The second sales channel, not order volume, breaks fulfilment
Ask when to automate fulfilment and the answer comes back in orders per day. Fifty, a hundred, whatever number sounds like a real shop.
The number is the wrong unit. One channel can be run by hand for far longer than people expect, because the stock figure lives in exactly one place and cannot disagree with itself. The day a second channel opens, every sale on one becomes a quiet lie on the other until somebody updates it - and that gap is where oversells live.
Amazon's own advice to sellers with too many cancellations comes down to the same thing: if you sell through other channels, you may have to update your inventory on Amazon several times a day.
The denominator, not the volume
Here is the arithmetic nobody puts in front of small sellers, and it runs against intuition.
Amazon's pre-fulfilment cancel rate measures seller-cancelled orders as a percentage of total orders over a seven-day window, and the requirement is to stay under 2.5%. Cancellations you initiate count. Cancellations the customer asks for do not.
Now do the division. An Amazon community manager, explaining to a seller how a single cancellation produced a 5% rate, spelled it out: a 5% figure "would imply an order volume of around 20 or so orders to have a single order reach 5%."
So at roughly forty orders a week, one oversell is 2.5% and you are at the threshold. At two hundred orders a week the same mistake is 0.5% and nobody notices. The metric is a fraction, small sellers own the small denominators, and the protection people assume they have from being small is the opposite of true.
Above the line, Amazon's policy summary is one sentence: keep your cancellation rate below 2.5% to prevent deactivation of seller-fulfilled offers. In practice a plan of action is usually requested first.
What the platforms do about it
The policies are unusually specific, and worth reading once with your own numbers in mind.
| Platform | Metric | Limit | Measured over |
|---|---|---|---|
| Amazon | Pre-fulfilment cancel rate | under 2.5% | 7 days |
| Amazon | Order defect rate | under 1% | 60 days |
| Amazon | Valid tracking rate | above 95% | - |
| Amazon | Late dispatch rate | below 4% | - |
| eBay | Transaction defect rate | no more than 2% | 3 months if over 400 transactions, otherwise 12 |
| eBay Top Rated | Transaction defect rate | no more than 0.5% | same, checked on the 20th |
| Walmart | Cancellation rate | 2% or below | - |
eBay is the most direct about the cause. Its policy counts a transaction defect when "the seller cancels the order unexpectedly (e.g. because it was out of stock, or because they sold it to someone else)". Selling the same unit twice is not an edge case in their model; it is the textbook example.
The consequence is money. Below Standard sellers face limits on selling activity "including charging higher final value fees", and sellers who qualify for Top Rated Plus - which also requires same or next business day dispatch and a free 30-day return option - get a 10% discount on final value fees that a rising defect rate takes away. A stale stock number has a price list.
The same mistake, two different outcomes
eBay publishes two worked examples worth more than any general advice. They differ in several details, but they land on the same defect rate and opposite verdicts, and eBay is explicit about which difference decides it.
A fabric shop finds a damaged bolt, cannot update the listing until the next day, and cancels 25 transactions - a 2.5% defect rate. Those cancellations hit two buyers. Result: Above Standard, no consequence.
A bike parts seller also cancels 25 transactions at 2.5%. Those cancellations hit 25 different buyers. Result: Below Standard, with the fee penalty that follows.
The difference is a protection in eBay's rules: you are only evaluated as Below Standard if defects are spread across more than four buyers. One bad SKU that disappoints a couple of people is survivable. The same error rate scattered across your customer base is not.
Which tells you something about which oversells to fear. It is not the big one-off; it is the slow drip across many buyers, which is exactly the shape produced by stock drifting out of sync on a second channel.
Why labels feel like the first thing to fix
Ask sellers what they automated first and shipping labels come up early. Thirty or forty orders a day across three channels, half needing different carriers, and printing them one at a time eats the morning. Automating that gives back hours you can point at.
It is a real gain, and it is the wrong first move, because of how the two kinds of error behave.
Label work is linear. Ten more orders is ten more minutes. Painful, but predictable, and it never gets worse than proportional.
Inventory error compounds. One oversell becomes a cancellation, which becomes a refund, a support conversation, a metric with a threshold attached, and sometimes a hit to account health that costs money on every future sale. The damage is not proportional to the mistake.
Nobody notices this ordering while it is happening, because labels hurt every morning and stock drift hurts once a quarter - in a way that is expensive and hard to trace back.
What the platforms solve natively, and what they leave to you
Before buying anything, know what you already have, because the answer is less than most sellers assume.
Shopify keeps stock per location, and its documentation is explicit that "each location's inventory is independent and can't be shared or pooled". It is a commerce platform with sales channels, not a stock synchroniser for marketplaces you sell on elsewhere.
Amazon's Multi-Channel Fulfilment will ship your Amazon-held stock for orders placed elsewhere, which solves the picking and packing, not the counting. Even Veeqo, which Amazon owns and gives away, puts inventory synchronisation behind paid add-ons for parts of what sellers expect from it.
So the gap between channels is real, and it is yours to close.
How fast can it close? Officially, nobody says. Neither Amazon nor eBay publishes a guaranteed interval for how quickly a stock update propagates through their systems. That absence is itself planning information: if the platforms will not promise a number, do not design a buffer-free system that assumes one.
Automate in the order the damage runs
The sequence follows from everything above.
First, one stock number. Whatever holds it - software, or a single spreadsheet everything else copies - there must be exactly one authority, and every channel must read from it. This is the fix that stops the compounding error, and it is first even though it feels less urgent than labels.
Second, orders into one view and tracking back out. Pulling every channel's orders into one place and pushing tracking numbers back is mechanical work with clear rules, and it is where labels stop being a daily tax.
Third, and carefully, returns. Returns are low volume and high judgement, which is the worst combination for rules. Mistakes are visible to the customer directly. Most sellers keep a person in the loop here long after the rest runs itself.
One honest exception to that ordering, because it comes up: if returns are a large share of your volume - clothing, sizing-heavy categories - the calculation changes, and a mishandled return that sits unprocessed for weeks can turn into stranded stock. Order the work by where your own hours and losses actually go, not by a general rule.
That is the same principle as where to start automating a business that already works: the first thing to automate is rarely the thing that annoys you most.
Before you buy anything
Four checks, none of which needs a vendor:
- Count your channels, not your orders. One channel and no plans for another? Most of this can wait. Two or more, and stock sync is already overdue.
- Do the fraction. Take your weekly order count and divide one cancellation by it. If that number is near 2.5%, a single oversell is a policy problem, not an inconvenience.
- Find where the real stock number lives today. If the honest answer is "in two places, and we reconcile them when we remember", that is the finding.
- Check your current defect and cancellation rates on each platform before deciding anything. They are the only unbiased measure of whether the manual method is still working.
And keep the alerting in mind as you add tools: an integration that silently stops updating stock produces exactly the same symptom as no integration at all, which is why automations that fail silently are worth designing against from the start. Amazon's own guidance makes the same point about third-party integrators - if you use one, take measures to prevent it making a mistake on your behalf.
If you would rather have someone map this with you, that is what the process audit is for: $299, three business days, and a written account of what to automate first and what to leave alone.
Sources
Amazon's thresholds are summarised on Amazon's selling policies page; the seven-day window and the arithmetic of one cancellation are from Amazon staff on its seller forums: how one cancellation reaches 5% and tracking and dispatch targets.
eBay publishes the seller standards policy, including both worked examples, and warns about overselling in its listings help.
Elsewhere: Walmart's seller performance standards and Shopify on stock held per location.