πŸ“Š Demand Variability

"We sell about 20 a day" is only half the truth. The other half is how wildly the days differ β€” and that second half is what forces you to hold safety stock.

What is it?

Demand variability is how much actual demand scatters around its average. Two products can both average 20 units/day: one sells 18–22 like clockwork, the other swings between 5 and 45. Same average, completely different planning problem.

The standard measure is the standard deviation (Οƒ): roughly, the "typical distance" of a day's demand from the average. If demand behaves normally, about 68% of days land within Β±1Οƒ of the average and 95% within Β±2Οƒ. To compare products with different volumes, divide Οƒ by the average to get the coefficient of variation (CV) β€” under ~10% is steady, 10–25% is moderate, above ~25% is lumpy and hard to plan.

Why it matters: safety stock scales directly with Οƒ (safety stock = z Γ— Οƒ Γ— √lead time). Halve the noise and you halve the buffer β€” which is why forecasting improvements and smoothing promotions are inventory-reduction projects in disguise.

Measure it yourself

The time bucket each observation covers β€” sets the units of the average and Οƒ
Paste or type observed demand, separated by commas or spaces

…or simulate a product

Generates random periods with that average and Οƒ, fills the history box above β€” try Οƒ = 2 vs Οƒ = 15 and watch the chart change character.
Daily demand Average Β±1Οƒ (β‰ˆ68% of days) Β±2Οƒ (β‰ˆ95% of days) Days outside Β± 1Οƒ
Observations (n)
–
Average
–
Std. deviation Οƒ
–
CV (Οƒ Γ· avg)
–
Outside Β±1Οƒ (~32%)
–
Outside Β±2Οƒ (~5%)
–

How the numbers are calculated

Average = Ξ£ demand Γ· n
Οƒ = βˆš( Ξ£ (demand βˆ’ average)Β² Γ· (n βˆ’ 1) ) Each day's distance from the average, squared (so ups and downs don't cancel out), averaged, then square-rooted back to units.
From spread to service level The intro's rule is two-sided β€” it counts unusually low periods too: β‰ˆ68% of periods fall within Β±1Οƒ, β‰ˆ95% within Β±2Οƒ. But you only stock out when demand runs high, so a service level looks at just the upper tail: buffer of +1Οƒ β†’ β‰ˆ84% service Β· +1.64Οƒ β†’ 95% Β· +2Οƒ β†’ β‰ˆ98%. That z-multiplier is exactly the z in the safety-stock formula below.
What your Οƒ implies for safety stock at 95% service:

Try this: add one freak period (say 90) to the history and watch Οƒ jump β€” a single outlier can inflate your buffer for months. This is why planners investigate extreme observations (promotion? data error? one-off bulk order?) before letting them into the calculation.