Safety stock is not a universal percentage. It is a deliberate buffer against the gap between what you expected to sell and receive, and what actually happens. A simple days-of-cover buffer can be appropriate for a young SKU. A variability-based model is more useful once you have enough clean demand and lead-time history.
Method 1: simple days-of-cover safety stock
The fast method starts with average daily demand:
average daily demand = units sold in period ÷ selling dayssimple safety stock = average daily demand × extra buffer daysIf a SKU sells 900 units over 90 days, average daily demand is 10 units. A 7-day buffer equals 70 units. If average lead time is 21 days, the simple reorder point is:
reorder point = 10 × 21 + 70 = 280 unitsThis method is easy to explain and maintain. It is useful when the business has limited history, stable demand or a supplier relationship where the chosen buffer is based on operating judgment.
Method 2: variability-based safety stock
When daily demand and lead time vary, a common planning approach uses a service factor and the combined uncertainty of demand and replenishment:
safety stock = z × √(lead time × demand variance + average demand² × lead-time variance)Here, z represents the chosen service target, demand variance describes daily sales volatility, and lead-time variance describes supplier or transit volatility. Units must be consistent: if demand is daily, lead time must be expressed in days.
Worked variability example
Assume:
- average daily demand: 10 units;
- standard deviation of daily demand: 3 units;
- average lead time: 21 days;
- standard deviation of lead time: 4 days;
- illustrative service factor: 1.65.
Demand variance is 3² = 9. Lead-time variance is 4² = 16.
combined variance = 21 × 9 + 10² × 16 = 1,789combined standard deviation = √1,789 = 42.30safety stock = 1.65 × 42.30 = 69.80 ≈ 70 unitsIn this example the advanced model lands close to the 7-day simple buffer, but that is not guaranteed. High demand volatility, unstable lead time or a different service target can move the result substantially.
When the simple model is enough
- The SKU has little history and a more complex model would create false precision.
- Demand and replenishment are reasonably stable.
- The business can review the buffer frequently.
- The cost of carrying a modest extra buffer is low relative to the cost of a stockout.
- One operator can explain why the chosen buffer days are appropriate.
When variability deserves explicit modeling
- The SKU has enough daily or weekly history to estimate variability.
- Lead time changes materially between orders.
- Stockouts are expensive, such as lost ranking, customer churn or production downtime.
- Inventory carrying cost is material and broad “30-day buffers” tie up too much cash.
- Different SKUs need different service targets based on margin and importance.
Do not let bad history create fake confidence
Before calculating standard deviations, clean the data:
- separate true zero demand from out-of-stock days;
- remove launch, clearance and one-off wholesale orders when they are not representative;
- record lead time from purchase order to usable inventory, not only supplier ship date;
- separate supplier delay from customs, port, freight and receiving delay where possible;
- use the same time unit for demand and lead time.
A polished formula cannot repair biased inputs. If the SKU was out of stock for 20 days, observed sales understate demand.
Connect safety stock to the reorder point
Safety stock is only one part of the decision:
reorder point = expected demand during lead time + safety stockThe Reorder Point Calculator supports a transparent planning estimate. Use it with current lead time, demand and buffer assumptions, then compare the result with cash and storage constraints.
Translate the buffer into cash
If safety stock is 70 units and landed cost is $18, the buffer ties up $1,260 before storage, financing and obsolescence.
cash tied in safety stock = safety-stock units × landed unit costThat does not make the buffer wrong. It gives you a price for the service level. Compare it with the expected cost of stockouts, lost sales and emergency replenishment.
Use an ABC decision rule
| SKU group | Typical role | Review cadence | Model preference |
|---|---|---|---|
| A | High contribution or strategic traffic | Weekly | Variability-based where data is clean |
| B | Steady mid-value range | Biweekly or monthly | Simple model with periodic validation |
| C | Long tail or low contribution | Monthly or quarterly | Small simple buffer or order-on-demand |
The classification should reflect contribution and strategic importance, not only revenue. A low-volume replacement part can still be critical.
Monthly validation loop
- Compare predicted demand during lead time with actual demand.
- Compare planned lead time with actual usable-inventory lead time.
- Record stockout days and emergency freight.
- Measure average inventory and carrying exposure.
- Adjust only one main assumption at a time.
- Document the service target and next review date.
Decision rule
Use the simplest model that is honest about the available data. Upgrade to a variability-based method when it materially improves the cash-versus-stockout decision—not merely because the formula looks more sophisticated.