Inventory velocity is the foundation of every restock decision, every demand forecast, every stockout prevention workflow. It is also one of the most commonly miscalculated numbers in Amazon analytics. Average it the wrong way and your restocks lag demand. Cut the wrong window and you over-buy slow movers.

This guide is about calculating inventory velocity properly.

TL;DR: Inventory velocity is units sold per day per SKU per marketplace. The right window depends on the use case — 7-day for daily ops, 30-day for restock planning, 90-day for category trends. Common mistakes include treating returns as sales, ignoring stockouts (which artificially lower velocity), and using marketplace-aggregate velocity for region-specific decisions. Properly calculated velocity is one query against your data layer, not a spreadsheet exercise.

The basic formula

Velocity = Units sold ÷ Days in window

That is it for the math. The complexity is in the inputs.

Choosing the right window

7-day velocity

Best for daily operations and short-term alerts. Reactive to recent demand spikes. Noisy on slow-moving products.

30-day velocity

The standard for restock planning. Smooths out weekly variation. Good baseline for most decisions.

90-day velocity

Useful for category trend analysis and quarterly planning. Less sensitive to recent demand shifts.

Trailing weighted velocity

Combines multiple windows with weights — typically last 7 days more heavily than older periods. Better for products with shifting demand patterns.

The four most common mistakes

Counting returns as sales

If you count gross orders without subtracting returns, your velocity is inflated. The fix: use net units shipped, or units sold minus units returned within the window.

Ignoring stockouts

If a SKU was out of stock for 5 days in your 30-day window, the velocity number understates real demand. The fix: calculate velocity over in-stock days only, or flag stockout periods and adjust.

Aggregating across marketplaces

Pan-EU SKUs ship from multiple warehouses. Calculating one global velocity hides the country-level demand differences that drive Pan-EU restock decisions.

Using gross sale price

Velocity is a unit metric, not a revenue metric. Mixing them confuses the math. Track units separately from revenue.

Where the data lives

• Units sold — Order detail reports per SKU per day.
• Returns — FBA Returns report or Order Return reports.
• Stockouts — Inventory snapshots over time, joined to days when stock was zero.
• Marketplace — Order detail includes marketplace ID for splitting.

Joining these into a clean velocity table per SKU per marketplace per period is the data engineering job. Once it is done, every restock workflow benefits.

How AI builders use velocity

With clean velocity data, AI builders ship restock recommenders, alert systems and forecasting tools as conversational builds. A typical prompt:

“Build me a daily report listing every SKU where 7-day velocity is more than 50% higher than 30-day velocity, sorted by units. Post to #ops-amazon at 9am.”

The AI builds the query, the schedule, the formatting. Your team gets the alert.

The bottom line

Velocity is simple math on top of clean data. Most teams stumble on the data part. Once it is in your data layer correctly, every downstream workflow — restock planning, stockout prevention, demand forecasting — inherits the right number.

DataDoe’s Amazon data layer exposes per-SKU per-marketplace velocity over multiple windows, ready for AI tools to query.