A few weeks after an inventory sync goes live, the same question tends to surface: is this actually working? The sync logs say success, nothing dramatic has gone wrong, and yet if somebody asked whether the numbers truly line up, you could not answer yes with any confidence. That discomfort is not a sign that your team is not paying attention. It is a sign that nobody has yet decided how the health of the sync gets measured.
In this article we will walk through four numbers that tell you whether your Shopify inventory sync is working, how to collect them without any analytics infrastructure, and how to turn what you collect into action. There is no complicated aggregation involved. Everything here fits inside one spreadsheet tab and a few dozen minutes a month.
Why inventory sync is something you cannot judge by feel
Before getting into measurement, it is worth being clear about why measurement is needed at all. Inventory sync has a failure mode all of its own, quite unlike most systems. Once you see that shape, measuring stops feeling like dull admin and starts feeling like a way of protecting your stock.
A broken sync never announces itself
Sync trouble does not arrive in the obvious form of a server going down. In most cases the sync itself completed perfectly well again today. It is just that one SKU dropped out of scope, or one location got pointed somewhere else, or a single row went missing from the sheet. Small discrepancies like that leave no error behind and trigger no notification. The numbers simply begin to drift.
Worse, the drift grows on its own. A SKU that falls out of sync scope still holds a correct number at the moment it falls out. The divergence begins afterwards, as goods arrive and ship. So the later you notice, the wider the gap, and by the time somebody spots it you can no longer trace when it started. That is the genuinely awkward part of inventory sync.
The cost of running stock on probably fine
Without measurement, stock gets managed on a feeling of probably fine — and that feeling misses in both directions. Believe things are fine when they are drifting and you oversell. Conclude instead that the numbers can never be trusted and you pad safety stock, carrying lost sales and excess inventory at the same time.
The tricky thing about running on instinct is that the cost never lands anywhere visible. Handling a cancellation after an oversell gets dealt with as a one-off and never gets counted. The decision to hold a little extra stock accumulates quietly as somebody's discretion. Added up it can be a substantial figure, yet no ledger anywhere carries a line item called the cost of not trusting our inventory sync.
Define what good looks like before you measure
The first thing to do when starting to measure is not to design a calculation but to put into words what good looks like for you. Should stock accuracy be 100 percent? Almost certainly not. If you run physical stores there is shrinkage and damage; in a warehouse things get put on the wrong shelf. Measure without a realistic benchmark and every number looks like a failure, at which point people quietly stop measuring rather than start improving.
Your definition of good can look completely different from anyone else's, and that is fine. It might be that the top 100 selling SKUs agree between sheet and Shopify at least 98 percent of the time, that oversells stay at one per month or fewer, or that a counted change appears on the storefront by the following business morning at the latest. The specific figures matter less than the fact that you chose them.
Four numbers worth tracking
There is plenty you could measure about an inventory sync, but four things genuinely earn their keep in practice. What matters is that two of them are outcomes (stock accuracy and oversell rate) and two sit closer to causes (time to reflect and manual corrections). Watch only the outcomes and you find out after things have already gone wrong.
Stock accuracy: how often the sheet and Shopify agree
Stock accuracy is the share of sampled SKUs where the on-hand quantity in your spreadsheet matches the on-hand quantity in Shopify. Check fifty and find forty-eight agreeing, and you are at 96 percent. The number you must compare is on hand, never available. Available is affected by unfulfilled orders, so it can legitimately differ even when the sync is doing everything right.
Go one level deeper and record the direction of each mismatch, because that pays off later. Is Shopify higher or lower than the sheet? A bias towards higher leads to overselling; a bias towards lower means goods you could have sold sitting invisible. The same 96 percent points at completely different priorities depending on which way it leans.
Oversell rate: orders taken for stock that was not there
An oversell is an order accepted against stock you did not actually have. Because it ends with asking a customer to accept a cancellation or a delay, it is the form of drift that converts into cost most directly. You can start counting it from cancellation reasons and from the records where fulfilment sent an order back saying the stock was not there.
As a rate, divide by total orders in the period. While the counts are small, though, do not fixate on percentages — raw numbers match people's intuition better. Going from one a month to two lands far more forcefully than going from 0.05 percent to 0.1 percent.
Whenever an oversell happens, make sure the SKU and the location get written down. Looking back, it is remarkably common to find them clustered at one site or within one product group, and that clustering is exactly where improvement begins.
Time to reflect: from counting a change to seeing it live
Time to reflect is how long it takes from recounting stock on the floor and writing it into the sheet, to that figure appearing on the online store. It is not determined by your sync interval alone. In practice three stages stack up: counting on the shelf, entering into the sheet, and the sync run itself. In most operations the longest of the three is one of the first two.
Measuring it can be crude and still be useful. Note the date and time of a stocktake and the date and time you could confirm the change in the admin, for a handful of cases, and look at the gap. If that exercise reveals that the sync runs hourly yet changes take a full day to appear, you have learned that the issue is not the schedule at all — it is how the entry flows.
Manual corrections: your earliest warning
The fourth is how often somebody fixes an inventory number by hand in the admin. It looks mundane, but it is the metric that flags a design problem soonest. When the system is working, people rarely need to touch quantities themselves. A rise in hand-fixing means the system has fallen behind the reality of the floor.
Recording corrections needs no special machinery — one extra row in a shared sheet will do. Date, SKU, location, and why it was fixed. Those four fields alone reveal a clear pattern once two months have accumulated. The trick with the reason column is to let people write the sentence they would say out loud: it was not in the sync scope, the row had been deleted from the sheet.
