Safety stock · 3 methods compared

Most online safety stock calculators give you one number from one formula. Real operators know that the right formula depends on what data you have and what you trust. This calculator runs all three standard methods on the same inputs and shows you the difference.

Max-min Average demand Service level (Z-score)

The three methods

Safety stock answers one question: how many extra units do I keep on hand to absorb a demand spike or a supplier delay? Three standard answers exist, each with different data requirements.

Method 1 — Max-min (basic)

Formula: (max daily usage − average daily usage) × lead time in days

Cheapest method to compute. Requires only a "max" day and an "average" day. Useful when you have very little data or very stable demand. Useless when your "max" day was a one-off promotion that won't recur.

Method 2 — Average & standard deviation

Formula: Z × σ_LT where σ_LT = std deviation of lead time demand

Uses your full demand history to compute the standard deviation of demand during lead time. More accurate than max-min. Requires you to have at least 8-12 lead-time periods of demand data.

Method 3 — Service-level target (Z-score)

Formula: Z × σ_LT — where Z is the number of standard deviations you need for your desired service level.

Same formula as method 2, but you pick the service level (e.g., 95%) and we look up the Z-score (1.65 for 95% one-sided). Lets you make a deliberate trade-off between stockout risk and carrying cost.

Why compare?

Real operators don't pick one and stick with it. They pick the one whose data they trust, then sanity-check it against the others. If the three methods give wildly different numbers, your data is too noisy for any of them to be reliable — time to collect more.

How to use the comparison

Look at the spread between the three numbers. Three things can happen:

All three close (within 20%). Your demand pattern is stable and the choice of method doesn't matter much. Pick method 3 with a 95% service level — it's defensible and easy to explain to non-operations people.
Method 1 is wildly higher. Your "max day" was probably a one-time event (a flash sale, a weather spike, an end-of-quarter push). Don't use it. Lean on method 2 or 3.
Methods 2 and 3 are wildly different from method 1 but close to each other. Your demand history is consistent but has long tails. Good — pick whichever service level your business demands.

Picking a service level

Standard targets: 90% for low-margin C items, 95% for B items, 97.5%-99% for A items and any SKU with customer-facing stockout consequences (long lead times, single-sourcing, etc.). The trade-off is dollars: going from 95% to 99% service level typically increases safety stock by 60% for the same demand pattern.

What this tool doesn't do

Doesn't model seasonality. If demand differs a lot by month, none of these formulas are right; you'd need a seasonal decomposition or a buffering model.
Doesn't model substitution. If a customer can substitute, the required service level is lower. Not modeled.
Doesn't model lead-time distribution shape. We assume normal distribution of lead time demand. In practice, supplier lead times are often skewed (longer than average more often than shorter).

FAQ

Q: I only have 4 weeks of demand data. Can I use this?

A: Yes, but the standard deviation estimate will be unreliable. Method 1 (max-min) is your best shot at that data depth.

Q: What's a "good" service level?

A: Depends entirely on the cost of stockout. A SKUs that costs you $50 in customer lifetime value per stockout deserves a 99% target. A throwaway item deserves 85%.

Related tools

Reorder Point — plug safety stock in directly.
Stockout Risk — estimate the probability of an out.
Carrying Cost — size the cost consequence of your chosen buffer.