Safety Stock for Amazon Sellers: How Much Is Enough?

This is a guest post from Jeremy Biron of Profit Hawk.

Two SKUs sell an average of 100 units a day. One ticks along steadily. The other has quiet stretches and sudden spikes. Give both the same 30-day buffer and you could end up overbuying one while still running short on the other.

Safety stock gives you room for surprises. The trick is deciding how much protection each SKU needs without letting “just in case” take over your cash.

Know what your safety stock is covering

Safety stock is extra inventory held above expected demand to cover unexpected demand or replenishment delays.

At 100 units a day and a 25-day replenishment lead time, 2,500 units cover expected sales. Safety stock is the buffer above that.

Planned promotions belong in your demand forecast. Supplier minimums constrain purchase size. Those extra units don’t automatically become safety stock.

For FBA, measure lead time from the action you’re deciding on until replacement units are sellable with your intended delivery promise. A supplier purchase uses the full PO-to-sellable journey; a transfer from a domestic warehouse uses the warehouse-to-sellable journey. A “delivered” status alone doesn’t tell you the stock can protect tomorrow’s sales.

ZonPrep’s Speed to Profit article explains how inbound delays tie up cash. Use your own shipment history to check whether the lead time in your plan matches what actually happens.

Use ABC tiers to decide what deserves protection

Group products by what a stockout would cost the business:

• A products: stockouts would seriously hurt contribution profit or a strategic priority, such as a new launch.
• B products: meaningful contributors whose interruptions are less costly.
• C products: lower-priority items where extra inventory may be harder to justify.

Revenue is a starting point. A big seller with thin margins may deserve less inventory investment than a smaller, more profitable SKU. Leave room for deliberate exceptions, including launches with little sales history.

You don’t need to force your catalog into neat percentages. Record why each SKU belongs in its tier and revisit it as sales and margins change.

Then look at demand variability. ABC tells you how much a stockout matters; variability tells you how much protection it takes to avoid one.

A steady A item may need close monitoring and a modest buffer. An erratic C item may be a candidate for a lower service target. “Erratic” doesn’t automatically mean “buy more.”

Measure the variation, not just the average

Start with a clean run of daily demand data that represents the period you are planning for. Eight to twelve weeks can be a useful first pass for a stable item, but it is not enough to understand an annual seasonal pattern.

Watch for two traps. Sales during stockouts understate demand. Promotions, price changes and seasonal shifts should not be mixed into a normal baseline without adjustment.

For a reasonably steady product, calculate:

• Average daily demand: total units divided by the number of days.
• Daily standard deviation: how widely daily demand varies around that average.

In a spreadsheet, use AVERAGE for the mean and STDEV.P for the variation in that cleaned historical range.

If average daily demand is 100 units and standard deviation is 30, use 100 to plan expected consumption and 30 to describe daily variation.

For seasonal products or a strong weekday pattern, forecast that pattern first. Then examine forecast errors, ideally across the full replenishment window. Simply feeding predictable peaks into a daily standard deviation can inflate the buffer.

Calculate a starting buffer

For a SKU with stable demand patterns and a reasonably fixed lead time, a useful starting model is:

Safety stock = service factor × daily standard deviation × square root of lead time in days

The service factor sets the protection target. A factor of 1.645 represents a 95% cycle-service level under this model: a 95% chance of getting through a replenishment cycle without stocking out. It doesn’t mean 95% of units filled.

Choose the target based on the cost of a stockout and the cost of carrying more stock. The 95% used here is an example, not a recommendation for every SKU.

Here’s a hypothetical supplier order, measured from PO placement until the replacement units are sellable on Amazon:

• Average daily demand: 100 units
• Daily standard deviation: 30 units
• Replenishment lead time: 25 days
• Service factor: 1.645

Safety stock = 1.645 × 30 × √25 = 246.75, rounded up to 247 units.

Expected lead-time demand is 100 × 25 = 2,500 units. Adding the buffer gives a reorder point of 2,747 units.

Compare that reorder point with your inventory position: usable on-hand stock plus relevant incoming supply, minus outstanding demand commitments. Count each unit once across locations, and check arrival dates. A purchase order arriving next month won’t rescue a stockout next week.

This model assumes independent daily variation, fixed lead time and approximately normally distributed total demand during lead time. It suits continuous inventory monitoring. For irregular demand or unreliable delivery times, test a more suitable model against your actual history.

If you only place orders weekly, allow for demand over lead time plus the seven-day review interval. That requires a periodic-review ordering policy; don’t apply this continuous-review reorder trigger unchanged.

Try these numbers in Profit Hawk’s free safety stock calculator, then check the result against your own replenishment history.

Shorter lead time changes both parts of the calculation

Now suppose that same PO-to-sellable lead time reliably falls from 25 days to 16. Demand variability and the service target stay the same.

Metric	Before	After
PO-to-sellable lead time	25 days	16 days
Expected lead-time demand	2,500 units	1,600 units
Safety stock	247 units	198 units
Reorder point	2,747 units	1,798 units

At 16 days, safety stock = 1.645 × 30 × √16 = 197.4, rounded up to 198 units. Of the 949-unit drop in the reorder point, 900 units come from lower expected lead-time demand and 49 from the smaller buffer. Lead time fell 36%; safety stock fell about 20%.

Shorter lead time also lets you react sooner to demand changes. Check the improvement across several shipments before reducing the buffer; a faster average can hide some very late deliveries.

ZonPrep’s guide to reducing Amazon FBA inbound times covers the operational side. Measure the effect on the full PO-to-sellable journey before changing your inventory settings.

Make the cash decision, then test it

At a hypothetical landed cost of $8 per unit, 49 fewer buffer units means $392 less inventory investment in the safety-stock target.

Actual cash timing depends on existing stock, purchase orders, payment terms and order constraints. Moving a buffer from FBA to another warehouse also does not remove it from the business.

Start with a small group of A and B products. Record the tier, demand variability, lead-time definition, service target, current buffer and proposed buffer. Keep supplier minimums, case packs and shipment economics in view.

Inventory planning software can keep stock levels, forecasts and replenishment recommendations together. Assign someone to review the settings and follow up when actual demand or delivery times drift.

Review stockouts, estimated missed demand, expedited shipments and inventory investment over several replenishment cycles. If the same SKU keeps needing emergency freight, check the forecast and lead-time assumptions before buying another month “just in case.”