Here is the conclusion first: available to promise is not "how much stock is left in the warehouse." It is "how much can you still commit to the next order without breaking a promise you already made." The formula fits on one line:
ATP = On-hand inventory + Incoming supply − Allocated quantity − Safety stock
Get any one of those four numbers wrong and the promise your OMS hands out at order intake is fiction. One confusion to clear up first: promised quantity and promised date are two different things. Quantity answers "how much can we take." Date answers "when can we ship it." You can get the quantity right and still miss the date, because the stock is on a truck that has not arrived. Get the quantity wrong and no delivery date, however nicely worded, is worth anything. This article is about getting the quantity right at order intake. What to do after you are already short — who gets cut first — is a different topic.
IHL Group research puts a number on what sloppy promising costs: inventory distortion, meaning stockouts plus overstocks, burns about $1.7 trillion globally a year. That sounds abstract until it lands on one warehouse. Customer service promises delivery in three days. The warehouse finds half the quantity. The rest is on a supplier's truck that has not docked yet. The promise is already in the customer's inbox, and every fix from that point on comes out of your margin.
Four numbers in the available to promise formula, none of them guesswork
On-hand inventory is the easiest number to overstate. The quantity in the system is a book number; the quantity on the shelf is the real one. Cycle-count variances, damaged goods never written off, returns sitting in receiving that never got put away — each one inflates the book figure. My rule: ATP only counts stock in pickable status. Quarantine, pending inspection, and damaged zones are excluded from the formula, full stop. Which statuses count as pickable gets decided before go-live, written into the configuration document, not left to verbal agreement.
Incoming supply only counts when it is confirmed and dated. A supplier saying "shipping this week" on the phone is not supply. A purchase order the supplier has acknowledged, with a tracking number and an expected arrival date in the system, is supply. Unconfirmed purchase orders and unscheduled production runs do not belong in the formula. Building a customer promise on them means moving the supplier's risk onto your delivery date. I have seen this exact failure three times, all on the night before a peak season sale: purchasing said the goods were on the road, the OMS took them at face value, the truck arrived two days late, and oversold orders blew up one after another.
Allocated quantity is the easiest to undercount. Allocation is not just confirmed sales orders. It includes reservations held for key accounts, quantities tied up in transfers in transit, and stock reserved for replacement orders. Plenty of systems go live with this landmine buried: reservation rules live in a spreadsheet, not in the OMS. Sales locks a batch for a big customer by hand, the system never sees it, and ATP promises the same batch to the next customer. Nobody finds out until picking day.
Safety stock comes out of ATP. It is a buffer held back for demand spikes and late deliveries, not stock for sale. Counting it as available to promise is selling the life jacket. Here is the counterintuitive part: a negative ATP is not the scary outcome — a wrong positive one is. A negative number is at least honest: the system blocks the order, raises an alert, forces a fix. An inflated positive number blocks nothing, and by the time you notice, the overselling has already happened. I would rather have ATP conservative than good-looking.
Here is a full worked example, with hypothetical numbers you can swap for your own:
| Item | Quantity (units) | | On-hand inventory (pickable status) | 800 | | Incoming supply (confirmed, arrives Friday) | 500 | | Allocated quantity (sales orders + key-account reservations) | 600 | | Safety stock | 200 | | Available to promise | 500 |
The math: 800 + 500 − 600 − 200 = 500. One detail matters: the 500 incoming units do not arrive until Friday, so the promised date for those 500 units cannot be today. Quantity and date are calculated separately and promised separately. That is what "two different things" means in practice.
How overselling actually happens: three root causes
The following project story is a composite scenario — the pattern is one I have seen recur across several clients, with details generalized. A home-goods client, the week before a big sale, me on site for go-live. Their OMS took orders fast and the ATP check was switched on. They oversold anyway. When we traced it, all three root causes were firing at once. I will rank them by damage.
Root cause one: incoming supply out of sync. The purchase order's arrival date changed in the ERP and the OMS never got the update. ATP still calculated on the old date, assuming stock tomorrow when it was really next Wednesday. The biggest risk with incoming supply is not that it is missing — it is that the date is wrong. One wrong date and every promised date downstream is wrong.
Root cause two: reservation rules in chaos. Key-account reservations lived in a sales spreadsheet. Rush orders from customer service grabbed stock by phone call. The OMS's ATP could see none of it, so the available quantity it calculated was far higher than what could actually be sold. Reservation rules have to live in the system, with an effective time, an expiry time, and an owner. A verbal reservation is no reservation at all.
Root cause three: safety stock eaten alive. Peak season arrives, sales complains ATP is too low, and someone agrees to "release the safety stock for now." Orders that day look great. A week later the supplier delivers two days late, the cushion is empty, and the shortage cascades. Releasing safety stock must go through approval, and it needs a cap — release at most half, and the rest does not move no matter what.
Of the three, out-of-sync supply is the most common, chaotic reservations the hardest to trace, and eaten safety stock the most damaging. When ATP is wrong, start here — nine times out of ten, you will find it.
The go-live checklist for available to promise
ATP is not a formula you write once and forget. It is four actions that have to run together. Walk this checklist the night before go-live and tick each box:
- Lock the definitions. One page that defines all four numbers: which statuses count as on-hand, which statuses count as incoming, which documents count as allocated, how much safety stock there is, and who is allowed to change any of it. Business, operations, and IT all sign it, and it goes in the project channel.
- Set the refresh rhythm. On-hand and allocated quantities change in real time, so the OMS calculates ATP live at order intake. A full recalculation runs at least once a day, ideally overnight, and hourly during peak season. ATP that only refreshes in a daily batch job will be wrong by the first morning of a big sale.
- Set the oversell circuit breaker. ATP below zero means the system blocks the order automatically and routes it to manual review — never auto-releases. ATP below the safety waterline for two hours straight alerts the operations owner. One SKU tripping the breaker three times in a day means every remaining order for that SKU goes through manual confirmation. The breaker is not there to slow the business down; it stops overselling at order intake instead of discovering it at shipping time.
- Set the reconciliation. First thing every morning: compare the ATP figure against the WMS physical count and investigate any gap over 2%. Once a week, reconcile incoming supply: has any purchase order's arrival date moved, and did the change reach the OMS? Skip reconciliation and ATP drifts silently until someone pays for it.
One last honest word. The technology behind ATP is not hard. The definitions and the discipline are. Vague definitions, refresh cycles that lag, reservations by phone call, safety stock released on a whim — no OMS, however expensive, can calculate accurately on top of that. Flip it around: definitions locked, refresh on schedule, circuit breaker standing, and an ordinary OMS will hold its promises. Give the ATP line on your go-live checklist a full half hour. Before you promise a customer, ask one question: would I put this number in a confirmation email? If the answer is no, do not let the system promise it for you.
By Lin Xiao, WMS Implementation Consultant







