that turns a bought quantity into a stock position: which units of a style, in which sizes, go to which door, channel or wholesale account, and when. It is made after the buy is placed and before the first receipt lands, and it is a forecast, because the units are placed on expected rates of sale, not actual ones. Replenishment is the rule that refills each position as it sells, from a warehouse holdback or from a supplier, on a trigger and a review cadence set before the season opened.
What is the difference between allocation and replenishment?
Allocation is made once per receipt and places units on a forecast; replenishment runs on a cadence and responds to actual sales. Allocation answers how the buy is divided across doors on day one; replenishment answers how much each door gets back as it sells. The two meet at the DC holdback, the share of the buy deliberately kept out of the initial allocation so that replenishment has units to work with. Reallocation and transfers move units between doors after the fact, and consolidation pulls a style back from doors that cannot finish it.
What is the fair-share allocation formula?
Door units = units to allocate × (door demand ÷ total demand across the doors). Demand is a forecast weekly rate of sale or a season demand figure built from comparable history. In this guide's worked example, 1,020 units are allocated across five doors whose planned weekly rates sum to 180, so door A1, planned at 60 a week, receives 1,020 × (60 ÷ 180) = 340 units before the presentation-minimum adjustment. Fair share is the baseline; presentation minimums, fixture capacity, size curves and pack rounding are applied on top of it.
What is a DC holdback and how is it sized?
A DC holdback is the share of a buy kept in the distribution center at the initial allocation so that replenishment can send units where the first weeks of selling show demand, instead of guessing all of it up front. Its size is a judgement set by how long the selling window is, how fast door-level sales become readable, how long the DC-to-door lead time is, whether a chase from the vendor is possible inside the window, and how costly it is to move units between doors later. In the worked example the holdback is 15% of a 1,200-unit buy, which is 180 units, and because it is the residual by size after the door curves are applied, it carries only 8 units of XL against 54 of S.
How is a replenishment order calculated?
Under need-based replenishment, need = (target weeks of cover × trailing weekly rate) − on hand − on order, floored at zero, and it is calculated per size, not per door. In the worked example, door A1 sold 136 units in two weeks, a trailing rate of 68 a week, and holds 197. With a four-week target its door-level need is 4 × 68 = 272 − 197 = 75 units. Calculated by size, the same door needs 78 units: 7 S, 38 M, 19 L and 14 XL, with a 3-unit surplus in XS that the door total silently netted off. Door B1 needs 49. The 127 units requested are filled from the holdback at 115 units, a fill rate of 115 ÷ 127 = 90.6%, because the holdback ran out of M and XL.
How is allocation and replenishment performance measured?
Five measures, read together at the door-and-size grain. In-stock rate is the share of planned door-size positions holding sellable stock. Fill rate is units shipped divided by units requested, which in the worked example is 115 ÷ 127 = 90.6%. Weeks of supply is on hand divided by trailing weekly sales, read by door, where the example runs from 2.64 weeks at door B1 to 6.57 at door B3 at the end of week two. Sell-through by door is units sold divided by units received, from 43.1% at B1 to 23.3% at B3 over the same two weeks. Lost sales are the units a position would have sold while it was empty, estimated from its rate before the stockout, and they never appear in sales history.
Share this guide with your team
Copy a link or a pre-written message for Slack, Teams, or email.
// Know where your operation stands
Apply this to your planning operation.
The free Apparel Planning Maturity Assessment benchmarks your operation and tells you exactly which gaps to fix first.