Operations

When the stock stops
agreeing with itself

This is usually the symptom that starts the conversation, and it is rarely the storefront's fault. Two channels disagree, someone reconciles by hand every morning, and the business oversells anyway.

Replatforming without addressing it moves the fault rather than clearing it, which is why the platform decision and the systems decision belong in the same conversation.

Why stock desynchronises

Three causes, in the order they are usually found. The first is by far the most common and the cheapest to fix.

No source of truth

Two or more systems each believe they own the stock figure and each will happily write to the other. Nothing is broken; there is simply no answer to "which one is right", so the last writer wins and the figures drift apart by the hour.

Latency at the wrong moment

The sync works and it is simply not fast enough for the sales rate. A fifteen-minute cycle is invisible on a slow catalogue and catastrophic on a product doing thirty units an hour during a promotion.

Reservation logic

Stock is decremented at the wrong point in the order lifecycle — at fulfilment rather than at checkout, or never released when a basket is abandoned. The count is arithmetically correct and operationally wrong.

What overselling actually costs

The refund is the smallest part of it, and it is the only part most businesses count.

What genuinely stops it is a single source of truth plus a buffer, not a faster sync. Faster syncs narrow the window; they do not close it, because two channels can always sell the last unit inside any window. A held buffer on fast-moving lines accepts a little unsold stock in exchange for never making that call, and for most operations that trade is obviously worth it once the cancellation cost is counted properly.

Count all four

The refund. The smallest of the four
Marketplace metrics. Cancellation and defect rates that throttle visibility, or suspend the account outright
Support time. Per incident, every incident
The customer. Told the order was confirmed, then told it was not

When a second location makes things worse

A second warehouse is usually opened to solve a space problem, and it introduces an allocation problem — which is a harder problem. Every order now needs a decision about where it ships from, every SKU can be in two places, and any line held in both locations can be sold twice unless the allocation logic is genuinely aware of both.

Worth saying: the fix here is frequently not software. Holding fast-moving lines at one site only, and accepting a slightly worse cube utilisation, removes the allocation problem entirely for a lot of operations.

The tell

A split-shipment rate nobody chose.

If orders are routinely going out in two parcels because the picking location was decided line by line rather than order by order, the business is paying twice the carriage to work around its own inventory model. It is measurable from the despatch data you already hold, and nobody has to buy anything to check it.

WMS or 3PL — how to work out which is cheaper

No published figure answers this, because 3PL pricing depends on the shape of the operation — order profile, unit sizes, storage footprint, seasonality — and in-house cost depends on premises a business may already be paying for. What can be given is the method, and the axes the two models genuinely differ on.

	In-house + WMS	3PL
What you actually pay for	Labour at peak, the fulfilment share of rent and rates, packaging, licence and implementation	A per-order and per-storage rate quoted against your real order profile
The line most under-counted	Implementation, and the management time that disappears into the warehouse	Surcharges outside the quoted profile — returns, oversize, peak
Who absorbs a peak	You do, in hours and in hiring	They do, at a rate agreed in advance
Where the risk sits	Operational, and yours to fix at 6am	Commercial — someone else's operation, still your customer
The wrong choice when	Volume is spiky and premises are already tight	The proposition depends on how the parcel arrives

Build the in-house cost honestly, using the first column. Businesses running their own fulfilment routinely under-count implementation and management time by a wide margin, and those two are what decide it.

Then get 3PL quotes on the real order profile rather than an average — the per-order economics of a five-line order and a one-line order are not close.

Compare at current volume and at twice current volume. The crossover usually sits between them, and the question that actually matters is which model you want to be in when the business is bigger.

The one non-financial factor worth weighting sits in the last two rows: 3PL makes fulfilment someone else's operational problem and your commercial risk.

What Shopify Plus does not solve

Plus is a strong storefront and checkout, with real multi-location inventory and a solid API. It will hold stock levels across locations and expose them to channels properly, and for a good number of merchants that is enough.

It is not a warehouse management system. It will not tell a picker the efficient route, will not direct putaway or replenish a pick face, and will not handle batch and expiry. It does now carry the purchasing record — purchase orders against supplier records, and receiving against a linked inventory transfer, in the admin since June 2026 — but that is a stock ledger, not a warehouse: there are no bin locations behind it, so nothing tells anyone where the pallet went. It has no opinion about your allocation policy beyond the rules you give it.

And a picking bottleneck is very often a process problem wearing a software costume: pick paths that follow the SKU list rather than the shelves, wave sizes set by habit, packing benches short of the packaging they need most. Those are worth exhausting before a licence is bought — they are free, and the improvement is measurable in a fortnight.

Diagnosis before purchase. A migration that fixes the storefront and leaves the warehouse untouched has moved the fault, not cleared it.

When the diagnosis does land on a purchase, the selection question — which system is nominated to be right, and the four numbers any candidate has to be able to show — is what a stock control system has to get right. The narrower version, for a merchant already on the platform, is what Shopify already covers and the four reasons to add to it.

When the stock stops agreeing with itself