Reorder Points When Your Lead Time Is 90 Days

Almost every reorder-point article assumes a domestic warehouse restocked in a week. This one is for a brand whose next receipt is on a vessel. It sits under demand forecasting for apparel brands and covers what the pillar stops short of: turning a forecast into an order date and a quantity your mill will actually run.

What a reorder point is, and the formula that produces it

A reorder point is a stock level, not a date: when on-hand inventory falls to it, you order.

Reorder point (ROP) = (average demand per period x lead time in the same periods) + safety stock

Two things go wrong in an import context, both inputs rather than arithmetic. The lead time is measured wrong — brands use the factory's PO-to-FOB quote, which excludes ocean transit, customs entry, drayage and DC receiving, commonly 40 to 50 days. And it is treated as a constant. A 33-day transit arriving on day 45 has eaten weeks of cover you never budgeted.

The safety stock formula long lead time importers need, and the Z-score most tables get wrong

Safety stock for imported apparel absorbs demand you did not forecast and time you did not plan. Two formulas.

Demand variability only:

SS = Z x sigma-d x square root of LT

Combined demand and lead-time variability (King's formula):

SS = Z x square root of ( LT x sigma-d squared + D-avg squared x sigma-LT squared )

Z is the service-level factor, sigma-d the standard deviation of demand per period, LT average lead time, D-avg average demand per period, sigma-LT the standard deviation of lead time. The first formula assumes a metronomic supplier: fine for a domestic 3PL, indefensible for an ocean lane.

The same calculation travels under several names, which is worth knowing before you compare two planning tools that appear to do different things.

One-tailed Z-scores for safety stock. Use the middle column.

Safety stock is one-tailed: you care only about demand exceeding supply, not symmetric deviation around a mean. Several widely copied tables, including SCMDojo's otherwise useful reference, publish two-tailed confidence-interval values. Using 1.96 where you meant 1.645 inflates safety stock by about 19% — on a 1,500-unit buffer, 285 units nobody decided to buy.

Decomposing a 90-day apparel import lead time into its real components

"Ninety days" is a rounding, not a lead time. Below is a repeat knit programme built stage by stage from Tirupur to a New Jersey DC, in two versions: fabric already dyed and held at the mill, and a custom dye lot run to order.

Repeat cotton knit order, India to US East Coast, calendar days. Production, sampling and transit stages come from published apparel lead-time and transit ranges; the last four rows are planning allowances to replace with your own receipt history.

Three rows carry the variance. Custom dyeing is the biggest controllable block: fabric and trim sourcing runs one to four weeks and custom dye adds four to eight on top, so holding a dyed platform for repeating colours removes about 40 days in one decision. Ocean transit is a range: India to the US East Coast is 30-45 days via Suez, India to the West Coast 25-35 and usually transshipped, with Red Sea routing unstable through 2026 — a 15-day spread is roughly two weeks of sigma-LT on its own. Customs is fast until it is not: cotton is one of U.S. Customs and Border Protection's named high-priority forced-labour sectors, and a detention belongs in sigma-LT, not the mean.

Do not estimate sigma-LT, compute it: two years of receipts, days from PO issue to sellable-in-DC, standard deviation, and do not exclude the bad ones. The Chinese New Year shipment three weeks late, the Eid shutdown, the container rolled in August peak are not outliers. They are your distribution.

The 85-day stage build and the 13-week planning input are two different numbers

The worked reorder point later on this page uses 13 weeks — 91 days — as the average lead time, not the 85 days this table sums to, and that gap is deliberate. The stage build is the sum of nominal stage durations: what each step takes when nothing slips. The 13 weeks is the mean of an actual receipt history for the same programme, and it carries the slippage the build cannot see — the port cut-off missed by a day, the container rolled, the re-work cycle after a failed inspection. Six days between them is a typical, not an alarming, gap.

Always plan against observed history, not the stage build. The build is a diagnostic: it tells you where the time sits and which stage is worth attacking. Only your own PO-to-sellable receipt dates tell you what belongs in the formula, and they are the only source that also gives you sigma-LT.

Why lead-time variability hurts more than demand variability when lead times are long

Look at the two terms inside King's formula. The demand term, LT x sigma-d squared, grows linearly with lead time. The lead-time term, D-avg squared x sigma-LT squared, is scaled by average demand squared. Once a programme has real volume, the second term takes over.

