Our company sources a lot of components from overseas, and supplier lead times are all over the place-sometimes 30 days, sometimes 60, with no clear pattern. We’ve been using a simple rule-of-thumb safety stock (e.g., 2 weeks of demand), but we’re constantly either out of stock or sitting on too much inventory.
I’ve seen formulas that include both demand variability and lead time variability, but I’m not sure how to apply them when lead times are so erratic. Do we just use the average lead time and standard deviation? What if the distribution is skewed or has outliers? And how often should we recalculate safety stock when lead times keep changing?
We’re trying to move from a gut-feel approach to something more data-driven, but I’m not sure where to start with the math and how to explain it to our planners.
Let me walk through a real example from our electronics division. We had a supplier with lead times ranging from 25 to 75 days. First, we pulled 12 months of lead time data and calculated the mean (48 days) and standard deviation (12 days). Then we used the formula: Safety Stock = Z × √(Avg LT × σ²_demand + Avg Demand² × σ²_LT).
For a part with average demand of 100 units/day, demand std dev of 20, and targeting 95% service level (Z=1.65), the calculation was: 1.65 × √(48×400 + 10000×144) = 1.65 × √(19200 + 1440000) = 1.65 × 1208 ≈ 1993 units. This was way higher than our old 2-week rule (1400 units), but it matched our actual stockout experience. The key was accounting for both sources of variability, not just demand.
I get the math, but for low-value C items, is it really worth the complexity? We have thousands of SKUs, and calculating safety stock with full variability formulas for every one is a lot of work. For items under $10 value, we just use a simple rule: 4 weeks of average demand. It’s not perfect, but it’s good enough, and it frees up our time to focus on the A items where the dollars matter.
My advice: use the sophisticated formulas for your top 20% of items by value or criticality. For the rest, keep it simple. Don’t let perfect be the enemy of good.
Data quality is critical here. Before you calculate anything, clean your lead time data: remove outliers caused by one-off events (strikes, natural disasters, data entry errors). We use a simple rule-any lead time more than 2.5 standard deviations from the mean gets investigated and potentially excluded.
Also, make sure you’re measuring the right lead time: order placement to receipt, not just shipment to receipt. And segment by product family or supplier if patterns differ. Once your data is clean, the formulas work much better. We refresh our lead time statistics quarterly and recalculate safety stock for any item where the std dev changed by more than 20%.