Safety Stock Calculation For Intermittent Demand: Techniques And Examples

Safety Stock Extra inventory held to reduce the risk of stockouts caused by demand or supply variability. When demand is intermittent—sporadic, lumpy, or zero-heavy—standard safety-stock formulas based on continuous normal demand break down and can produce misleading results.

Intermittent demand is common for spare parts, slow-moving SKUs, and many B2B items. The challenge is that variance and average demand are poor predictors when long runs of zero demand are followed by occasional large orders. For these items, planners must use specialized statistical techniques and business rules to avoid over- or under-stocking.

Why Intermittent Demand Requires Different Methods

Standard safety-stock formulas assume demand per period follows a continuous distribution (often approximated by normal). Intermittent demand typically has many zero periods and a skewed distribution for non-zero demand. That makes mean and standard deviation unstable and can generate excessively large safety stock if you blindly apply normal-based service factor calculations.

Calculation Methods

Choose from several methods depending on data quality and system capability:

• Croston’s Method: Separates occurrence intervals from demand sizes and forecasts both independently. Good for forecasting intermittent demand, and you can base safety stock on forecast error of the demand-size series.
• Syntetos–Boylan Approximation (SBA): An adjustment to Croston that reduces bias in certain intermittent patterns.
• Bootstrapping & Simulation: Use historical demand to generate many simulated lead-time demand scenarios, then set safety stock to the percentile matching your target service level (e.g., the 95th percentile of simulated lead-time demand).
• Poisson/Compound Distributions: For purely random occurrence rates, model demand occurrences with a Poisson distribution and demand size with an empirical distribution—combine them to derive lead-time demand distribution.
• Rule-Based Buffers: For very low-volume SKUs, assign minimum safety stock as fixed units (e.g., 1–3 units) or based on criticality rather than statistical calculation.

How To Choose The Right Approach

Selection depends on SKU criticality, data volume, and system capabilities. Use Croston/SBA when you have at least a year of transactional data and moderate intermittency. Use simulation where you need explicit service-level guarantees and can afford compute. Apply rule-based buffers for thousands of extremely low-volume SKUs where statistical methods are unstable.

Practical Example

Consider a spare-part SKU with 24 months of demand: 18 months had zero orders; six months had orders of sizes 1, 3, 2, 5, 1, 4. Lead time is 30 days. A planner using average and standard deviation will find misleadingly large variance. A better approach: model occurrence rate (6 occurrences / 24 months = 0.25 per month) and average order size (mean ~2.67). Using Poisson for occurrences and empirical distribution for size, simulate 10,000 lead-time periods and take the 95th percentile of the simulated lead-time demand; if that percentile equals 4 units, set safety stock accordingly (plus review minimum stock level to avoid stockouts for critical parts).

Operational Steps

• Label: Segment SKUs by intermittency (e.g., measure percent-zero and coefficient of variation).
• Label: Apply statistical methods where data supports them and rule-based buffers where it does not.
• Label: Automate simulation or Croston calculations in your forecasting module or use add-on analytics if the ERP lacks functionality.
• Label: Regularly review and reconcile spare-parts buffers with maintenance schedules and criticality matrices.

Tips For Practitioners

• Label: Combine statistical outputs with business rules—critical SKUs often need a safety margin above statistical recommendations.
• Label: Don’t ignore lead-time variability—small suppliers and international shipments magnify risk for intermittent items.
• Label: Use SKU classification (ABC/XYZ) to prioritize analytical effort; focus advanced methods on high-cost or mission-critical SKUs.

In short, the Safety Stock approach for intermittent demand must be tailored: use Croston/SBA or simulation where possible, but lean on simple, defensible business rules for very sparse SKUs. That balance prevents stockouts while keeping inventory investment reasonable.