Most sellers on five marketplaces don't die from running out of stock. They die from five platforms selling the same stock at the same time.
I watched it happen last November. A seller I know had one hot car charger listed on Amazon, Walmart, eBay, his own Shopify store, and TikTok Shop. An influencer video took off overnight and pulled in over 3,000 orders. By morning, every platform was still selling against the full inventory number — which was, of course, fiction. Refunds, negative reviews, account health hits. The full combo.
Here's the playbook I hand to clients for exactly this situation. It has two halves: how to split the inventory, and how to stop the bleeding when demand spikes.
Cut the pool: five channels, five ledgers
The rule is simple: you can store everything together, but you must account for it separately.
Two ways to do it. Physical segregation means dedicating bin locations per channel — great for big sellers doing tens of thousands of orders a month, but it wastes space and makes transfers painful. Logical allocation means mixing inventory physically and carving virtual pools in the WMS — cheaper, flexible, and what I recommend for anyone under roughly 10,000 orders a month. Don't over-engineer this.
Set the split by actual sales weight, not gut feel. Take the last 30 days of sales per channel and divide the total inventory accordingly. If Amazon drove sixty percent of volume, it gets sixty percent of the sellable units. Then carve out a 5–10% flex pool that belongs to nobody — when one channel surges, you reassign from the flex pool. Review it weekly and shift the pool toward whatever is running hot.
Two things people get wrong here. First, units sitting in Amazon's FBA warehouses are no longer yours to control. Keep them on a separate ledger, never mixed with your self-fulfilled pool. Second, in-transit inventory is not sellable inventory. Ocean freight on the water and cartons the factory just shipped go on their own line of the spreadsheet. They enter the pools only after receiving and putaway. Counting in-transit as sellable is the second most common cause of oversells I've seen.
Everything above assumes you have sales history. New products are a different game. A launch SKU has zero days of sales, so the 30-day rule gives you nothing. Here's what I do: borrow the channel mix of the closest existing product in the same category — if your old power bank moved seventy percent of its volume on Amazon, allocate the first 5,000 units of the new charger the same way, but double the flex pool to around 15% for extra margin of error. Then recut the split after two weeks of real data, not at month-end. Clearance stock is a third case: don't spread it across five platforms. Concentrate it on the one or two channels that move fastest and show sold-out everywhere else. Slow movers die from being spread thin; concentrated firepower turns them into cash fastest.
Circuit breakers: three levels, in order
Pools alone won't save you. You need brakes, and they have to fire in order.
Level one is a sellable-quantity cap per platform. Your WMS pushes an updated sellable number to each marketplace every 15 to 30 minutes: allocated units minus sold-not-shipped minus reserve. Write that sync cadence into the SOP. Marketplace feeds lag — treat the delay as normal and plan around it, and check each platform's current inventory-feed limits before you build the integration.
Level two is the flex-pool alarm. When the flex pool runs dry, or a channel's sellable number drops below its safety line, everything tightens: all platforms get their sellable numbers cut automatically, or your ops lead approves any further release by hand. This step saved a client of mine during Black Friday 2023, when TikTok moved 8,000 units overnight — level two locked Amazon and the DTC store's sellable counts before the whole thing went oversold.
Level three is the full stop. When there's genuinely nothing left to sell, don't hesitate: pause the listings or stretch the promised ship date, and have the customer-service script ready. One oversold order's refund-plus-bad-review cost dwarfs the margin on a few missed sales. Do that math once and you'll never skip level three again.
One more thing: write down who pulls each brake. Level one runs automatically. Level two is the ops lead's call. Level three goes straight to the owner. Don't make it a committee decision — when oversells are happening, there's no time for a meeting.
Level two has a side effect you have to design around: false trips. A promo spike is normal — a flash sale can burn through hundreds of units in ten minutes and drag the sellable count below the safety line in one sync cycle. Trip the breaker then, and you've delisted healthy stock and killed your own traffic. So add a second confirmation: when the trip condition fires, freeze for fifteen minutes without cutting sellable numbers, and have the system recompute actual remaining stock — on-hand minus sold-not-shipped minus a rough allowance for carts not yet checked out. If there's still room, release and send the ops lead a warning only. If it's genuinely dry, run the three levels. The second confirmation runs on numbers, not on somebody's gut call, so you block real oversells without friendly-firing your own promos.
The warehouse handshake: if the numbers don't tie, nothing else works
All of this rests on one assumption: the WMS numbers are right. So the warehouse relationship is the real foundation.
Reconcile weekly. On-hand in the WMS should equal sellable across all platforms plus sold-not-shipped plus in-transit plus the flex pool. When those don't add up, something is leaking. The most common leak is returns — units come back into the warehouse but never get added back to anyone's sellable pool, so stock sits on the shelf while the listings show zero.
I recommend a fixed five-column sheet: SKU, on-hand, total sellable across platforms, sold-not-shipped, in-transit plus flex pool. Post it in the group chat every Monday morning. Anything that doesn't tie gets flagged red, and somebody owes an explanation that same day.
Another trap: the 3PL's reserved inventory. Some warehouses quietly reserve stock for big clients inside the WMS. If that reservation logic isn't transparent to you, your sellable numbers are fiction. Ask about it when you sign the contract and get it in writing.
When the numbers don't tie, I walk three checkpoints. First, returns not re-pooled: compare the returns area's on-hand against the sellable pools and find the gap; check whether receiving and the system putaway happened the same day — an overnight gap is where units go missing. Second, in-transit counted early: look for any SKU whose sellable count is higher than its on-hand — that almost always means somebody counted freight on the water. Third, opaque 3PL reserves: make the warehouse export its reservation detail, line it up SKU by SKU against the contracted reserve amounts, and make them explain anything over the line. Walk all three and you'll find the root cause nine times out of ten.
Tonight, do one thing: take your top SKU and put three numbers in a table — on-hand, sellable per platform, in-transit. Call your warehouse tomorrow morning and walk through it line by line. If it ties, your multi-channel inventory is healthy. If it doesn't, forget about growth for a week and fix the books first.
