Multichannel inventory management
Our products › Multichannel inventory management
One stock position, several channels, and a decision on every order about who gets the unit.
Allocation only works on top of real detail rather than a total, which is the argument the inventory pillar makes in full. Inventory management software.
The moment it starts
Two orders inside a minute, one unit left. Your warehouse knows immediately. The channel finds out somewhere between forty seconds and four minutes later, depending on how the update travels and how busy the queue is.
In that gap you sold it twice. Nobody made a mistake, no integration was broken, and no amount of accuracy in the stock figure would have prevented it — the figure was right at both moments.
Overselling on a multichannel operation is not usually a data-quality problem. It is a timing problem, and timing is not fixed by counting more carefully.
The gap does not close
Measure it rather than assume it. Take the time between a pick confirmation in the warehouse and the availability figure changing on the channel, and take the worst case rather than the average — the average is a comfort, the worst case is what happens in peak.
Multiply that by your peak order rate for the item. That number is how many orders you can take in the window that you have no mechanism to prevent. For a fast mover on a busy Monday it is rarely zero.
Faster connections shrink the window; they do not remove it. There is always a last unit and always an interval.
So you list less than you hold
The conclusion feels wrong the first time. You have the goods, and you are deliberately not offering all of them.
It stops feeling wrong once you price the alternative. Holding a small buffer back from a channel costs you the margin on the units you did not sell that day. Selling a unit you cannot deliver costs the repick, the apology, the refund handling, and whatever the channel does to you afterwards.
The right size of the buffer is a commercial decision per channel, not a technical default. A marketplace that penalises cancellations gets a conservative threshold; your own webshop, where you control the conversation, can run closer to the line.
Set the threshold where the cost of a cancellation on that channel outweighs the margin on a unit you did not list. That number differs per channel, and treating them all the same is the actual error.
What a cancellation costs on a rated channel
The repick is the small part. On a channel that rates sellers, a cancellation attributed to you feeds a defect measure, and that measure moves placement.
Ranking suppression outlasts the mistake by weeks, and its cost is invisible in a way the refund is not: you never see the orders that stopped arriving. There is no line in any report for demand that was never shown to you.
That asymmetry is the whole argument for listing less than you hold — and it is why the classical allocation models, which price a refusal at the lost margin, understate the answer on a rated channel.
Who wins when stock is short
When two channels want the same unit, somebody decides. Today that is usually whoever's order arrived first, which is a decision nobody made deliberately.
The commercial answer is rarely first-come. It might be the contract customer with a penalty clause, the channel where a cancellation costs most, the order that completes rather than splits, or the one shipping from the location that can still make today's cut-off.
Those are business rules, and they belong somewhere a commercial person can change them without raising a ticket.
Allocation is a commercial decision that has been left to arrival order by default. Making it deliberately is most of the available gain.
Frequently asked questions
What is multichannel inventory management?
Holding one stock position and deciding, per order, which channel gets the unit — with availability published per channel rather than one shared number. It exists because several channels drawing on the same goods will otherwise sell the same unit twice.
Why do we still oversell when our stock figure is correct?
Because overselling is a timing problem, not an accuracy problem. Between the pick confirmation in the warehouse and the channel's availability changing there is a window, and orders that arrive inside it cannot be refused. A more accurate figure does not shorten the window.
Should we publish all of our stock to every channel?
Usually not. A per-channel threshold, set commercially, holds back a small buffer where a cancellation is expensive. The correct size depends on what that specific channel does to a seller who cancels — it is not one global setting.
What does a cancellation actually cost on a marketplace?
Far more than the refund. It feeds a seller defect measure, and the resulting placement suppression outlasts the incident by weeks. The cost that matters is the demand you were never shown, which appears in no report.
One stock position, several channels
Half an hour on your channel mix and your real latency — including whether the problem is worth solving yet.