What Is Normal Distribution in Inventory Planning?
Normal distribution in inventory planning is the statistical model that describes how daily demand for a product clusters around an average value in a predictable, bell-shaped pattern. When your SKU's sales history forms this bell curve, you can use standard deviation and Z-scores to calculate exactly how much safety stock to carry for any target service level.
The bell curve tells you that most days your sales will land close to the average, with fewer days at the extremes. If your SKU averages 25 units per day with a standard deviation of 5, roughly 68% of days will see sales between 20 and 30 units. About 95% of days fall between 15 and 35 units. That predictability is what makes normal distribution in inventory planning so useful: it converts uncertainty into a calculable range.
Without this statistical foundation, reorder point calculations and safety stock formulas are guesswork dressed in math. The normal distribution gives you a principled way to translate "I want to avoid stockouts 95% of the time" into a specific number of extra units to hold.
Normal Distribution in Inventory Planning: Z-Score Lookup Table
The core output of assuming a normal distribution is the Z-score lookup. Each Z-score maps to a cumulative probability, which becomes your service level target:
| Target Service Level | Z-Score (Service Factor) | What It Means |
|---|---|---|
| 90.0% | 1.28 | Stock covers demand in 9 out of 10 replenishment cycles |
| 95.0% | 1.65 | Most common target for mid-tier FBA SKUs |
| 97.5% | 1.96 | Typical for hero ASINs with high contribution margin |
| 99.0% | 2.33 | Near-zero stockout tolerance; carries significant holding cost |
These Z-scores feed directly into the safety stock formula:
Safety Stock = Z × σ_dLTWhere σ_dLT is the standard deviation of demand during lead time. The Z-score is the bridge between your business decision ("I want 95% service") and the math ("carry 1.65 standard deviations of buffer"). See the Z-score / service factor entry for the full derivation.
Worked Example: Normal Distribution for an FBA SKU
You sell a kitchen gadget at $34 ASP with a 60-day ocean freight lead time from Shenzhen. Over the last 12 weeks, your daily unit sales fluctuate between 17 and 33 units.
Step 1: Calculate the mean. Average daily demand = 25 units/day.
Step 2: Calculate standard deviation of daily demand. σ_daily = 4.8 units/day. (This is your demand variability.)
Step 3: Scale to lead time. σ_dLT = σ_daily × √(lead time) = 4.8 × √60 = 4.8 × 7.75 = 37.2 units.
Step 4: Pick your Z-score. For a 95% service level, Z = 1.65.
Step 5: Calculate safety stock. Safety Stock = 1.65 × 37.2 = 62 units (rounded up).
Step 6: Calculate reorder point. ROP = (25 × 60) + 62 = 1,500 + 62 = 1,562 units.
That 62-unit safety stock buffer is what the normal distribution gives you: a statistically grounded number instead of a gut-feel guess of "add 10% extra." The difference matters. A flat 10% buffer on this SKU would be 150 units of safety stock, more than double what the math says you need, locking up over $5,000 in unnecessary inventory.
Why Normal Distribution Matters Specifically for FBA
Textbook inventory planning assumes clean demand data. FBA adds noise. Prime Day, Lightning Deals, coupon stacking, and competitor stockouts all create demand spikes that break the bell curve temporarily. Before applying normal distribution in inventory planning to your FBA data, strip out these anomalies from your demand history. Use the last 8 to 12 weeks of organic sales only.
FBA receiving delays also matter. Your supplier shipped on time, but Amazon's inbound process took 14 days instead of 7. That variability belongs in the lead time variability side of the safety stock formula, not the demand side. Mixing the two inflates your standard deviation and overestimates safety stock, tying up cash that could fund your next purchase order.
For sellers tracking forecast accuracy (MAPE), the normal distribution assumption also underlies most statistical forecasting models. If your MAPE is consistently above 30-40%, your demand data may not be following a normal pattern, and it is worth investigating seasonality or external drivers before trusting the Z-score math.
Common Mistakes
1. Including promotional spikes in your demand history. A Lightning Deal that tripled your sales for two days will inflate your standard deviation by 40-60%. Your safety stock calculation overshoots, and you end up with excess inventory collecting aged inventory surcharges.
2. Assuming normal distribution for slow-moving SKUs. If your SKU sells 0 to 3 units per day, you get a skewed distribution piled up at zero. The normal curve produces nonsensical results here, like negative lower bounds. Use a Poisson or discrete model for SKUs under roughly 5 units per day.
3. Forgetting to scale standard deviation to lead time. A common error is plugging daily standard deviation directly into the safety stock formula without multiplying by the square root of lead time in days. For a 60-day lead time, this underestimates safety stock by a factor of 7.75x and virtually guarantees stockouts during replenishment.
Related Terms
Frequently Asked Questions
Why do inventory planners assume demand follows a normal distribution?
Most SKUs with steady repeat-purchase demand produce daily sales that cluster around an average with roughly symmetric scatter. The central limit theorem reinforces this: the sum of many independent purchases over a day or week converges toward a bell curve, letting you use Z-scores and standard deviation to calculate safety stock instead of guessing.
What if my demand is not normally distributed?
If your SKU has heavy seasonality, frequent promotions, or very low daily velocity (under 2 to 3 units per day), the normal assumption breaks down. Consider Poisson distribution for slow movers or empirical simulation. For most mid-velocity FBA SKUs selling 10+ units per day, normal distribution works well enough for practical safety stock calculations.
How does normal distribution connect to safety stock?
Safety stock equals Z times the standard deviation of demand during lead time. The normal distribution provides the Z-score lookup that translates a service level percentage into a concrete safety stock multiplier. Higher service level targets require higher Z-scores, which means more buffer inventory.
What Z-score should I use for Amazon FBA?
Most FBA sellers target a 95% service level (Z = 1.65) for top sellers and 90% (Z = 1.28) for slower SKUs. High-margin hero ASINs may warrant 97.5% (Z = 1.96). The right Z-score depends on your stockout cost relative to your holding cost.
Do I need to test whether my sales data is actually normal?
For most practical FBA planning, formal normality testing is overkill. Plot a histogram of your last 8 to 12 weeks of daily sales. If it looks roughly symmetric around the mean without extreme outliers or multiple peaks, the normal assumption is reasonable. Strip out known anomalies like Lightning Deals or Prime Day before plotting.