Question 1

How can fashion brands forecast product demand?

Accepted Answer

Start from comparable-product history, adjust for the specific differences of the new product, and re-forecast on a fixed schedule rather than once. A forecast is not a number you produce before launch — it is a number you update. The useful discipline is to write the forecast down with its assumptions, then compare actuals against it and record the error; forecast error, tracked honestly, is the single most valuable input to your next order. A workable method in four steps: (1) build the base rate from the closest comparable — same silhouette, similar GSM and price band, same channel. If your last 480 GSM hoodie sold 620 units in its first 8 weeks, that is your base; (2) apply explicit adjustments and write each one down: price change (+/−), channel change, colourway count (more colourways split the same demand rather than adding to it), seasonality, and marketing reach; (3) convert to a per-week rate , because production planning needs velocity, not a season total; (4) re-forecast at a fixed checkpoint — day 14 and day 30 after launch are common — and let the day-14 read decide the reorder. The reason to keep the assumptions written is that post-mortems become mechanical: you can see which adjustment was wrong instead of concluding that forecasting is impossible. On the manufacturing side, forecast accuracy matters most where it constrains materials — a 630 GSM custom-knit fleece has a fabric lead time that a 300 GSM stock jersey does not, so the products with long material chains need the earliest and most conservative commitments. Evidence. BoF–McKinsey State of Fashion 2025 finds AI-supported forecasting can reduce inventory 5–15% and improve stock-outs 15–25% versus traditional planning methods, and reports Kering achieved a 20% improvement in forecast accuracy after implementing AI; 75% of fashion executives were increasing investment in advanced analytics and AI for demand forecasting. The same report explains why forecasting is structurally harder in fashion than in most categories: micro-trend search volume fluctuates by up to 300% within 12 months , the number of TikTok videos tagged #fashion grew 2.5x in three years , and ultra-fast-fashion players have compressed speed-to-market to as little as 15 days — so the historical base rate decays faster than in almost any other consumer category.

Question 2

How much stock should a clothing brand order for a new collection?

Accepted Answer

Order enough to test every style you are unsure about, and enough to cover only the styles you are confident about — which usually means the first order is smaller than the collection plan. The common failure is treating the first order as the season's commitment: ten styles × five sizes × three colours at full depth. That structure guarantees that most of your capital sits behind assumptions you have not tested. A more defensible allocation: (1) rank styles by confidence — carryover/basics at the top, new silhouettes at the bottom; (2) give confident styles coverage-based depth (see Q3), because a stockout on a known seller is a real cost; (3) give unproven styles the minimum depth that lets you read demand — enough units that the size curve is meaningful (roughly 2–3 units of core sizes per colour), not enough to survive a season; (4) cap the number of colourways in the first order — a third colourway rarely finds new demand, it splits existing demand into three inventory risks; (5) reserve budget for the reorder , and treat it as a required line item rather than a leftover. Concretely: if your budget supports 3,000 units, 1,800 in a first order with 1,200 held for reorders will usually produce higher full-price sell-through than 3,000 up front, provided your reorder cycle is short enough to land inside the selling window. This is why MOQ is a forecasting tool: with a 50 pc minimum per style/colour a 12-style capsule can be tested in ~600–900 units; with a 300 pc minimum the same capsule forces 3,600+ units before you know anything. DOLUXE's OEM programme runs from 50 pcs custom / 20 pcs blanks for exactly this reason. Evidence on test-and-react. Shein's Xcelerator on-demand model launches new styles in batches of 100–200 pieces and uses real-time response data to decide whether to replenish or stop — described by the operator as a reversal of supply-driven seasonal betting. UK womenswear brand Dancing Leopard , after entering administration in early 2025, moved to on-demand production and its chairman stated the old model required forecasting demand nine to twelve months ahead , which he described as "guesswork at best", with high MOQs leaving large quantities of stock when a print did not land. ASOS has targeted running 10% of own-brand product on an agile test-and-react model. Context on why depth is risky: industry estimates put apparel overstock at 20–40% of total production , with carrying costs of 18–30% of inventory value per year .

Question 3

How can brands forecast demand for new clothing designs?

