How to Handle Split, Partial and Backordered Shipments Across Channels

Every multichannel seller hits the same wall eventually. An order comes in for three items, and for one reason or another you can't put all three in one box. Maybe two are in your Warsaw warehouse, and the third is in Berlin. Maybe one is simply out of stock. Maybe none of them are in stock, but you took the order anyway because your supplier delivers Thursday.

Each of those situations has a name, a different fulfillment path, and a different message that has to go back to Amazon, eBay, Shopify, or bol.com. Get that message wrong, and the marketplace thinks the order is finished when it isn't.

This guide walks through all three scenarios, explains the logic that decides which one you're in, and gives you a workflow you can actually implement.

The three scenarios, in plain terms

Let's use one order for three items and see what happens under different stock conditions.

Split shipment. All three items exist, but they're not in the same place. Two sit in Warsaw, one sits in Berlin. You can't physically combine them, so you send two parcels with two tracking numbers. The customer gets everything they paid for, just in two deliveries, possibly on different days.

Partial shipment. Only two of the three items are available anywhere in your network. You ship those two now rather than making the customer wait, and the third stays open as an outstanding balance on the order.

Backorder. None of the three are in stock. You accept the order anyway, reserve the incoming units against it, and give the customer a ship date based on when your supplier delivers.

The difference between a split and a partial trips up a lot of teams, so it's worth stating clearly: a split shipment delivers everything the customer ordered, just in multiple parcels. A partial shipment delivers less than what they ordered, with the rest still owed.

Here's how the three compare across the things that matter operationally:

What actually decides which scenario you're in?

The deciding factor is a number called available-to-promise, usually shortened to ATP. It's the stock you can genuinely commit to a customer right now, which is not the same as the stock sitting on your shelves.

The calculation looks like this:

ATP = on-hand stock − already allocated stock − safety stock + confirmed inbound units

Say you have 40 units of a product on the shelf. Twelve are already reserved for orders you haven't shipped yet. You hold 8 back as a buffer. Your ATP is 20 units, not 40.

Now an order arrives for 25 units. You're 5 short at that location. Your system checks the second warehouse. If it finds 5 there, you get a split shipment. If it finds nothing, you ship 20 now and backorder the remaining 5. That's a partial shipment with a backordered line attached.

This is why ATP has to be calculated per product, per location, and why it has to be current. Stock figures that are ten minutes stale will happily commit units that were sold on another channel nine minutes ago. Our guide to syncing inventory across Amazon, eBay and Shopify covers how tight that sync interval needs to be.

Why do splits happen more than you'd expect?

Splits aren't a sign that something's broken. They're usually a side effect of decisions you made deliberately.

Holding stock in three countries means faster delivery promises, and it also means more orders will draw from two locations at once. Oversized items often can't share a parcel with anything else. Dangerous goods have packaging restrictions. Dropship lines come straight from a supplier, so any order mixing dropship and warehouse stock splits automatically. Carrier weight limits do the rest.

The question isn't how to eliminate splits. It's how to avoid the ones you didn't need: orders that could have shipped complete from one warehouse but got split because your allocation rules ran in the wrong order.

How allocation rules work (and why their order matters)

When an order arrives, your order management system runs through a ranked list of rules to decide which warehouse fills which line. The ranking is the whole game.

• Single-source priority tries to fill the entire order from one location before considering a split. This keeps parcel count and packaging costs down.
• Nearest node picks whichever warehouse is closest to the delivery address. Faster transit, more splits.
• Highest availability first pulls from wherever you have the deepest stock, which protects you against stock-outs on the next order.
• Lowest shipping cost picks the cheapest carrier route.

Dropship last only reaches out to suppliers when your own warehouses can't cover the line.

Here's the practical consequence. A system that checks "nearest node" before "single-source priority" will split orders that one warehouse could have filled completely, because proximity won the tiebreak. You pay for a second parcel to save a few hours of transit time. Sometimes that's worth it. Often it isn't.

Most teams never look at this sequence. It's usually left on whatever the default was during setup. Reviewing it is one of the cheapest wins available. Automating order management across multiple marketplaces goes deeper into how these rule sets get configured.

Safety stock: the buffer that prevents most backorders

Safety stock is inventory you deliberately keep out of ATP. It absorbs two kinds of surprise: a demand spike you didn't forecast, and a supplier who delivers late.

