Shopify Inventory Sync Across Online and Retail

Merchants moving to Shopify POS expect inventory sync to be the hard part and are usually surprised that it is not. Retail and online share one pool — a sale in the shop decrements the same stock the website sells from, with nothing to synchronise.

What replaces the sync problem is an allocation problem, and that one is a business decision rather than a setting. Two channels want the same unit. Which gets it?

Which locations fulfil online orders

The first decision, and it has consequences on the shop floor.

Enabling a retail location for online fulfilment makes its stock sellable online. That is often desirable — you have inventory, someone wants it, and shipping from the nearest location is faster.

It also means someone in the shop picking and packing during trading hours, competing with walk-in customers for both stock and attention. And it means the last unit on the shelf can be sold to someone in another city while a customer is holding it.

The options in practice:

Warehouse only. The simplest arrangement, and it leaves retail stock entirely untouched. It also wastes availability if your shops hold significant inventory.

Warehouse plus selected shops. The usual compromise. Larger or quieter locations participate; busy flagship stores do not.

Everything enabled, with buffers. Maximum availability, and it requires the buffer discipline described below to be workable.

Safety buffers are not optional at retail

A buffer is a quantity held back at a location so that channel does not sell out from under the other.

Without one, the failure is specific and bad: a customer in your shop is holding the last unit, an online order takes it, and either the online customer is disappointed or the walk-in customer is. Both outcomes are avoidable.

The number is a judgement. It should reflect how fast that location sells the item and how quickly you can replenish. Fast-moving items in busy shops need more; slow-moving items in quiet locations need little.

The fifth item is the one most people miss. If every unrestricted location is out and only a buffered one has stock, the system needs a rule — hold the order, notify someone, or release the buffer. Deciding it in advance beats discovering it during a busy week.

Promotions consume stock unpredictably

The most common source of unexpected stockouts, and it is worth treating separately because ordinary forecasting does not describe it.

Free gifts are consumed by demand for qualifying products, which may be most of your catalogue. If the threshold is easy to hit, every large basket is effectively ordering one.

Bundle components deplete from two directions when the same product is also sold separately, and neither channel's forecast accounts for the other.

Buy X get Y drains the Y product at a rate governed by demand for X, and Y is frequently the item nobody is watching precisely because it was chosen for being slow.

Before any promotion, set alerts on every gift, component and Y product, and decide in advance what the offer does when one runs out — stops applying, substitutes, or oversells. All three are defensible; discovering which one your setup does mid-campaign is not.

Continue-selling is a per-product decision

Allowing sales at zero stock is right for made-to-order items and where inbound stock is reliable. It is wrong for most physical products.

The setting is harmless all year and dangerous during any high-volume period, because volume that would normally arrive over weeks arrives in hours. A product that would have oversold by three units across a fortnight oversells by three hundred over a weekend.

Review it product by product before any peak rather than as a global setting, and pay particular attention to anything a supplier has recently been slow on — the products where continue-selling feels safest are usually the ones where your own inbound assumptions are most out of date.

Stock held during an abandoned checkout

A pattern that confuses staff more than any other, and it is worth explaining plainly because the confusion produces bad decisions rather than just curiosity.

When a shopper reaches checkout, Shopify reserves the items in their cart for a short window so that two customers cannot both be sold the last unit while one of them is still typing their address. That reservation is temporary and releases automatically if the checkout is abandoned, but for the duration it holds, the stock appears unavailable to everyone else, including a shop assistant looking at the same product on a location enabled for retail.

The confusion this causes on the shop floor is specific: a staff member sees a product showing zero available, a customer is standing in front of a shelf with three of that product visible, and the staff member concludes the inventory system is wrong. Usually it is not. Several online checkouts may be holding the remaining units temporarily, and the count will correct itself within the reservation window once those checkouts complete or expire.

This matters most during high-traffic periods, when the number of concurrent checkouts against a popular item rises sharply and the reservation effect becomes visible rather than a rare edge case. A product with real stock on the shelf can show as unavailable online, or appear unavailable to staff, for the several minutes it takes abandoned reservations to expire, and repeated during a sale this can look like a sync failure when it is the reservation system working as designed.

The practical response is not a setting to change but a piece of staff training worth having before it comes up during a busy weekend: if a product shows unavailable but is visibly on the shelf, the likely explanation is a held online reservation rather than a data error, and it is usually worth waiting a few minutes and checking again before escalating it as a fault. Where this happens often enough to disrupt retail service on a specific fast-moving item, that is itself useful information, and it belongs in the buffer conversation above rather than being treated as a bug to be reported each time.

Transfers and the in-transit gap

Moving stock between locations creates a window where units are neither where they were nor where they are going, and how your system represents that window matters.

Two failure modes: stock counted at both ends briefly, which oversells; and stock counted at neither, which suppresses sales you could have made.

Get the process right rather than the setting. Transfers should be recorded when they leave and received when they arrive, by people who know that is their job. Most transfer problems are process problems wearing a technical costume, and the tell is that they cluster around particular locations or particular people rather than particular products.

Multi-location changes what "in stock" means

With one location, a product is available or it is not. With several, availability becomes a question with a location attached, and that has consequences for what you show shoppers.

On the storefront, a shopper generally wants to know whether they can have the thing, not where it is. Showing aggregate availability is usually right. Showing per-location stock is useful only when the shopper cares — click-and-collect, or a category where visiting the shop is normal.

For click-and-collect, per-location accuracy suddenly matters a great deal. A customer who reserves for pickup at a shop that does not actually have the item has had a worse experience than one who was told it was unavailable. This is where buffer discipline and counting accuracy stop being back-office concerns and become customer-facing.

For staff, visibility across locations is the feature that changes daily work. A customer asking for a size the shop does not have can be told, immediately, that another location has one — and that conversation converts a lost sale surprisingly often.

For fulfilment, priority order matters as much as eligibility. Shipping from the nearest location is faster and cheaper, but repeatedly drawing from one shop while others sit on stock creates an imbalance that nobody notices until that shop is empty and the others are overstocked.

Worth reviewing periodically rather than setting once: fulfilment patterns drift as your traffic and your locations change, and a priority order that made sense at two locations often does not at five.

Drift is a process problem

Inventory counts diverge from reality on every retail operation, and it is worth being clear that no software prevents this.

The causes are consistent: shrinkage, receiving errors, returns refunded but never restocked, damaged goods removed from the shelf but not the system, and transfers that happened physically before they happened in software.

Cycle counting — checking a subset regularly rather than everything annually — catches drift while it is small. Annual stocktakes find a year of accumulated error at once, by which point the cause is unknowable and the only available response is to write it off and start again.

A workable cycle count is less onerous than it sounds. Count your fastest-moving and highest-value items monthly, the middle of your catalogue quarterly, and the long tail annually. That concentrates effort where error is both most likely and most expensive, and it turns stocktaking from an annual shutdown into a routine task that takes an hour a week. The by-product is that when a count is wrong, you find out close enough to the cause to actually fix the process that produced it.

For the wider platform picture, see POS and operations. Promotional mechanics and their inventory consequences are covered in promotions and offers, and peak-period planning in the BFCM 2026 series.