Why do my sales channels disagree on how much stock I have?

Ask what each number is measuring

A stock number without a definition is an invitation to confusion. Suppose a fictional homewares store has 24 lamps on a shelf, four allocated to orders, two damaged, and one held for inspection. Its physical count is 24, but that does not mean 24 customers can place new orders. A channel displaying 17 might be working exactly as intended. Another channel displaying 24 may be publishing the wrong state.

Shopify distinguishes inventory states including on hand, available, committed, unavailable, and incoming. Use the terminology in your own systems rather than assuming that similarly named columns mean the same thing. Put a short definition beside each number in your reconciliation worksheet. Your first task is to compare equivalent quantities, not to force every screen to display identical figures.

Match one variant at one location

Start with a narrow example. Choose one sellable variant, such as the sage lamp with a particular plug type, and one shipping location. Record its internal SKU, platform variant identifier, marketplace listing reference, and unit of measure. A parent product containing several colors is not a useful comparison point if each channel sells the colors separately.

Location differences are equally easy to miss. One tool may total all warehouses, while a marketplace offer draws from a single fulfillment location. Inventory in a showroom may be physically present but excluded from online orders. A transfer can also appear in two places during its journey if you compare incoming stock at the destination with on-hand stock at the origin. Make the location boundary explicit before investigating the integration.

For example, a retailer might hold twelve blue lamps in its main warehouse and four in a showroom that does not ship online. A channel showing twelve could be correct even though the stock report totals sixteen. If two warehouse units are already committed, ten may be the appropriate sellable quantity. Write those three facts separately before changing anything. The exercise often reveals that a disagreement is a difference in scope, not a failed update.

Build a tiny timeline instead of taking more screenshots

Screenshots show what a page displayed; they rarely explain how it got there. Build a sequence with the last known correct quantity, the sale or adjustment that changed it, the time the central record changed, the time an update was sent, and the time the channel acknowledged it. Keep the time zone visible. A five-minute difference can look like a missing order when two interfaces use different clocks.

For example, a sale at 10:03 reduces available stock from 17 to 16. The central workspace records the change at 10:03, and a channel still shows 17 at 10:04. That could be a normal delay within your measured operating window. If the channel shows 17 after repeated successful acknowledgments, investigate mapping or competing writers instead. The sequence directs your next action; another refresh does not.

Look for the second tool holding the pen

Inventory problems often survive repeated corrections because more than one system is allowed to overwrite the quantity. A warehouse application sends 16, a scheduled spreadsheet import sends yesterday's 20, and a marketplace connector later sends 15. Each job can report success while the business sees an apparently random result. The issue is ownership, not simply speed.

List every process that can change inventory: receiving, point-of-sale sales, manual adjustments, bundles, returns, transfer receipts, bulk imports, and installed applications. Assign an authoritative writer for each kind of change. That does not require one tool to do everything. It requires a clear rule for where an adjustment starts and how the other systems learn about it. Remove redundant schedules only after you understand the work they currently perform.

Reconcile the movement, not just the ending balance

A useful reconciliation begins with a trusted opening quantity and accounts for movements. Add received stock and approved restocks. Subtract sales, damage, transfers out, and other adjustments according to your stock model. Keep reservations separate if your system already subtracts them from available inventory. Otherwise, the same order can be deducted twice in your spreadsheet.

When the math does not close, investigate the unexplained movement. Was a return received without inspection? Did someone edit a bundle component? Was an order canceled in one system but still reserved in another? A manual correction can restore today's count while hiding the underlying cause. Record why you made it, which evidence supported it, and whether the same defect could affect other products.

Contain the risk while you investigate

Treat a discrepancy differently depending on what remains for sale. A difference of two units on a slow-selling item with 400 available is not the same operational risk as a difference of two on your last three units. Consider temporarily reducing the exposed quantity for the affected listing or removing an uncertain location from its allocation while the count is checked.

Avoid making broad catalog changes in response to a narrow problem. A storewide quantity reset can spread a bad mapping much further than the original discrepancy. Work from a small, documented sample and verify both the outgoing update and the receiving channel's view. If you need to pause a process, record the pause time and the orders that arrived during it so the recovery has a defined boundary.

Turn the investigation into a repeatable check

Once the example is resolved, classify the cause: definition, location, timing, mapping, missing event, or competing update. These categories make a weekly review more useful than a folder of screenshots. A cluster of mapping issues suggests catalog cleanup. Growing delays suggest synchronization work. Frequent receiving errors suggest a warehouse process change.

Track the number of affected SKUs, the age of unresolved differences, and the number of customer orders disrupted. Set a review owner and a realistic threshold for intervention. The goal is not an impossible world in which every system changes at precisely the same instant. It is a business that knows which quantity is safe to sell, notices exceptions quickly, and can explain how a correction was made.

Use a shared discrepancy record

Give each unresolved discrepancy a short record that another person can understand without a meeting. Include the item, location, expected quantity, observed quantity, last verified time, owner, and next check. Attach links to the relevant source records instead of copying unnecessary customer information. A clear record prevents two colleagues from making different corrections to the same item.

Set an explicit closing condition. For example, the physical count is confirmed, the central quantity is corrected with a reason, and each affected channel has acknowledged the intended amount. A ticket is not resolved merely because the number looks better in one tab. If a later order changes the quantity again, compare the new movement against the corrected baseline. This turns reconciliation into controlled work rather than a contest between whoever last refreshed their screen.