Inventory Forecasting for Small Clothing Brands

Why inventory forecasting for small clothing brands is a variant problem

A candle brand forecasts one SKU per product. A clothing brand forecasts a matrix. If you sell a tee in sizes XS through XL and four colors, that is twenty variants sharing one product page, one photo shoot and one demand pool — but twenty separate stockout risks. The core discipline of inventory forecasting for small clothing brands is deciding which level each decision lives at: demand is forecast per style, purchase orders are cut per variant, and stockouts are counted per variant.

The most common spreadsheet failure is forecasting each variant independently from its own thin sales history. A variant that sells two units a month produces a velocity number so noisy it is closer to a coin flip than a forecast. Pooling demand at the style level gives you a stable signal; the size curve (covered below) turns that signal back into variant-level order quantities.

Sales velocity: the 30-day and 90-day windows

Sales velocity is the engine of inventory forecasting for small clothing brands: units sold divided by days in the window. The window you pick changes the answer. A 30-day window reacts fast: it catches a style that just got tagged by a creator, and it catches a collapse just as quickly. A 90-day window is calmer: it smooths out one good weekend and one dead week into something closer to the underlying rate.

The spreadsheet-native answer is to compute both. Put 30-day velocity and 90-day velocity in adjacent columns for every style. When they roughly agree, either number is fine. When the 30-day number is well above the 90-day number, demand is accelerating — forecast with the faster number and buy sooner. When the 30-day number is well below, the style is cooling — forecast with the slower number, or lower, and let stock run down rather than reordering into a decline.

A worked example: a hoodie sold 84 units in the last 90 days (0.93 per day) and 42 units in the last 30 days (1.4 per day). The style is accelerating. Forecasting the next 60 days at 1.4 per day projects 84 units; at 0.93 it projects 56. If your supplier needs 45 days, the difference between those two forecasts is the difference between a full size run and a broken one in the middle of the acceleration.

Forecast the style, allocate by size curve

A size curve is the percentage split of a style’s sales across sizes. Compute it from your own history: total units sold per size for the style (or for a group of similar styles) divided by total units sold. If the last 200 units of your tees sold 18 XS, 44 S, 62 M, 50 L and 26 XL, the curve is 9% / 22% / 31% / 25% / 13%.

Now forecasting a variant becomes two clean steps. Step one: forecast style demand from pooled velocity — say 120 units over the next buying period. Step two: multiply by the curve — 11 XS, 26 S, 37 M, 30 L and 16 XL, rounded to your supplier’s pack sizes. This is how inventory forecasting for small clothing brands stays workable in a spreadsheet at 200-plus variants: one velocity forecast per style, one shared curve per category, and multiplication does the rest.

Recompute curves quarterly, not weekly. Size mix drifts slowly, and a curve rebuilt from every small sample will thrash. New styles with no history borrow the curve of the closest existing category until they earn their own.

Seasonality without statistical models

Seasonality is where inventory forecasting for small clothing brands most often goes wrong, because apparel demand is seasonal on two clocks at once: the weather clock (fleece in October, linen in May) and the retail clock (gifting spikes, back-to-school, sale events). You do not need exponential smoothing to respect either one — you need last year’s monthly sales laid out beside this year’s plan.

Build a simple index: each month’s share of the year’s units, computed from whatever history you have. If November carried 14% of annual units and February carried 5%, then a raw velocity measured in February understates what a November-facing buy needs by roughly a factor of 2.8. Multiply the base forecast by the ratio of the target month’s index to the current month’s index. It is crude, it is visible in a cell, and it beats pretending July velocity predicts December.

One season of history is thin but usable; zero seasons means you borrow. Use category-level intuition, ask your suppliers what comparable brands reorder ahead of, and buy conservatively with a planned reorder rather than one large speculative order.

Lead time buffers: forecasting is really a calendar problem

A forecast has no value until it is attached to a date. If your cut-and-sew supplier quotes 60 days and sea freight adds 30, the stock you will sell in November was ordered in July. The operative question is never “how much will this style sell” in the abstract — it is “what is the last calendar day I can place this purchase order and still receive it before the projected stockout.”

The spreadsheet version takes four inputs per style: current stock, daily velocity, total lead time in days, and a safety buffer in units. Runway equals stock divided by velocity; the latest order date is today plus runway minus lead time, pulled earlier by however many days of buffer you hold. ShelfOwl’s latest order date calculator runs this exact arithmetic for one SKU, and the reorder point calculator expresses the same logic as a trigger stock level instead of a date.

Apparel lead times also deserve their own buffer. Fabric availability, dye lots and factory queues move quoted dates, so track the lead time you actually experienced on the last three purchase orders — not the one on the quote — and size safety stock against the worst of them.

Returns: forecast what stays sold, not what ships

Apparel return rates tend to run higher than most other ecommerce categories, because fit is unknowable until the parcel is opened. Whatever your own rate is — measure it from refunded orders divided by orders over the last 90 days — it changes forecasting in two ways.

First, net demand is what matters for reordering. If a style ships 100 units a month and 15 come back in sellable condition, net depletion is 85, and reorders sized on 100 will slowly build a stock mountain. Second, returns arrive late: units sold in a November peak reappear in December. A December reorder that ignores the incoming return wave double-counts demand exactly when the season is ending. Keep a returns column per style in the spreadsheet, lag it by your average return transit time, and forecast on net units.

When the spreadsheet stops paying its way

A velocity sheet with size curves genuinely works at boutique scale, and this guide is written so it can. But the failure mode is predictable: every number above — stock, velocity, lead time, the latest order date — goes stale the moment you stop re-entering sales data. At 20 styles that upkeep is an hour a week. At 300 variants it quietly becomes the job you skip, and the week you skip it is the week a winner stocks out.

That maintenance burden, not spreadsheet math, is the real graduation trigger for inventory forecasting for small clothing brands. ShelfOwl’s paid forecaster is built as the smallest possible step up: you export one CSV of sales and stock levels — no integrations, no migration — and it returns the latest order date for every SKU at once, then emails you a reminder when any SKU approaches its order-by date. It is 29 dollars a month with a 14-day trial that needs no card, and the free single-SKU calculators stay free either way.