Oversell combines quantity, time, and concurrency

ERP on-hand includes stock already allocated, held, damaged, or too late for the promised date. Concurrent channels can also sell the same stale snapshot.

ATP is a dated allocation by item, location, channel, and promise, backed by reservation and a conserved event ledger.

Govern item, location, and inventory state

Map warehouse, store, in-transit, consignment, and virtual nodes to fulfilment regions. Separate on-hand, allocated, reserved, blocked, quality, damaged, inbound, and available states.

Unify SKU, pack, unit, and bundle components. A bundle must consume every underlying item.

Declare the dated ATP formula

Subtract allocation, reservation, and buffer from sellable on-hand, add reliable supply arriving before date t, and subtract existing future commitments.

Use promise time buckets. Next-week supply cannot support today, and day-level data cannot honestly promise by hour.

Treat buffers and channel quotas as policy

Calibrate buffers from item value, service impact, accuracy, and lead time. Excess buffer creates false stock-out; too little creates cancellation.

Version channel quotas and release them when campaigns end. Commercial priority should not be hidden as a system error.

Use atomic reservation for concurrent checkout

Create a reservation with ID, quantity, order, channel, expiry, and version. Use atomic decrement or optimistic locking, then convert, expire, or release idempotently.

Periodic synchronization alone cannot prevent two channels from promising the final unit.

Grade inbound by reliability and date

A purchase order, shipped supply, received goods, and quality-released stock have different confidence. Include only supply usable before the customer promise.

Production plans also depend on material, yield, and schedule. Present uncertain supply separately.

Close cancellation, return, and transfer events

Cancellation release differs after pick or ship. Returns become sellable only after receipt and quality. Transfers remove source, create inbound, then add destination.

Do not count a transfer at both sites or restore a shipped cancellation immediately.

Govern channel lag and caches

Track source watermark, last success, backlog, and delay. Under excess lag, reduce promise, pause risky items, or use a conservative policy.

Cache by item, location, bucket, policy, and watermark, then validate atomically on reservation.

Accept under concurrency and failure

Test simultaneous last-unit checkout, payment timeout, duplicate cancellation, partial receipt, quality return, transfer, count loss, API lag, and restart.

Inject lost and reordered events, lock conflict, and throttling. Reconcile every difference to an order and event.

BI0 operating boundary

Pilot a small high-volume assortment across one warehouse and two channels, then monitor accuracy, oversell, false stock-out, reservation leak, lag, and fulfilment.

BI0 may support inventory integration and analysis. Atomic reservation and transaction writeback require explicit POC verification; analytics alone does not prevent oversell.

Public references

Take the next step with BI0.AI

Talk through a real business scenario and see how governed AI BI can fit your team.

Explore BI0.AI