Here's the verdict up front: FBA replenishment comes down to three terms — average daily sales × total lead time in days + safety stock. In all my years doing this, every replenishment blowup I've seen got the first two terms right and crashed on the third.
Start by getting average daily sales right. A lot of sellers just take 30-day sales from Seller Central and divide by 30. That's the first landmine. Promo spikes have to come out: a big-event week can move several times your normal volume, and once it's in the denominator, your "average" comes out 20–30% too high — which means a month of extra inventory sitting in an Amazon warehouse. My method: take the last 60 days, throw out promo weeks and stockout days, then average what's left. New product with no history? Borrow the first-month ramp of a comparable ASIN in your category, estimate conservatively, and plan on reordering sooner rather than stuffing the first shipment. A pet-supplies seller I worked with in 2021 did the opposite — he sized his second shipment off his launch-week spike, back when his ads were still in the honeymoon phase. Sixty days of cover turned into a hundred and ten. The listing survived; his cash flow didn't enjoy those two months.
Then total lead time. Note the word "total": it's not "how many days on the water." It's factory production + first-leg freight + Amazon receiving and check-in, added together. The third leg is the one everyone underestimates. In Q4 2023, a home-storage seller walked me through his post-mortem: freight left his 3PL in Ontario — a 90,000-square-foot building — sailed 25 days without a hitch, then sat in Amazon's receiving queue for nearly three weeks during peak. He missed the first holiday traffic wave entirely. Since then, "receiving/check-in" gets its own line in every replenishment sheet I build: pad two to three weeks in peak season, one to two in normal times. And calibrate it against your own account's receiving history — check-in speed varies wildly by fulfillment center and category, so don't copy someone else's number.
Lead Time Is a Sum, Not a Single Number
Most sellers remember "25 days on the water" and forget everything around it. Factory production slots take time. Booking and waiting for a vessel takes time. Customs clearance — and if your container gets pulled for exam, that time belongs in somebody's plan, not nobody's. I break the first leg into five pieces: factory production, booking and vessel wait, ocean transit, customs and drayage, and final delivery into the Amazon building. Each gets its own days, and each gets a peak-season buffer. Exams are luck, but your model needs a line for luck — one exam costs far more than a day or two, and it's cheaper to carry the buffer up front than to explain the stockout later.
One more detail: LCL and FCL don't share a clock. LCL waits to consolidate and gets devanned on arrival; a full container pulls straight through. Run the math on the channel you actually use, not the faster one you wish you used.
Safety Stock Is Where the Math Actually Breaks
Safety stock is insurance against sales volatility, so the coefficient should be set by volatility — not by gut feel. The experience-based values I start from (illustrative — tune them to your own data): a steady year-round staple does fine at 1.2; a seasonal category runs 1.2 off-peak and 1.5 or higher into peak. The classic way to die is one coefficient all year: overstocked in the slow months, stocked out in the hot ones, losing on both ends.
The second trap is mixing in-transit and on-hand into one number. The formula tells you how much to ship this round. How many days of cover are left in the building, and the latest date you can place the next order — that's a separate sheet. One is replenishment quantity, the other is the reorder point, and they shouldn't be mashed together. The reorder point itself is simple: average daily sales × total lead time. When sellable inventory at Amazon drops to that many days of cover, the next factory order is already late. Safety stock sits on top as a cushion, not blended into the total.
Then the trade-off: when do you push the coefficient up? Compare the two costs. Q4 storage rates run higher than the rest of the year (check Amazon's current fee schedule — it changes), so extra units cost real money. But a stockout day costs rank, ad weight, and organic momentum that money can't buy back. My rule is blunt: if the ASIN is climbing and ads are scaling, bump the coefficient a notch — the cost of stocking out dwarfs two extra weeks of storage fees. If it's a long-tail veteran with stable organic traffic, you can afford to be conservative. Here's a gut-check I run with sellers (illustrative): if a thousand extra units cost you one month of peak storage, but one stockout week costs you the number-three slot on your main keyword — a slot you spent three months climbing into — the storage fee is the cheaper insurance. Run it with your own numbers; the answer is usually not close.
Peak and everyday are two different games. Everyday replenishment optimizes for turns — lean, frequent, lower coefficients. Peak replenishment optimizes for never going dark — better to clear it on discount in January than to stock out in December. Mix the two logics and you get the worst of both: piles of inventory in the slow season, stockouts in the hot one.
Worked example (all numbers illustrative — plug in your own): say a kitchen storage bin sells 40 units a day after stripping out the July promo spike. Total lead time: 10 days production + 5 booking and vessel wait + 28 LCL ocean + 4 customs and drayage + 12 Amazon receiving = 63 days. Base replenishment = 40 × 63 = 2,520 units. Safety stock: the category swings hard October through December, so take a 1.4 coefficient — an extra 40 × 63 × 0.4, about 1,000 units. Ship roughly 3,500 this round. Notice the formula only asks you to make two judgment calls: spike-adjusted daily sales, and the safety-stock coefficient. Everything else is addition.
One question to close: when was the last time you recalculated your safety-stock coefficient against your category's peak-season volatility? If the answer is "never," this article earned its keep.
