Turn lead-time history into actionable variability insights fast. Compare suppliers, processes, and routes with confidence. Export reports, tune safety stock, and plan steadier deliveries.
| Shipment | Lead time (days) | Notes |
|---|---|---|
| A-001 | 6.0 | Baseline lane |
| A-002 | 7.2 | Customs hold |
| A-003 | 5.4 | Expedited pick |
| A-004 | 8.1 | Supplier constraint |
| A-005 | 6.6 | Normal variation |
In engineered operations, variability often comes from handoffs: release queues, picking windows, carrier cutoffs, inspection holds, and batch scheduling. If your mean is 6.5 days but P95 is 8.8 days, planning to the mean will miss deadlines about 5% of the time for that lane.
Standard deviation measures spread in the same unit as lead time. CV normalizes it for fair comparisons: a CV of 0.10 is far steadier than 0.35. Percentiles translate directly into promises: P90 supports “on time in 9 of 10” commitments; P95 is better for critical assemblies.
A practical rule is to set internal plan dates near P90 and customer-visible dates closer to P95, then work the gap down with process improvements. If P95 − mean exceeds 20% of the mean, the lane has a heavy right tail and needs root-cause categorization of late arrivals.
When you enter average daily demand and a one-sided service level, the calculator estimates safety stock using both demand variability and lead time variability. With d = 120 units/day, σd = 25, L = 6.5 days, and σL = 1.1, a 0.95 service level can require a buffer on the order of several hundred units.
Split the histogram into segments by supplier, shift, carrier, and product family. Stabilize the highest-variance segment first: enforce release cutoffs, standardize packaging, instrument dwell time at inspection, and set WIP caps. A 15% reduction in σL can materially reduce safety stock while improving service.
Recompute metrics weekly for high-volume lanes and monthly for stable ones. Track mean, CV, and P95 as a trio. Trigger investigation if CV rises by 0.05 or if P95 increases by more than one day versus the prior period. Store exported CSVs to maintain an auditable planning trail.
Use at least 20 observations for stable percentiles. With fewer samples, P95 can swing with single outliers, so pair it with cause codes and repeat measurement frequently.
Only remove points with documented, non-repeatable causes. Otherwise, keep them; they represent real tail risk. Consider separate lanes (normal vs. exception handling) instead of deletion.
CV compares spread relative to the mean. Two lanes can have the same standard deviation, but the lane with the smaller mean has higher relative instability and needs more buffer or tighter control.
Inventory protection focuses on the “late” tail, not early arrivals. One-sided service levels map directly to the probability of not stocking out due to demand during lead time.
Yes. Choose the unit that matches how you plan and record events. Keep demand expressed per day for the buffer model, or convert consistently before interpreting reorder points.
Segment the data by season, capacity regime, or promotion periods. Compute separate means and percentiles for each segment and switch planning parameters when the operating regime changes.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.