Order Management Across Marketplaces
The second channel is easy. The third is where stock, orders and returns stop agreeing with each other, and the fix is structural.
One marketplace is a spreadsheet. Two is a spreadsheet somebody updates twice. Three is where the spreadsheet starts being wrong, and the first sign is usually an oversell on the channel that matters most.
Distributed order management is the unglamorous name for the answer: one place that knows what stock exists, which channel sold it, and where it has to ship from.
What Actually Goes Wrong
Overselling. Two channels sell the last unit within the same few minutes. One of those orders has to be cancelled, and a cancellation is a metric on every marketplace, not just an apology.
Stale availability. Stock is there, the channel does not know, and the listing sits unavailable while a competitor takes the sales.
Split fulfilment. The same product held in two places, so which warehouse ships depends on which channel sold, and nobody knows why one channel's delivery times are worse.
Returns nobody reconciles. A return goes back into stock on the channel it came from and stays invisible to the others, so the numbers drift a little every week.
What One Oversell Costs on Each Marketplace
A cancellation for stock that was not there is counted on every marketplace, and the allowances are small enough to be measured in single orders.
Amazon. For orders the seller ships, the pre-fulfilment cancellation rate is held under 2.5%, beside a late shipment rate under 4% and a valid tracking rate of at least 95%, with the order defect rate under 1% across every order.
Walmart. Cancellation rate is one of the measures in Walmart's Pro Seller tiers: 2.5% or lower for Rising, 2.0% for Advanced, 1.5% for Pro. The Walmart+ badge on seller-fulfilled items is reported to need a cancellation rate under 2.5% over a rolling 14 days, so one bad week shows within a fortnight.
eBay. A seller cancellation because the item was unavailable is a transaction defect, and Top Rated status on eBay's US site allows a defect rate of no more than 0.5%.
TikTok Shop. The seller fault cancellation rate allowed is 2.5%, which on forty orders is one.
Read in orders rather than percentages, the point is plain. A shop with forty orders a week on a channel has one oversell to spend before it crosses the line, and the oversell that spends it usually happens on the busiest day, on the channel hardest to recover. That is why the rule below comes before everything else on this page.
The One Rule That Fixes Most of It
One system owns the stock number, and every channel reads from it. Which system does not matter much. That there is exactly one matters enormously.
The most common failure is two systems that both think they are authoritative, usually an accounting package and a listing tool, reconciled by a person. That works until the person is on holiday.
Buffers, and Why They Are Not a Fix
Holding back a few units per channel prevents oversells and costs sales on everything with thin cover, which is usually the products worth the most.
Use a buffer while the sync interval is long, and treat it as a symptom. When the sync is quick enough, the buffer comes down, and the products that were quietly capped start selling again.
What Not to Centralise
Pricing. Each marketplace has its own competitive position and its own fee structure, so one price pushed everywhere is either uncompetitive in one place or unprofitable in another.
Listing copy, to a point. The product is the same and the way people search for it is not: a title written for Amazon's search box is not the title that works on eBay, and a feed that pushes identical copy everywhere is leaving ranking on the table on at least two channels.
Customer messages. Each marketplace has its own response time expectations and its own rules about what may be said, and a single templated reply is how a seller gets a policy warning for offering something off-platform.
Map the SKUs Before Anything Else
Every system downstream depends on one decision made first: what the business calls each sellable item. One internal SKU per item, owned by the business, and every channel's own identifier mapped to it: the marketplace's product number, its fulfilment label code, the store's variant ID. A channel's identifier is not a substitute, because each channel has its own and they change when listings are rebuilt.
Bundles and multipacks need the most care. A three pack sold on one channel and a single unit sold on another are the same stock, so the bundle has to be defined as a kit that uses up three of the component when it sells. Systems that treat the bundle as its own stock item are how a warehouse full of singles shows as out of stock for the pack. How the marketplaces identify products, and where barcodes fit, is in ASINs, barcodes and product identifiers.
Do the mapping in a sheet before choosing any tool. It is usually a day's work for a catalogue of a few hundred items, it exposes the duplicates and orphans every catalogue has, and it makes the tool's setup a matter of importing a list rather than discovering the catalogue during implementation.
When a Tool Pays for Itself
Roughly: more than two channels, or more than a few hundred SKUs, or any product held in more than one place. Below that a disciplined spreadsheet and one person is genuinely fine and much cheaper.
What to ask of a tool: how often it syncs, what it does when a channel's API is down, whether it maps your own SKU across channels rather than making you rename everything, and whether it handles returns back into stock. The last question is the one that gets skipped and the one that causes the drift.
Dispatch Deadlines Differ by Channel
Each marketplace sets its own handling time and measures its own on-time dispatch, and they do not agree. A warehouse that picks orders in the order they arrived will miss the strictest channel's deadline on a busy day while shipping a more forgiving channel's orders early.
The fix is to pick by deadline, not by arrival: every order carries the time it must leave by, from its channel's rules and the handling time set on the listing, and the pick list is sorted on that. A single daily cut-off for the warehouse has to be set by the strictest channel it serves, not the most common one.
The handling time on each listing should be honest about what the warehouse can do. Setting it shorter to win a faster delivery promise works until the first busy week, and the late shipment metric it produces is measured on the channel, where it costs far more than the conversion it bought.
Routing an Order When Stock Sits in Two Places
Once stock is held in more than one place, every order needs a rule for where it ships from, written down rather than left to whichever warehouse sees the order first.