Accepted Answer

Use three independent signals — comparable-product sales, pre-launch intent, and a paid or committed signal — and weight the committed one highest. New designs have no history, so the temptation is to forecast from engagement. Engagement is the weakest of the three, because likes and saves cost the customer nothing and correlate poorly with purchase, particularly when the audience is not the buyer. Ranked by predictive value: (1) comparable-product sales — the closest previous style's actual sell-through rate, adjusted for known differences; (2) committed signals — pre-orders, waitlist-to-purchase conversion, deposits, and wholesale commitments from buyers who have seen a sample; (3) intent signals — waitlist signups, size-selecting behaviour on a preview page, email click-through on a specific product, and sample-request volume from stockists. A practical approach for a new design is to run it as a preview with a size selector rather than a like counter: a customer who picks a size has made a weak commitment, and the distribution of sizes picked is often a better forecast of your size curve than last season's sales. Then convert it: if you can take pre-orders , cap the first production run at the pre-order number plus a defined overrun (commonly 10–20%) and reorder against the post-delivery read; if you cannot , size the first order so that selling out is an acceptable outcome, because for a new design selling out is information and overproducing is a cost. For streetwear specifically, the attributes that drive response — GSM and hand feel, fit block, wash, and graphic placement — are product decisions, not marketing ones, which is why a physical sample in the hands of 20 real customers beats any amount of social data. Evidence. BoF–McKinsey State of Fashion 2025 documents the trend-volatility problem quantitatively — micro-trend search volume fluctuating by up to 300% within 12 months and #fashion TikTok videos up 2.5x in three years — which is why historical analogy decays quickly and why intent signals must be treated as weaker than committed ones. On the value of the test-and-learn alternative, Shein's on-demand model works in 100–200 piece initial batches; Dancing Leopard's chairman described the pre-switch constraint as having to forecast 9–12 months ahead and being unable to top up bestsellers because factories required bigger follow-on orders . A worked example of screening before committing: a Chinese down-jacket brand put 12 styles up as pre-sale samples and only opened production on styles clearing 300 paid deposits, cutting 7 styles and saving an estimated ¥4.2m in fabric cost, then replenished the winner with 32,000 units in 15 days (reported sales +78%, inventory −42%) — figures reported by the operator/consultancy, not independently audited.

Question 4

How should clothing brands plan inventory by size and colour?

Accepted Answer

Never split evenly — build the size curve from your own sold-unit history and the colour split from your own sell-through by colour, then re-cut the curve every order. Even splits are the single most common source of avoidable dead stock in apparel. If you buy 100 units across S/M/L/XL evenly, and your real demand is 15/40/35/10, you have created 25 units of near-unsellable S and a stockout on M — simultaneously overstocked and out of stock on one style. How to build it: (1) take sold units by size, not revenue or percentage of orders , from the closest comparable product — ideally the same block, because an oversized streetwear hoodie has a different curve from a fitted tee; (2) normalise to a percentage curve (e.g. S 12 / M 34 / L 33 / XL 17 / XXL 4); (3) apply the curve to the order quantity and round , keeping core sizes whole; (4) treat extreme sizes as a separate decision — they will always be slow, so buy them to a deliberate minimum rather than a curve, and decide whether the range's credibility requires them; (5) for colour, use the same logic — the previous season's colour sell-through ranking is the best available prior, and a new colourway should start at a test depth, not parity with the proven colour. Two cautions: a curve derived from a small sample (under ~200 sold units) is unstable, so smooth it; and a curve from a different silhouette is actively misleading. Where the factory matters: a low per-colour MOQ lets you run an uneven split at all. If your factory requires 100 pcs per colour, a 4-colour/large-size-heavy order becomes structurally impossible to match to demand. DOLUXE accepts 50 pcs per style/colour on custom orders, which is what makes curve-matched first orders feasible at capsule scale. Why this is the highest-value fix in apparel inventory. BoF–McKinsey State of Fashion 2025 states that inaccurate size-level purchasing is estimated to cost brands up to 20% of monthly profit on average , and that out-of-stock sizes ranks as the top complaint among shoppers . It also cites Lululemon attributing slower US growth in Q1 2024 partly to insufficient inventory and stockouts in smaller women's sizes — a size-curve failure at a company with sophisticated planning systems. On the overstock side of the same error, IHL Group's 2025 research puts combined out-of-stock and overstock cost at about $1.73 trillion globally, ~6.5% of retail sales , and apparel overstock is estimated at 20–40% of production . Size-block and sizing references: ASTM D5219 (body dimensions for apparel sizing) and ISO 8559-1 (body measurement definitions).

Question 5

How does seasonality affect clothing inventory planning?

