Capital & Cash

Three clocks run between the day money leaves for your supplier and the day it is back in your bank.

1. Until it lands. The lead time: made, shipped, checked in and sellable. Nothing sells before then.
2. Until it sells. An order sells over the order ÷ units a day. The average unit sells halfway through.
3. Until you are paid. Amazon holds a sale's money until 7 days after the customer receives it, its "DD+7" reserve (Amazon's terms). Then it settles every 14 days unless you request a payout, and the transfer takes up to five business days to reach your bank (Amazon).

Hubricon's engine counts 10 days from a sale until Amazon can pay it (two to deliver, then DD+7), half a settlement cycle of waiting on average, and 4 days at the bank: 21 days from a sale to cash.

days a dollar is gone = days until it lands + half the days to sell the order + days until you are paid − days of supplier credit

On Shopify Payments the third clock is shorter: payouts typically arrive 3 to 5 business days after the payment (Shopify).

Supplier terms are days of credit

A supplier who takes a deposit at the order and the balance before shipment is lending you the balance until it is paid. Weight each payment by the share it is: 30% at the order and 70% at shipment 25 days later is 0.7 × 25 = 17.5 days of credit, and the dollar is gone that many days less.

Why it matters more than margin

The cycle sets how much cash the business needs to stand still. Every unit a day you sell ties up its landed cost for the whole cycle: 17.18 units × $6.20 × 148.5 days is $15,818 out at any moment. Grow 20% and that grows 20%, before the first extra payout arrives.

Fixed costs leave every day. Payouts arrive on their cycle. A wire leaves all at once. Put the three on one line and the bank balance goes up in steps and down in a slope, with a cliff on the day of the wire.

An invented listing · Not a client · Not a result

The trough is the number

Revenue tells you what came in over a month. Profit tells you what is left after it. Neither tells you the lowest the account goes on the way, and that is the number that bounces a supplier's wire or a card payment. A listing can be profitable every month and still hit zero on the day before a payout.

low point = balance on the wire's day − the wire − costs until the next payout

Growth deepens it

The wire scales with the order and the order scales with sales. A listing selling twice as much wires twice as much on the same day, while its payouts double only as the new stock sells. Growing faster than your cash cycle can carry is how a profitable seller runs out of money.

Three numbers decide the wire: when to order, how much, and so on which day and for how much. All three come from units a day and lead time.

When: the reorder point

Order when what you hold, on hand plus inbound, falls to what you could sell before the new order lands, in a bad case. Hubricon's engine reads that bad case as the 95th percentile of demand over the lead time. That covers two uncertainties: units a day vary, and lead times run late (the engine lets a lead time vary by about 20%).

reorder point = the 95th percentile of units sold over the lead time

How much: the order

order = units a day × (lead time + 30 days)

Enough to cover the lead time and a month after it lands. For the garlic press: 17.18 × (75 + 30) = 1,804 units, a wire of $11,185 at $6.20 each.

When, to the day

days of sales before the wire = (on hand + inbound − reorder point) ÷ units a day

With 2,500 on hand: (2,500 − 1,983) ÷ 17.18 = 30.1 days of sales above the reorder point, so the wire goes on day 31, and every 105 days after. Put the date in the calendar now: a wire you know is coming is cash you can plan.

A service level is the chance you do not run out before the next order lands. Hold more and it rises; it also ties up more cash and pays more storage. The right level is where one more unit's chance of a sale is worth exactly what it costs to hold.

service level = cost of one unit short ÷ (cost of one unit short + cost of one unit left over)

• Short: the margin you lose, plus Amazon's low-inventory-level fee on the units you do ship while cover is thin. For standard sizes in 2026 it runs from $0.32 to $1.11 a unit, by size and by how thin; the garlic press, a large standard unit, pays $0.97 under 14 days.
• Left over: storage for the cycle, the cash tied up (12% a year), and a small charge for the unit that never sells (2% of its cost a cycle).

On Shopify there is no low-inventory fee and no Amazon storage: a unit short costs its margin, and a unit left over costs the cash tied up, the write-off, and whatever your own shelf or 3PL charges.

What it changes

The order-up-to level is that percentile of demand over the lead time plus the week until the next review: 2,142 units for the garlic press against 2,134 at a flat 95%, and 314 for the spatula set against 379. The thick margin buys more stock and the thin one less, and the cash moves from the second to the first.

A unit on the shelf cost what it cost; that money is gone whatever you do next. The question is only which choice brings more cash from today: sell it at full price as it goes, paying storage until it does, or take Amazon's liquidation now.

hold = each month's units sold × cash per unit − storage and surcharges on what is left, discounted 1% a month

liquidate now = units × price × 10%

The tenth is Hubricon's engine's assumption for what Amazon's liquidation returns; what it really returns varies by product and is quoted in Seller Central before you accept. On Shopify there is no liquidation program: the same arithmetic holds with a clearance sale, or a liquidator's offer, in place of Amazon's tenth.

When liquidation wins

Storage and the aged surcharge are charged by the cubic foot; the cash a sale brings is by the unit. So the bet turns on size and speed. A small unit that sells, even slowly, almost always pays to hold. A bulky one that sells a few a month, past the 271-day cliff, can cost more a month to keep than it will ever bring: that is the one to liquidate, or to mark down, and The Fee Staircase's lesson 7 prices the cliff.

The trap

Do not count the landed cost on one side and not the other. It is sunk under both choices. A sheet that subtracts it from "hold" and not from "liquidate" will tell you to dump stock that would have netted several times more sold.

Sales wobble from day to day, so the path in lesson 2 is only the middle of many. Hubricon's engine runs the same 90 days 10,000 times, each with its own sales, and reads the spread.

An invented listing · Not a client · Not a result

The two numbers to read

• The 5th percentile: the bad case one month in twenty is worse than.
• The average of the worst twentieth: how bad those months are. A percentile says nothing about the tail beyond it; this does. It is the number a cash decision should rest on.

cash to hold = starting balance − the average of the worst twentieth of low points + the floor you want to keep

For the garlic press with a zero floor: $11,000 − $317 = $10,683 is what this cycle needs. Keep a floor for what the model cannot see, a late payout or a surprise invoice, and add it.

The reorder point already holds a safety stock against a slow lead time and a fast month. Wire late and you spend that safety stock before the order is even placed.

An invented listing · Not a client · Not a result

What the number leaves out

The lost margin is the part the arithmetic can see. A listing that runs out also loses sales rank, which it buys back with ads and time, and pays Amazon's low-inventory-level fee on what it ships while cover is thin. None of that is in the $343; all of it makes late worse.

The asymmetry

Early costs a little storage and a little interest, and you can see both coming. Late costs margin, rank and fees, and you see them only after. When the date is in doubt, wire early.