Bottom line up front: the partial shipment decision — whether to ship available items first or wait for the complete order — is not a customer-service judgment call. It's an OMS rules configuration problem. When the rules are configured, the system decides automatically. When they're not, CSRs argue about it in chat every day and the warehouse waits for instructions. I've done 30+ OMS implementations, and every operation that left partial shipment to human judgment ended up a blame-shifting mess.

Why Human Judgment Doesn't Scale

Two projects I've lived through.

The first: a beauty brand in 2023, warehouse in Los Angeles. Their policy was "ship what's available" — sounds customer-friendly, right? One order had 5 SKUs, 1 out of stock, and the warehouse ended up shipping it in 3 waves, because the missing SKU arrived in three separate inbound batches. The brand ate the second and third shipping costs — over $20,000 extra in freight that month. Worse: the first box arrived with no note inside, so the customer assumed it was a mispick and left a one-star review.

The second: a 3C accessories seller in 2024, with the exact opposite policy — "never split, always wait." One big promo order: 9 SKUs, 1 out of stock, waited 11 days. The marketplace flagged it as a late shipment, fined them, and they lost a major customer's repeat business. I've seen this failure three times, and the root cause is always the same: the decision logic lives in someone's head instead of in the system.

So step one isn't picking a policy. It's writing the policy as rules the OMS can execute.

The Partial Shipment Decision Logic: Four Gates

The rule set I configure for clients is four gates, evaluated in order. Any gate that says "wait" stops the split:

Gate 1: SLA. Does the order carry a promised ship time? Marketplace orders with SLA penalties — ship what's available, no debate. The penalty is certain; the second-shipment cost is controllable. This math doesn't need a spreadsheet.

Gate 2: Shortage ratio. If the out-of-stock value exceeds a threshold of the order value (I usually recommend 30%), wait. Shipping when a third of the order is missing loses on both freight and experience. Below the threshold, ship.

Gate 3: Customer tier. VIPs, key accounts, subscription customers get their own policy. These customers would rather wait a little and receive everything at once — they buy peace of mind, not one-day-faster.

Gate 4: SKU attributes. Bulky, high-value, fragile items: never split. Sets sold as a unit (a 4-piece dinnerware set): if one piece is missing, the set doesn't ship. Consumables and standard items: split freely.

These four gates are four configurable rule conditions in the OMS, with thresholds each client sets themselves. On my go-live-night checklist, I always make the client fill in all four thresholds and sign off — because in the first month after launch, customer service will absolutely ask "why did this order split," and that's when you pull out the signature.

Backorder Auto-Generation: Three Non-Negotiables

Once you decide to ship the available portion, the missing items must auto-generate a backorder. Three non-negotiables — miss one and it blows up:

  1. It must inherit the original order's priority and promises. A backorder is a continuation of the original order, not a new order. Customer tier, promised ship date, shipping method — all inherited. I've seen a system treat backorders as ordinary new orders, dropping them into the standard queue. A VIP customer's replenishment waited two weeks. CSRs got destroyed.

  2. It must recalculate the promised date. Take the out-of-stock SKU's ETA, add outbound processing time, produce a new estimated ship date, and notify the customer automatically. No "will ship soon" vagueness — give a real date. If the ETA moves, the date updates automatically.

  3. It must stay linked and traceable. When the customer tracks the package or the CSR looks up the order, it must be instantly clear: "this is the backorder for order X, with Y items on the way." Break the link and the customer assumes you charged them twice.

The Partial Shipment Cost Math: Calculate Before You Split

Most people can't do this math, so here's the framework. True cost of a partial shipment = second freight + second pack labor + CSR communication cost + hidden cost of bad reviews and returns.

Take the beauty brand (real magnitudes from that project): one order split into 3 shipments. First leg $8, each follow-up leg $6 — average freight per order jumped from $8 to $20. Two thousand split orders a month meant $24,000 in extra freight. Had they waited for complete orders instead, average delay was 4 days, and late penalties plus churn cost under $8,000 a month. The right answer for that client was "wait" — but they started with a gut-feel "ship."

Now the 3C seller: waiting 11 days cost over $30,000 a month in fines and lost orders. Shipping partial would have cost barely $10,000+ a month in second freight. Their right answer was "ship."

See the pattern? There's no universal answer — only an answer you've done the math on. That's exactly why the decision logic belongs in the system: humans go with gut feel, systems follow thresholds.

One Counterintuitive Take

Here's the one most people miss: the biggest cost of partial shipment usually isn't freight — it's the customer thinking you shipped the wrong items.

Customer orders 5 items, receives 3, with zero explanation. Their first thought isn't "oh, 2 more are on the way." It's "this seller messed up my order." Then comes the bad review, the refund request, the support ticket — all three. So anyone doing partial shipment must put a backorder notice card in the box (physical paper — don't count on the customer reading email), and trigger an SMS/email the moment the backorder is created. Skip that step and the partial-shipment experience scores negative.

One more item for the go-live-night checklist: go to the warehouse and watch a split order get packed with your own eyes, and confirm the notice card actually goes in the box. I've seen the system configured perfectly and the SMS sent — but the packer never put the card in, because nobody told the pack station this order was a split.

By Lin Xiao, WMS Implementation Consultant