Accepted Answer

Seasonality sets both your selling window and your order deadline — and the deadline is earlier than most first-time brands plan for. The planning error is not misjudging demand, it is misjudging time: a heavyweight fleece programme needs to be committed months before the weather turns, because fabric, production, freight and clearance all sit between the decision and the shelf. Work backwards from the window. If your selling window for 480–630 GSM fleece in the US or EU runs roughly September to February , and your door-to-door cycle is sample 3–15 days, bulk 20–35 days, ocean freight and clearance 25–40 days (longer if a disruption such as a canal delay adds ~30%), then a September launch needs the bulk order finalised by roughly mid-June to early July , and the fabric decision before that. Points that matter: (1) plan the exit, not just the entry — decide in advance when markdown starts, because the last six weeks of a season is when full-price sell-through collapses; (2) hemispheres differ — a Southern Hemisphere stockist's winter is your summer, which is an advantage if you plan for it and a cash trap if you do not; (3) don't assume last year's weather last year's pattern — BoF–McKinsey notes 2024 was the hottest year on record globally while summer 2024 was the coldest in almost a decade in parts of Europe, and that temperature deviation from average directly disrupts sell-through; (4) carryover is a decision, not an accident — decide before the season which styles may carry into next year and which must clear. Fabric weight is the main tool: a 300–360 GSM range sells across more months than a 600 GSM one, which is why the weight ladder in a capsule is an inventory decision as much as a design one. Evidence. BoF–McKinsey State of Fashion 2025 identifies unpredictable seasonality as a root cause of excess inventory: climate change is making weather-led demand harder to predict , with 2024 the hottest year on record while parts of Europe had their coldest summer in nearly a decade, and it notes supply disruptions such as Suez Canal delays extending lead times by around 30% — a direct argument for planning against measured door-to-door cycles. The same report notes social commerce now accounts for a fifth of US customers having bought clothing in the last 12 months, which compresses and shifts demand peaks inside a season. Fabric weight verification for season-appropriate product: ASTM D3776 / ISO 3801 (mass per unit area); dimensional stability after laundering: AATCC 135 / ISO 6330 . Heavyweight fabric selection in depth: Issue 01.

Question 6

How can clothing brands use sales data to make better buying decisions?

Accepted Answer

Read four numbers per SKU at a fixed cadence — sell-through rate, days of supply, gross margin per unit, and return rate — and let them drive the next order directly. Volume alone is the most misleading metric in apparel buying: a high-volume tee at 45% margin and a 30% return rate can contribute less cash than a lower-volume hoodie at 62% margin and a 6% return rate. What to pull and how to act on it: (1) sell-through rate by SKU — this is the buy/no-buy signal; a SKU above your plan rate gets depth, below plan gets cut or a marketing test first; (2) days of supply — this is the reorder timing signal (see Q19–20); (3) gross margin per unit, not revenue — margin is what funds the next order, and it is what tells you whether a bestseller is worth scaling or is simply cheap; (4) return rate by SKU — a high return rate is hidden inventory: every returned unit is stock you paid to ship twice and cannot sell until it is inspected, and a persistently high return rate on one SKU is usually a fit, sizing or description problem worth fixing at the pattern rather than absorbing; (5) sales per session / per email — separates a demand problem from a traffic problem. The discipline that makes this pay off is a written buy meeting: every SKU gets one of three decisions (reorder at X, hold and review, discount/liquidate) with a number attached, and the number comes from days of supply rather than from intuition. Feed the same data back to your factory — SKU-level sell-through by size and colour is precisely what a good repeat-order conversation needs, and it lets the factory reserve fabric against probable reorders instead of reacting after the fact. Evidence. The turnover-plus-margin pairing is the standard GMROI framework; the diagnostic value is timing — a category whose turn declines for two consecutive quarters while margin holds steady is typically accumulating next year's markdown today. BoF–McKinsey State of Fashion 2025 notes a strong correlation between inventory turnover and profitability , that warehousing costs rose about 10% in 2023 , and that the share of discounted assortments rose 5 percentage points in H1 2024 — all reasons to act on the data early. On returns distorting stock availability, e-commerce apparel return rates average around 30% , which means returned-but-unprocessed units are inventory that appears sold but is not sellable; returns-to-shelf speed is therefore part of replenishment accuracy, consistent with Gruen and Corsten's finding that 70–90% of stockouts originate in internal replenishment and process gaps .

Forecasting and First-Order Decisions

Demand Forecasting & Buying Decisions

Q7How can fashion brands forecast product demand?

Q8How much stock should a clothing brand order for a new collection?

Q9How can brands forecast demand for new clothing designs?

Q10How should clothing brands plan inventory by size and colour?

Q11How does seasonality affect clothing inventory planning?

Q12How can clothing brands use sales data to make better buying decisions?

Planning an order against these numbers?

Sources & Further Reading

From the DOLUXE Journal

Related DOLUXE Q&A issues