A retailer upgrades to real-time inventory sync, and the oversells continue. An online order was cancelled because the unit it claimed was already promised to a store pickup, even though the feed that sold it was current to the second.
Real-time sync solves a freshness problem, and overselling is rarely a freshness problem. The number most systems publish counts units that are already reserved for other orders, held back for walk-in shoppers, or sitting in locations that do not ship online. The storefront keeps selling stock the channel could never honor.
This blog explains why publishing raw on-hand stock causes overselling no matter how often it syncs, why the flat-buffer workaround fails in both directions, and what it takes to publish a number each channel can honor.
Why raw on-hand is the wrong number to publish
Most inventory systems broadcast raw on-hand, the count of units physically sitting in stores and warehouses. That number is easy to trust because it is literally true: those units exist.
Existing and being sellable are not the same thing. A unit can sit on a shelf and still be unavailable to sell, because it is reserved for a pickup order, held back as safety stock, or parked in a location that does not fulfill online orders. Raw on-hand counts all of it, so the number a channel can actually honor is smaller. None of this is exotic. Every retailer already runs reservations and safety stock; raw on-hand hides them inside one figure.
The safe-to-sell number has a name: Available To Promise, or ATP. It is on-hand minus everything that makes a unit unavailable to a given channel, and publishing raw on-hand instead of ATP is what produces oversells. A real-time feed does not fix that; it publishes the same figure more often.
The scale of the miss is not small. Global inventory distortion, out-of-stocks plus overstocks, reached $1.77 trillion in 2023. For a single brand, even a low oversell rate becomes a steady stream of cancellations as volume climbs, and each cancellation costs a shipment, a refund, and some customer trust.
Freshness tells you how closely a system mirrors the shelf. It says nothing about whether that figure should have reached a shopper, and that second question is the one overselling turns on.
Why a flat percentage buffer does not stop overselling
The usual response to overselling is two moves: sync faster, then hold back a flat percentage of stock on each channel. You set the buffer at 10% and hope the cushion absorbs the mistakes, but it rarely does.
A flat buffer fails in both directions. Set it too low and the oversells continue, because the buffer has no idea which units are already spoken for. Set it too high and sellable stock disappears from the site, so the brand sits on inventory it could have sold. One blunt number cannot manage a problem with this many moving parts.
The reason is that raw on-hand ignores everything that decides whether a unit is sellable: reserved units, safety stock, the network-level threshold, and facility and channel rules. A flat percentage cannot tell that a unit is reserved for pickup, or that a store is not part of the online channel. It spreads one guess across all of them, and faster syncing delivers the same flawed figure to every channel sooner.
The result is common enough to measure. About 34% of retailers have shipped an order late after selling a product that was out of stock. Many operations leaders have watched a flat buffer fail and suspected the published number itself was the problem, without a clean way to prove it.
How Online ATP decides what each channel can sell
Fixing this starts by separating two jobs most systems blur together: seeing stock, and deciding what to promise. HotWax Commerce is an omnichannel order management system built for the second job. It sits between eCommerce, the ERP, and store systems as an orchestration layer, pulling raw counts from every source and then computing what each channel can safely sell.
The first step is a single pool. HotWax unifies on-hand stock across every location, including stores, warehouses, and distribution centers, into one inventory picture. That pool is only the input. The published number comes from the Online ATP calculation that runs on top of it.
The calculation nets out what raw on-hand ignores. Reserved units come out first, the stock already committed to pickup and open orders. Safety stock comes out next, the per-store buffer that protects walk-in shoppers. A network-level threshold holds back a company-wide cushion against count discrepancies, and units claimed by orders that are captured but not yet routed to a location are earmarked as well. Stock at facilities that are not enabled for online selling never enters the pool at all.
The number left after those deductions is Online ATP, a count of genuinely sellable units rather than the raw shelf count. HotWax recomputes it as orders, receipts, transfers, and counts change, so the storefront keeps reflecting what the network can actually deliver.
Retailers set the rules behind each deduction in the ATP app, by product, facility, or channel, and the buffers do not have to be uniform. A new launch that sells out fast can carry a larger safety stock than the rest of the catalog, a flagship store can hold back more units than smaller locations, and a returns facility can sit outside the online channel entirely. Rules apply by product tag, so a rule set once covers every matching product, and HotWax nets all of it into one Online ATP number per channel. That is the difference from a flat buffer: the reserve matches the product and the location it protects.
Because the rules are per channel, the same on-hand count can produce different promises. eCommerce might see the full sellable figure, while a marketplace with stricter cancellation penalties carries a higher threshold and sees a more conservative one. Nothing changes physically in the warehouse when a brand moves to Online ATP; it simply publishes the number it can deliver instead of the number it can count.
Each deduction is on the record, so the team can trace how a published figure was reached: how many units are sellable, how many are protected by safety stock or the threshold, and which locations contributed nothing. A number the team can trace holds up when a customer, a marketplace, or finance asks how it was set.
Accurate promises also sell more. When accurate delivery timing appears in the cart, 60.7% of shoppers are more likely to buy. A number the operation can keep protects the brand from cancellations and gives shoppers a reason to check out.
Key takeaways
- Real-time sync makes an inventory number fresher. It does not make the number safe to promise.
- Trustworthy inventory comes from publishing Online ATP, the count of units a channel can actually sell.
- On-hand stock becomes sellable only after reservations, safety stock, the network-level threshold, unrouted orders, and facility participation take their cut.
- A flat percentage buffer either keeps overselling or hides stock the brand could have sold, because it cannot see which units are already spoken for.
- Online ATP rules are set per channel, so the same physical stock can back a different promise on each storefront.
- Every deduction behind the published number is on the record, so the team can explain how the figure was reached.
* * *
Book a HotWax Commerce demo and bring one order that was oversold. Walk through what was published that day, what Online ATP would have subtracted, and the number that would have reached eCommerce instead.