The channel's promise first. Only a location that can meet the delivery date the channel showed the buyer is a candidate. A cheaper warehouse that ships late saves money on the label and spends it on the late shipment rate.
Then the cost. Among the locations that can meet the promise, the cheapest to ship from, which is usually the one nearest the buyer.
Then keep the order whole. Splitting one order across two locations doubles the shipping and the chances of a problem. Split only when no single location can fill it and the channel allows it.
Respect stock that is spoken for. Units in a marketplace's own fulfilment network serve that marketplace's orders, and other channels' orders only through its multichannel service, on that service's rules and fees. Amazon's and Walmart's multichannel services compares the two.
Whatever the rules, the order layer should apply them automatically and record which location shipped each order, so the weekly read can tell a warehouse problem from a routing one.
Fulfilment Is Part of the Same Question
If Amazon holds your stock and another channel sells it, something has to raise a fulfilment order and pass the tracking back before the channel's dispatch deadline. That is a real integration and not a manual job at any volume.
The alternative is holding separate stock per channel, which is simpler to run and ties up more money. Neither is wrong; what is wrong is drifting into one of them without deciding.
Returns Back Into Stock, Graded
Returns are where the stock number drifts, because a returned unit is not stock until somebody decides it is. Give that decision a fixed path.
Receive against the order. Log each return against the order and the channel it came from, so the refund, the stock and the reason all point at the same sale.
Grade it. Sellable as new, sellable only as used where a channel allows that condition, repairable, or written off. The grade decides which stock number the unit joins, and a used unit put back into the new count is the next buyer's complaint.
Restock in the central count. A sellable unit goes back into the one stock number every channel reads, not into the channel it came from, so the next sale can come from anywhere.
Watch the units the marketplaces hold. A return to a marketplace's own warehouse is assessed there and comes back as sellable or not. The ones that are not need a decision, removal or disposal, before they cost storage, and they belong in the weekly read below.
Keep the reason. The return reason, by product and channel, is the cheapest product research a business has, and the license plate number on an Amazon return is how a single returned unit is traced.
The Weekly Numbers That Show Drift Early
Order management fails slowly, so a handful of numbers read weekly catch it before a customer does. Cancellations by channel, and the reason for each, because an oversell shows up here first. Sync failures, from whatever moves stock between systems, since a sync that failed quietly on Tuesday is a stock count that has been wrong since Tuesday.
Stock accuracy, from a small rotating count of real shelves against the system: ten SKUs a week, chosen differently each week, is enough to show whether the number can be trusted. Late shipments by warehouse rather than by channel, which separates a warehouse problem from a channel rule. And returns waiting to be restocked, because every unit sitting there is stock the channels cannot sell.
None of these takes long to read, and together they say whether the order layer is holding. When one moves, it usually points at a single missing sync or step rather than at the whole system.
What We Do With This
The order layer is what decides whether a third channel is a growth or a mess, which is why it comes up before anybody adds Walmart or eBay to an account that already sells in two places.
It is also the argument for keeping your own store honest: a store that reads the same stock number as the marketplaces is a channel, and one that keeps its own is a fourth thing to reconcile.
Every one of those channels is a service we run, and the reason they are described together on one page rather than sold separately is that the order layer underneath them is one problem, not four.
If your numbers stopped agreeing and nobody is quite sure when, it is usually one missing sync rather than everything.
Related reading: Amazon FBA Explained: How It Works, What It Costs, and Whether It Is Worth It, Using Amazon to Fulfil Orders You Did Not Sell on Amazon, eBay Inventory Management: Where Cancellations Come From and Shipping Hazmat From a Marketplace: Lithium Batteries, Perfume, Aerosols and the USPS Fee.
Frequently Asked Questions
The questions that come up most often on this subject.
One system that knows what stock exists, which channel sold it and where it ships from, with every marketplace reading from that single number instead of keeping its own. The distributed part is the stock and the fulfilment; the management is having one authority over both.
Make one system the owner of the stock number and have every channel read from it, then shorten the sync interval until a buffer is no longer needed. Buffers prevent oversells by capping sales on exactly the products with the thinnest cover.
Roughly above two channels, a few hundred SKUs, or any product held in more than one place. Below that a disciplined spreadsheet and one responsible person genuinely works and costs nothing.
How often it syncs, what happens when a channel's API is unavailable, whether it maps your existing SKUs rather than forcing a rename, and whether returns go back into stock automatically. The last one gets skipped and is what causes numbers to drift.
Yes, through multi-channel fulfilment, and it needs a real integration: something has to raise the fulfilment order and pass tracking back before the other channel's dispatch deadline. At any volume that is not a manual job.
It is simpler to run and ties up more money, and it is a legitimate choice. What causes trouble is drifting into it without deciding, so that nobody can say which pool a given unit belongs to.
Often enough that the buffer can come down. The right interval is the one where two channels cannot both sell the last unit, and on fast-moving lines that is minutes rather than hours.
More than the order. A cancellation is a performance metric on every marketplace, it can affect the buy box on Amazon and seller standing on eBay, and a run of them attracts attention that a single one does not.
They are the part most setups handle worst. A return that goes back into sellable stock on one channel and stays invisible to the others makes the numbers drift a little every week until somebody counts by hand.
It should be. A store reading the same stock number as the marketplaces is one more channel; a store keeping its own numbers is a fourth thing to reconcile, and it is usually the one that gets reconciled least often.