The same programme under two supply models. Demand 400 units a week, sigma-d 120; service level 95%, Z = 1.645.

At 13 weeks, three quarters of the buffer covers uncertainty about when goods arrive, not how many you sell. Improve one input at a time on the import case:

• Halve demand sigma, 120 to 60 (genuinely hard): SS = 1.645 x root(46,800 + 640,000) = 1.645 x 828.7 = 1,363 units. Saving: 133 units.
• Halve lead-time sigma, 2 weeks to 1 (a fabric platform and a booked sailing): SS = 1.645 x root(187,200 + 160,000) = 1.645 x 589.2 = 969 units. Saving: 527 units.

Reducing lead-time variability is worth roughly four times as much as improving the forecast, and it is usually cheaper. That is the case for platforming fabric, a nearshore replenishment tail, or air-freight optionality: each cuts sigma-LT, not sigma-d.

A fully worked reorder point for a core tee programme

Core cotton crew tee, black, one of a six-colour core book, made in Tirupur and shipped to a New Jersey DC. Landed cost $6.08 a unit at non-peak freight, built line by line in landed cost for apparel imports.

The lead-time input is the 91-day receipt-history mean, not the 85-day stage build in the table above. Use the history.

Step 1 — demand-only safety stock, for comparison. SS = 1.645 x 120 x root(13) = 712 units

Step 2 — King's formula, the one to use. LT x sigma-d squared = 187,200; D-avg squared x sigma-LT squared = 640,000. SS = 1.645 x root(827,200) = 1.645 x 909.5 = 1,496 units. That King's formula safety stock number is 784 units above the demand-only answer, and the gap is entirely the lead-time term.

Step 3 — the reorder point. ROP = (400 x 13) + 1,496 = 6,696, call it 6,700 units

Step 4 — read it in weeks. 6,696 ÷ 400 = 16.7 weeks of forward cover. You reorder while still holding four months of stock. A planner on domestic instinct sees 6,700 tees on hand and concludes there is nothing to do, which is how brands run out of their best colour in February. The wrong formula in step 1 would leave you 784 units short at the precise moment a vessel is late.

Your reorder quantity is bounded below by fabric and dye-lot minimums

Here is the constraint no reorder-point article models: the quantity the maths returns may not be one your mill will run.

A dye vessel has a minimum load. Below it the liquor ratio is wrong and shade reproducibility fails, so the mill declines at any price. Custom-colour minimums are typically 300-500 kg of fabric, and tier-1 Vietnamese mills commonly quote 800-1,200 kg per colour; piece-dye minimums run 500-1,000 metres. These come from set-up-cost amortisation and physics, not factory size: how fabric and dye-lot minimums set apparel MOQs.

Sage colourway of the same tee: demand 55 units a week, target coverage 12 weeks, fabric consumption 0.27 kg per unit from the standard single-jersey formula at 180 gsm with 10% wastage, dye-lot minimum 400 kg.

The plan asked for 660 units; the smallest order the mill accepts is 1,482. The reorder point was not wrong — it produced an answer that does not exist as a purchase order.

What to do when the reorder point returns a quantity the mill will not run

1. Aggregate the colour across every style sharing the lot. The dye vessel does not care which style the fabric becomes, so meet the minimum at programme level, not style level.

(Consumption other than the tee is illustrative; take yours from the factory's sheet.)

2. Round up to the lot minimum and carry the excess as planned coverage. Buy 1,482 and book the extra 822 units as coverage rather than a rounding error, priced: at $6.08 landed that is $4,998 of working capital held roughly 15 weeks longer than planned, plus markdown risk if the colour softens. Usually right for a stable core colour, wrong for a fashion colour, and the difference is whether the demand repeats.

3. Change the colour strategy for the tail. Stock and greige-dyed standard colours carry no dye-lot minimum, and garment dye moves it to 50-200 pieces per colour. What does not work is ordering below the minimum and hoping: below-MOQ pricing carries a 15-30% unit premium and stretches lead time to 14-20 weeks PO to FOB against a normal 8-12 weeks PO to FOB on a repeat order.

Three checks before the purchase order goes out

Confirm the receipt month has budget (open-to-buy planning), the size ratio is derived rather than inherited (size curves and the SKU explosion), and the inspection plan is set, because a failed final inspection adds a re-work cycle to a committed lead time (AQL inspection for brand owners). Terms above are in the Yarnstick glossary.

Frequently asked questions