A purchase order is two objects

Most sellers keep supplier management in two places that never meet: a margin spreadsheet, and a folder of invoices with dates on them.

That split feels natural. Profitability is an analysis question; paying the factory is an admin task. They sit with different tools and often different people.

It's also the reason profitable Amazon businesses run out of money.

Because a purchase order isn't one thing. It's an inventory commitment and a cash schedule wearing the same document. The inventory half determines what you'll earn. The cash half determines whether you can get there. Look at either alone and you'll make a decision that's correct in one dimension and ruinous in the other.

Here's how to hold both at once — and a number that settles the argument sellers have most often, which is which supplier is cheaper.

01Two objects, one document

Split every PO deliberately, because the two halves have different owners, different questions, and different failure modes.

• The inventory commitment — units, landed cost, when they become sellable, what they'll contribute. Feeds product economics and reorder planning.
• The cash schedule — a set of dated obligations, each with an amount and a trigger. Feeds the cash timeline.

They share exactly two facts: the unit cost and the arrival date. Everything else diverges. And those two shared facts are precisely where systems drift apart, so they're worth guarding — §08 comes back to it.

The PO you can afford and the PO you should place are different documents until you put them on one screen.

02The quote is not the cost

A supplier quotes $8.00. That number belongs in a negotiation, not in a margin calculation.

Landed cost is everything required to turn a quote into a sellable unit: manufacturing, packaging, inspection, labelling, prep, inland transport, international freight, duty and tariffs, insurance, and whatever the last shipment taught you to expect.

The practical consequence shows up when comparing suppliers. A $7.50 quote against an $8.00 quote is not a saving until you've added:

• MOQ — a lower price at 5,000 units when you need 2,000 is a worse deal, not a better one.
• Incoterms — EXW and DDP quotes are not comparable numbers. One of them contains the freight.
• Packaging and prep — who does it, and is it in the price?
• Defect rate — 3% unsellable is 3% on your effective unit cost.
• Payment shape — the subject of the next two sections.

Compare the economic cost of the inventory, not the number on the quote.

03Payment shape, and why it doesn't show up in margin

Two suppliers, identical landed cost, identical product.

• Supplier A: 30% at order, 70% before shipment.
• Supplier B: 20% at order, 80% thirty days after shipment.

Your product margin is byte-for-byte identical. Your business is not.

The reason is the shape of the gap between paying and being paid. An Amazon seller pays a factory, waits for production, waits for a boat, waits for check-in, sells over weeks, then waits again for Amazon's disbursement cycle. Cash leaves early and returns late, and the interval is measured in months.

Two numbers describe that gap. Almost nobody computes either:

cash conversion gap = days from the first supplier payment
                      to the first Amazon disbursement
                      from that inventory

peak exposure       = the largest amount you are out of pocket
                      at any point in the cycle

Margin tells you what a unit earns. These two tell you what it costs to wait for that, and how much of your business is tied up while you do.

04So which supplier is actually cheaper?

Now make the two suppliers differ on price as well, the way they do in reality — the cheaper one is slower and wants its money sooner.

A is 5.6% cheaper per unit. Every instinct says take it. Watch what the cash does.

Turn the cycle length into how many times a year the same money works:

The cheaper supplier earns $3,200 less per year on the same capital — and that's before counting what its 20 extra days of lead time cost in safety stock and stockout risk.

Add a second product competing for the same cash and the gap widens, because the constraint isn't the price. It's how long each dollar is stuck.

05Turning a PO into dated payments

A $21,000 purchase order is not a $21,000 cash event. It's a set of obligations, each with an amount and a rule for when it lands. Four triggers cover essentially every real contract:

• At order — dated on the PO date. Known immediately.
• At shipment — dated when the goods leave. Estimated until they do.
• N days after shipment — the same anchor, plus a fixed offset.
• On a fixed date — dated as agreed.

Two of those four hang off an event that hasn't happened yet, which produces the trap:

This is the difference between a schedule that ages well and one that quietly rots. It's the same discipline as the rest of the stack: keep the inputs, derive the outputs, and label what's derived.

06Committed, scheduled, due

"How much do I owe suppliers?" has three legitimate answers, and mixing them is how a comfortable-looking month turns into a scramble.

Three products in production at once — $10,000, $15,000, $20,000 — is $45,000 committed. At 30% deposits it's $13,500 due now, which reads as comfortable. The remaining $31,500 arrives weeks later, and it arrives alongside the ad spend, the freight invoices, the software and the existing balances that were already scheduled.

Nothing went wrong. Three reasonable decisions simply landed in the same fortnight.

