This is the second piece in our series on retail stock. If what you need is the technical setup of the sync itself, it is in the guide to syncing stock between Shopify, Amazon and a physical shop; here we look at what comes before: understanding where the mismatch is born.
What overselling is, and why you only see it when the customer writes
Overselling is charging for a unit that has already sold through another channel, or that is no longer in the stockroom. The odd thing is that you never notice it at the time: the order comes in normally, the payment goes through normally, and the problem shows up hours later, when someone goes to pick it and the shelf is empty. You know the rest: an apology email, a cancellation, a refund and, sometimes, a review.
That is why almost nobody measures it. It is lived as a string of one-off incidents —“that order crossed over”— rather than what it is: a process that fails the same way every time. And it appears exactly when you sell the most: sales, campaigns, end of season. In other words, at the worst possible moment.
The six doors it gets in through
In the shops we have looked at, the mismatch almost always comes in through one of these six doors, and often through several at once:
- The time gap between channels. If the website learns about counter sales once an hour, for that hour the website keeps offering what the counter has already sold. The gap is the overselling.
- The last unit. With twenty units in stock a delay does not matter. With one or two left, every channel offers them at the same time and the first one to sell does not tell the others in time.
- Returns. A garment that comes back is added to stock before anyone checks it, or is never added and the system believes it sold. In fashion and footwear this is the widest door.
- Variants. Size 41 in black is one reference on your till, another on the marketplace and sometimes none on the website. What is not identified the same way everywhere cannot be deducted the same way everywhere.
- Reserved stock that looks available. A web order placed at 10:00 and waiting to be picked still counts as stock at the counter at 10:30. Physical stock and available stock are not the same number.
- Adjustments nobody records. Monday’s manual count, the display unit that got sold, the “I’ll note it down later”. Every unrecorded adjustment is a ghost sale waiting to happen.
The two numbers that make it visible
You do not need any system to know what it costs you. You need two numbers written down for one week, in a spreadsheet or on paper:
- Orders cancelled or changed for lack of stock. Every time a customer gets a “we’re sorry”, one mark.
- Stock corrections made by hand. Every time someone goes in to change a quantity because “it didn’t add up”, another mark.
The first measures the damage the customer sees. The second measures the work nobody sees. With both in front of you the conversation changes: it is no longer “sometimes an order crosses over”, it is a weekly figure that can be brought down and checked. They are the same two numbers we use in the one-week pilot before paying for any licence.
Stopping it in a week: what to do before building anything
There is an order that works, and it starts with decisions, not tools.
Days 1 and 2: decide who is in charge. Which of the three figures wins when they disagree —the till, the website or a central stock— and through which channel the last unit is sold when only one is left. That is what we call a single source of truth, and until it is decided everything else is patching.
Day 3: a buffer per channel. For references down to one or two units, publish the available stock on marketplaces minus one. You lose the odd sale of the last unit; you stop cancelling. It is a patch, and it comes off the day the sync runs instantly.
Day 3: reserve when the order is placed, not when it ships. The stock shown to the other channels has to be the physical stock minus what is already sold and waiting to be picked. If your system only deducts on shipping, every pending order is a potential oversell.
Day 4: returns with an intermediate state. Nothing goes back into sellable stock until someone checks it and marks it. A “pending review” state stops you selling a garment that came back stained, and stops the opposite case: the one that came back fine and nobody booked in.
Day 4: the channel that sells the most notifies, it does not wait. If one channel concentrates the sales, that is where the sale notice should go out instantly rather than waiting for the next scheduled read. How that notice is set up is explained step by step in the technical guide.
Days 5 to 7: the same two numbers again. If they have dropped, you know which door it was. If they have not, you know too: the cause is behind another of the six, and you have spent nothing finding that out.
What does not fix overselling
- Changing platform. The time gap between channels moves with you to the new platform.
- Buying a licence before deciding who is in charge. The software will overwrite one number with another in whatever order the notices arrive, which is exactly what already happens.
- Getting someone to count more often. It reduces ghost adjustments, not the gap between channels, and it collapses the day that person is off.
- Switching a channel off mid-campaign. You lose the sales and the cause is still there for the next one.
When it stops being a patch and becomes a system
If after the week both numbers are still high, the problem is no longer about decisions: information is taking too long to travel between systems, and only a sync that writes to every channel at the moment of sale, queues what it could not send and tells a person when something gets stuck will solve that.
We build it on top of what you already use —till, online store, marketplace— without changing systems, starting with a single product family and showing it working on video before anyone signs anything. How each piece connects is described in automation for retail.
Frequently asked questions
Is overselling the marketplace’s fault or mine?
It comes from the gap between your systems, but your account pays for it: marketplaces measure a seller’s cancellation rate, and a run of orders cancelled for lack of stock lowers the visibility of your whole catalogue, not just the reference that failed.
Won’t a stock buffer make me lose sales?
Some, yes: the last unit of each reference. In exchange you stop cancelling orders that were already paid, which costs more. The buffer is temporary: it comes off the day the sync runs instantly and the stock the channels see is the real one.
How often does stock have to sync to avoid overselling?
Before the next customer can buy the same unit. With high stock and slow sales, hourly is fine; during a campaign, with references down to one unit, only instant works, triggered by the channel that sells.
Does this apply if I only have a physical shop and a website?
Yes. Two channels already produce the gap: the till sells and the website takes a while to find out. The six doors and the two numbers apply just the same; there are simply fewer places to look.
What do I do with an order that has already been oversold?
Tell the customer before they ask, offer an alternative or the refund in the same message, and record the case as one more mark in the first number. The order that gets cancelled without an explanation is the one that ends up as a review.
Tell us your channels on our contact page and we will look at which of the six doors it gets in through. If you want the technical setup, it is in the sync guide.
You might also like