Inventory replenishment in apparel: methods, size ratios and timing.

What replenishment covers

Inventory replenishment is the decision system for restocking: when to reorder, how much, for which SKUs, into which locations. Done well it’s invisible - stock simply doesn’t run out, and neither does it pile up. Done badly it announces itself twice a year: as stockouts on the styles that were selling, and as markdown budget consumed by the styles that weren’t.

In an apparel business the word covers two distinct loops. Supplier replenishment: reordering from factories or distributors into the warehouse, on production lead times measured in months. Channel replenishment: allocating from the warehouse out to stores, concessions or channel inventories, on lead times measured in days. The math is related; the stakes and cadence are completely different, and conflating them is a category error that produces either starving stores or panic POs.

The four standard methods

Real operations mix them: min/max on proven core, periodic review aligned to supplier shipping cycles, demand-driven where the data supports it. The method matters less than the inputs being current - a perfect formula fed last month’s velocity is worse than a crude one fed yesterday’s.

What changes in apparel

Standard replenishment literature assumes a product that sells continuously and gets reordered indefinitely. Apparel breaks that assumption three ways.

Most styles are never replenished. Seasonal fashion is bought once, sold through, and exited - the plan is sell-through management, not restocking. Replenishment applies to the core/continuity slice of the range, and the share varies wildly by brand: a basics-led brand may replenish 70% of volume, a fashion-led one 15%.

The unit of replenishment is the SKU, not the style. One style is a matrix of sizes and colours, each with its own velocity. Style-level replenishment - reordering “300 units of the hoodie” without a size curve - restocks the sizes that weren’t selling along with the ones that were.

Lead times are long and asymmetric. Offshore production runs 60-120+ days; a replenishment decision today serves demand next quarter. The cost of being wrong is asymmetric by direction and by product: under-ordering core costs sales now, over-ordering seasonal costs markdown later - and the same formula shouldn’t treat those errors as equal.

Core vs seasonal: the first decision

Before any formula runs, every style needs a replenishment classification - and this is a planning decision, not a systems default:

• Core / continuity: carried across seasons, replenished on rules, protected against stockout. The list should be explicit and reviewed - core status is earned by velocity and margin, not by sentiment.
• Seasonal: bought to a sell-through plan, replenished only by exception (a genuine in-season chase when the factory calendar allows one), exited on schedule.
• Test / new: small initial buys with a decision gate - promote to core, extend the season, or exit - at a defined sales checkpoint.

The classic failure is core-by-drift: styles nobody decided to continue getting reordered because they’re familiar, while genuinely fast new styles miss their chase window because nobody was watching for the promotion signal. An explicit classification, revisited each season with the velocity data, is the cheapest fix in this entire article.

Size ratios and broken runs

The apparel-specific heart of replenishment. Sizes sell on a curve - the familiar bell across S/M/L/XL, shifted by market and category - and stock breaks from the middle outward: M and L sell out while S and XL remain. The style still shows inventory; the size run is broken; and because a customer who can’t find an M doesn’t buy an L instead, the style’s sell rate collapses to the velocity of its remaining fringe sizes.

Broken runs are expensive twice. The stranded sizes tie up capital and eventually eat markdown, and the missing sizes represent the highest-probability sales in the range - demand for the middle of the curve is the most reliable demand a brand has. A wall of S and XL is not 40% in stock; it’s functionally out of stock with carrying costs.

Ratio-aware replenishment is the discipline that addresses it: reorder quantities calculated per size to restore the target curve given current holdings - not the original buy ratio, and not a flat top-up. It requires size-level velocity data, size-level reorder logic, and honesty about when a run is too far gone to fix (late-season, the right answer to a broken run is often consolidation and exit, not a reorder that arrives after the season does).

Safety stock and lead times

Safety stock is the buffer between forecast and reality - against a demand spike, a late vessel, a factory slip. The textbook sizes it from demand variability, lead-time variability and a target service level. The apparel-practical version adds two adjustments.

First, set it per tier and per size. Core styles in core sizes justify generous buffers - the M and L of a permanent hoodie should effectively never stock out, and the carrying cost of two extra weeks is trivial against the lost-sale cost. Fringe sizes and anything seasonal justify thin buffers, because their excess exits through markdown.

Second, treat lead time as a managed input, not a constant. Replenishment math silently assumes the lead time it’s fed; production slips break it silently. This is where replenishment meets the critical path - a two-week factory delay on a core restock is a replenishment event, and systems that connect production tracking to inventory projections surface it while there’s still time to air-freight the size curve’s middle, rather than discovering it as a stockout.

How much to automate

The calculation layer should absolutely be automated: velocity by SKU, size-curve state, projected stockout dates, suggested order quantities - this is continuous math no spreadsheet review cadence can match. The commitment layer deserves more care, because apparel’s lead times make a bad automatic order expensive for months.

The pattern that works: system proposes, planner approves - replenishment proposals generated on rules, reviewed in minutes because the math is visible, committed with judgment applied to what the formula can’t see (the retailer conversation, the fabric problem, the marketing push next month). Full automation then earns its way in for proven core styles with stable demand and reliable suppliers - the segment where human review adds latency and no insight. Automating in that order builds trust in the numbers before the numbers act alone.

The signals worth watching

• Weeks of cover by SKU - stock divided by current velocity, the basic health number, watched per size
• Size-run integrity - share of core styles with full runs in core sizes; the earliest broken-run warning
• Projected stockout date vs replenishment arrival date - the two dates whose gap is the whole game
• Sell-through vs plan for seasonal styles - feeding the chase/exit decision at the checkpoint
• Forecast bias - systematic over- or under-forecasting by category, the correction that improves every downstream number
• Inventory turnover by tier - confirming core earns its shelf space and seasonal exits on time

Every one of these is computable from data the business already generates - sales, stock, POs, lead times. Whether they get computed continuously by the platform or quarterly by a heroic spreadsheet is, in practice, the difference between replenishment as a system and replenishment as an argument.

Frequently asked questions