- Stock accuracy: share of sampled SKUs where the sheet and Shopify agree on on-hand quantity
- Oversell rate: share of orders in the period accepted against stock that was not actually there
- Time to reflect: elapsed time from recounting on the floor to the storefront figure changing
- Manual corrections: how often inventory is fixed by hand in the admin — a rise signals a design problem
Collecting the numbers without building a reporting project
Most measurement efforts collapse because building the collection mechanism turns into something enormous. The moment somebody proposes a dashboard, this has become a separate project. Here are three ideas for gathering the numbers within what you can genuinely start today.
Sample; do not count everything
Hearing measure stock accuracy, you might picture reconciling every SKU against Shopify. With a few thousand products that is neither practical nor necessary. Without getting statistically rigorous about it, sampling thirty to fifty items a month is quite enough to see the trend.
There is a small knack to how you sample. Deliberately mixing in the cases you want to learn about yields more information than picking purely at random. Take a few from your best sellers, a few from products stocked at multiple sites, a few from a location you added recently, and fill the rest randomly. That way you look hardest where problems are likely while still picking up the overall picture.
- 01Fix a review day each month and choose thirty to fifty SKUs to check
- 02Deliberately mix in best sellers, multi-location products and items at newly added sites, then fill the rest at random
- 03Compare the sheet's on-hand figure against the per-location on-hand figure in the admin, one by one
- 04Record how many matched, and for each mismatch the SKU, the location and the direction of the gap
Start from the sync logs you already have
Before you start recording anything new, look at the logs your sync runs already leave behind. When it ran, how many rows were in scope, whether it succeeded. That alone is a surprisingly good starting point. The row count in particular is worth watching: if last month was 1,200 and this month is 1,150, then fifty rows have dropped out of scope.
The knack when reading logs is to care less about the success flag and more about the difference from last time. A sync can succeed while its scope shrinks, and the drift widens all the same. Once a month, open the recent runs and note nothing more than the counts and the run times. It takes five minutes and works remarkably well as a net for catching quiet failures. What to keep and how to read it is in sync logs as an audit trail.
One owner and one cadence per number
The single biggest reason measurement stops is that nobody owns it. Let us all keep an eye on it is functionally identical to nobody looking at all. For each of the four numbers, name one person and one moment. In a small team that can be the same person four times over. What matters is that a name is written down.
Match the cadence to the nature of each number. Stock accuracy needs one sample a month. Oversells are naturally logged as they happen and counted at month end. Time to reflect is fine measured once a quarter across a handful of cases. Manual corrections are the exception: they have to be written at the moment they occur, so keep the record somewhere the floor can reach easily.
Turning the numbers into action
Numbers produce nothing at the moment they are collected. Measurement becomes work only once it is settled what you do when you see a given value. Here are three principles for converting the four numbers into real improvement.
Set thresholds that trigger action, not aspirational targets
We are aiming for 99 percent stock accuracy sounds admirable and achieves almost nothing, because nobody has decided what happens at 98. What you want instead is a threshold: cross it and a specific action follows. A target points a direction; a threshold pulls a trigger.
Always write the matching action alongside the threshold. If stock accuracy falls below 95 percent, review the sync configuration for every mismatched SKU. If oversells exceed two in a month, revisit the settings for the location involved. A threshold with no action attached simply reverts to being a number. Monitoring and alerts for inventory sync shows how to be told when a threshold breaks.
- Stock accuracy drops below your line: check the sync scope and location mapping for every mismatched SKU
- Oversells exceed your allowance: investigate the location involved and its time to reflect first
- Time to reflect is lengthening: review both the sync schedule interval and when entries reach the sheet
- Manual corrections are rising: read the stated reasons together and ask what the system could absorb instead
Problems cluster — fix the worst SKUs first
Keep a record of discrepancies and one thing becomes apparent almost without exception: problems are not spread evenly across the catalogue, they gather in particular places. One location, one supplier's product group, a handful of SKUs stocked at multiple sites. It is common for a few items at the top of the list to account for a large share of all the drift.
That property is good news for how you improve. You do not need to lift everything a little; you identify the worst few and fix those carefully. Usually the cause is shared — the same location misconfiguration, the same misuse of a sheet column, the same habit in one person's routine. Fix one and every SKU hanging off it improves at once.
So when you record, always keep the SKU and the location. Three mismatches this month hides the cluster entirely. Three mismatches at the East Japan warehouse is the point at which the next move becomes obvious.
Settle into a review rhythm
Finally, make measurement and improvement a repeating rhythm rather than a one-off event. A two-tier arrangement works well: monthly and quarterly. The monthly pass simply lines up the four numbers and checks whether any crossed a threshold. It is a check, not a debate, and should be finished in ten to fifteen minutes.
The quarterly pass puts three months side by side and looks at the trend. Even when nothing has deteriorated, this is where you notice manual corrections creeping upward or time to reflect stretching out. Use the same session to ask whether the thresholds themselves still fit reality. One that has not been crossed once in three months is probably too generous.
Whether your inventory sync is working ought to be something you know rather than something you sense. Stock accuracy, oversell rate, time to reflect, manual corrections. Keep those four to hand and the next time stock comes up in a meeting you can answer with facts instead of impressions. More importantly, you notice the moment they start to slide — and against a system that fails quietly, there is no better preparation than that. The reconciliation itself is covered in reconciling physical counts warehouse by warehouse.