07The danger is overlap, not any single order

One PO almost never causes a cash problem. The problem is the calendar view nobody builds.

Per-supplier questions — when do I pay Supplier A? — are the wrong grain. The business doesn't experience suppliers one at a time; it experiences a week in which $18,400 leaves.

So the useful view is a forward window across everything:

Next 7 days     $18,400 due          2 instalments
Next 30 days    $47,200 scheduled    5 instalments + freight
Next 60 days    $91,500 committed    incl. 2 POs with no schedule yet

And the one exception check worth running automatically: is any payment due before the inventory it's paying for turns into Amazon cash? That's not a rule of thumb — it's a comparison of two dates you already have, and it's the specific shape of every inventory-driven cash squeeze.

A supplier dashboard that lists fifty POs has told you nothing. One that says "two payments land before the cash they depend on" has done the whole job.

08One cost, one owner

Back to the two facts the halves of a PO share. Here's how they come apart.

A supplier raises the price from $7.20 to $7.80. The new PO records $7.80 correctly. The margin model still holds $7.20, because it was populated separately — or because it recalculates from a cached figure that nothing told to refresh.

Nobody made a mistake. There was simply no path from the purchase order to the profitability number, and the business now looks 8% more profitable than it is on every screen that matters.

The rule is structural, not procedural:

• The PO is the source of truth for unit cost. Margin reads from it; it never holds its own copy.
• A cost change recomputes what depends on it, automatically. If recomputation is a button someone has to remember, every number downstream is stale by default.
• Estimates are labelled as estimates. Freight projected, duty assumed, a quote that expires in 30 days — mark all three, so a margin built on them declares itself.
• A real invoice replaces its estimate, rather than being added beside it. Otherwise freight gets counted twice, once as a guess and once as a bill.

Chain it: supplier → PO → cost → product → margin. Every link is a read, not a copy.

09The supplier scorecard you already have the data for

Most supplier records are contact cards — a name, an email, a WhatsApp number. That's an address book, not a supply chain.

The useful record is a history, and almost every field in it is a by-product of orders you already ran:

• Price history per SKU, with dates. Answers "has this crept up?" without opening old emails.
• Payment terms, so a new PO inherits them instead of being re-typed.
• Promised vs actual production time — quoted 25 days, averaged 34 across the last six orders.
• Promised vs actual arrival, which is the number your safety stock should be sized from.
• Estimated vs actual landed cost, so freight assumptions get better.
• Defect and rejection rate, which is a silent addition to unit cost.
• MOQ, currency, incoterms, so quotes are comparable at a glance.

The consistent-25-quoted-34-actual finding is the one that pays for the whole exercise. It shifts every reorder date for that supplier by nine days — and nine days is the difference between a comfortable restock and an airfreight decision.

You are not collecting supplier data. You are keeping the receipts from experiments you already ran.

10The one that ambushes importers

If you pay a factory in USD, freight in EUR and duty in your home currency, a landed cost that sums the raw amounts is wrong by however much the market moved.

Three rules keep it honest:

• Store the currency and the rate on every payment, alongside the converted amount. Not a global setting — a property of the transaction.
• Convert at the rate on the date the money moved, not today's rate. A deposit paid in March was paid at March's rate, permanently.
• Scheduled future payments carry an assumed rate, and should be labelled as such — the amount is known in the supplier's currency, not in yours.

For a seller with meaningful FX exposure this quietly moves landed cost by several percent, which is enough to reorder a ranking of which products are worth restocking.

11Expected against actual

Every PO is a set of predictions that later resolve. Recording how they resolved is what makes the next PO better — and it's the cheapest thing on this list, because reality supplies the data for free.

Systematic error is a settings problem. Random error is the weather. You can't tell them apart without the record, and after five or six orders per supplier the pattern is unmistakable.

Which is the real answer to the accuracy question: the system doesn't get better at predicting. Your assumptions get better informed by your own history.

12Five numbers, together

For any product you'd consider reordering, these five belong on one screen. Separately they're reporting. Together they're a decision.

• 01Contribution per unitWhat it earns after landed cost, Amazon fees and advertising.
• 02Cover remainingUnits on hand and inbound, at current velocity, in days.
• 03Order-by dateWhen the next PO must be placed, given measured lead time.
• 04Cash required, and whenThe instalments that order creates, on their dates.
• 05Cash recovery dateWhen the resulting inventory starts returning money.

Numbers 4 and 5 are the pair almost nobody has, and they're the ones that turn "should I reorder?" into a question with a real answer. The distance between them is your cash conversion gap. Everything in this article exists to make that distance visible before you commit to it.

Frequently asked