The size of that buffer should reflect how unpredictable each product is. A product with a 14-day lead time and volatile demand needs a much bigger cushion than one with a 3-day lead time and steady weekly sales. Applying the same buffer to everything means you're simultaneously over-protecting your reliable products and under-protecting your risky ones. The safety stock formula breaks down the inputs.

There's a real cost to getting this wrong in the other direction. Every unit held in reserve is a unit you can't sell, and that shows up in your inventory turnover and your cash position. Bigger buffers mean fewer backorders and more capital sitting still.

Telling the channel what happened

This is where most of the damage gets done, and it's almost always the same mistake.

Amazon, eBay, Allegro and bol.com all accept multiple tracking numbers against a single order. What they need from you is line-level detail: which line item, how many units of it shipped, which carrier, which tracking number.

Confirm the order at order level instead, with a single "shipped" flag, and the marketplace marks the whole thing complete. The undelivered item vanishes from the order record. Nobody chases it. The customer waits, then contacts you, then leaves a review about it. You find out about the problem several days after you caused it.

That's the failure mode to design against. Every partial dispatch needs line-level confirmation and a customer message that explains what's coming and when. Handled manually, this is tedious and easy to skip under pressure, which is exactly why it's worth automating. AI agents for order management cover how those notifications get triggered automatically.

Managing backorders without losing the customer

A backorder is a promise. The customer has paid, and they're waiting on your supply chain. Five things keep that promise credible:

1. Reserve against the purchase order. The incoming units should be earmarked for that specific order, not thrown into general stock when they arrive.
2. Give a real ship date. Derived from supplier lead time, not a guess.
3. Update the channel. The listing should reflect that this product ships later, not immediately.
4. Recalculate on every goods receipt. Partial deliveries from suppliers are normal. If you receive 60% of what you ordered, release backorders in the order they came in.
5. Escalate before the date slips. A purchase order confirmed for day 12 that moves to day 19 needs a revised customer message before day 12 arrives, not after.

That last one matters more than it looks. Customers accept delays reasonably well when they hear about them in advance. They cancel when they find out by discovering that nothing has arrived. An AI purchasing agent can watch supplier confirmations against open backorders and flag the slip before it becomes a support ticket.

Do splits cost you money?

Yes, every extra parcel adds a carrier base rate, packaging materials, and another pick-pack cycle in the warehouse. Splitting an order in two carries roughly two shipping charges against one order's revenue.

Free-shipping thresholds make it sharper. An order that qualifies for free delivery generates two carrier charges while you absorb the shipping entirely.

The key detail is where you can still act. A split prevented at the allocation stage costs nothing. A split discovered after picking has started costs the full second parcel. Cost control lives in your allocation rules, not in your warehouse. If you're building a business case, multichannel automation cost analysis puts these per-parcel numbers alongside platform spend.

Metrics worth tracking

Six numbers tell you whether your setup is working:

Read split rate and perfect order rate together. Rising splits with a stable perfect order rate means your distributed fulfillment is working as designed. Rising splits with a falling perfect order rate means allocation is failing and customers are feeling it. Measuring automation ROI shows how to put a cost figure against these.

The workflow, start to finish

Pulling it together, here's the sequence that handles all three scenarios across every channel you sell on:

1. Capture the order with full line-level detail from the channel.
2. Calculate ATP per product, per location.
3. Run allocation rules in a deliberate order, starting with single-source priority.
4. Reserve units and assign a source warehouse to every line.
5. Release pick instructions, one pick list per parcel on a split.
6. Post line-level dispatch confirmation and tracking back to the originating channel.
7. Recalculate open backorders every time goods arrive.

Step 6 is the one that breaks. It's also the one nobody notices breaking until customers start writing in.

Where a platform fits

Handling this manually works up to a point. Past a few hundred orders a month across several channels, the arithmetic stops being the problem, and the coordination starts being the problem.

ExactFlow runs ATP and allocation across your connected marketplaces, couriers and warehouses from one place. Orders from Amazon, eBay, Shopify, bol.com, Back Market and Refurbed land in a single queue, and stock updates flow back to every channel after each dispatch. Courier connections to DHL, DPD and FedEx generate one label per parcel when an order splits, and the AI operational agent watches for allocation exceptions and aging backorders without anyone checking a dashboard.

If you're comparing options, our guide to choosing an order management system covers what to look for, and plan details are here.