All guides

InventoryAugust 21, 20266 min read

Common Shopify Inventory Forecasting Mistakes

The recurring mistakes that quietly wreck a Shopify store's inventory forecast — from stale data windows to ignoring lead time changes entirely.

Most forecasting mistakes aren't a bad formula. They're a good formula fed a number that was already wrong before the math started. This is a checklist, not an essay — each item names the mistake and links to the post that teaches the actual fix.

One of these deserves more than a line before the list starts, because it's the one merchants live through repeatedly without ever having a name for it.

A stockout doesn't just cost the sales you missed that week. It teaches your forecast to expect less. If a SKU sells out and stays out for three days, your sales history shows three days of near-zero demand — not because customers stopped wanting it, but because there was nothing on the shelf to buy. A forecast built on that history reads those three flat days as real, orders a little too little next cycle, runs out again a little sooner, and repeats. Researchers call this censored demand: the sales record shows what you were able to sell, not what people actually wanted, and a forecast that treats the two as the same thing will quietly, systematically under-order the exact SKUs that keep going out of stock.

A stockout doesn't just cost this week's sales — it teaches your forecast to expect less, permanently, until someone catches it.

Why each stockout produces a smaller restock than the lastA line chart of one SKU’s stock level across three repeated stockout cycles. In each cycle the recorded stock level, drawn as a solid emerald line, depletes in a straight run down to zero, then sits flat on zero for a visible stockout window while there is nothing left to sell. During that flat window a dashed brass line continues below the zero line, representing the demand that still existed but could not be recorded, because a sale needs stock and there was none. When each stockout ends, stock jumps back up at a restock — but each restock lands lower than the one before it, so the highest point of the line drops visibly after every cycle, following the thin dashed line connecting the three restock peaks. This is illustrative of a pattern, not a real dataset, so there are no numbers on either axis. The argument: a forecast built only from recorded sales reads each stockout’s flat, near-zero demand as if it were real, so it orders a little less next time, produces a smaller peak, reaches the next stockout a little sooner, and repeats — the system is learning the wrong lesson from its own failure.Stock level after repeated stockoutssame SKU, three cycles — each restock lands lower than the lastStockoutagainagainRestocksmaller reorderRestocksmaller stilleach restock lands lower than the lastcycle repeatsstock level — what’s on the shelf, and what gets recordeddemand that existed but couldn’t be recorded — nothing was there to sell
None of the three restocks in this chart look wrong in isolation — each one matches what the sales data showed. The data itself was already lying, one stockout at a time.

Data Mistakes

  • Treating stockout days as normal low demand. A sold-out SKU's sales history understates real demand, biasing every future forecast downward and making the next stockout more likely, not less. Building the habit of catching this belongs in measuring whether your forecast is actually accurate.
  • Averaging a full year instead of a recent window. A stale annual average tells you nothing about a velocity that doubled last quarter. See the reorder-point post's own take on recalculating from a recent 60–90 day window.
  • Skipping the outlier cleanup before forecasting. A single bulk or wholesale order left in the data inflates every average built from it. Cleaning that out is covered in using your own Shopify sales data.
  • Applying an established product's method to a SKU with no history. There's no velocity to average when a product has zero prior sales — a different set of methods applies. See forecasting a brand-new product.

Method Mistakes

  • Running a flat moving average on a trending or seasonal product. A plain average assumes no trend; it lags a genuine upward or downward shift and misses seasonal swings entirely. See when moving average breaks down.
  • Never comparing methods against your own catalog's pattern. The method that fits a steady seller is often the wrong one for a spiky or seasonal SKU. Start with comparing forecasting methods.

Process Mistakes

  • Setting the forecast to match a sales goal instead of the likely outcome. A forecast quietly adjusted to hit a target stops functioning as an early-warning signal and leads to overstocking against reality. Covered in improving forecast accuracy.
  • Publishing a single number with no range around it. A point forecast with no stated uncertainty gives no basis for sizing a safety stock buffer. See choosing your safety stock number.
  • Never checking the forecast against what actually sold. A forecast that's never reviewed against real results can run wrong for months before anyone notices. The habit worth building is in measuring forecast error.

External-Factor Mistakes

  • Not noticing a supplier's lead time quietly changed. A reorder point built on last year's 10-day lead time is wrong the moment it becomes 18. See how lead time affects your forecast.
  • Letting a promotion's spike distort the ongoing baseline. One unflagged promo week folded into normal sales history overstates "typical" demand for every future non-promo week. See how promotions affect forecasting.
  • Missing a genuinely seasonal pattern entirely. A flat average across a full year hides a swing that a seasonal index would have caught. See forecasting seasonal inventory.

A handful of these are structural, not effort problems — a stale velocity window, an unnoticed lead-time change, a stockout quietly biasing the next forecast down. Recalculating every SKU by hand, every week, catches these in theory; in practice it's the task that slips first once a catalog passes what one person can hold in their head. StockCue recalculates every SKU's forecast nightly from live order data, which removes several of these mistakes structurally instead of relying on someone remembering to check.

STOCKCUE

Stale windows and unnoticed lead-time changes are easy to miss by hand and hard to miss when a forecast recalculates itself every night. StockCue does that for every SKU, on a Free plan that covers your first 50.

Install StockCue on Shopify →

Devmerx

Devmerx is a Shopify and WordPress development agency helping DTC brands build faster stores, cleaner migrations, and higher-converting experiences. Based in London, UK, serving clients worldwide.

See our work

Need help with your Shopify store?

Devmerx builds and optimises Shopify stores for DTC brands. Book a free 20-minute consultation.

Book a free consultation