Your supplier says 14 days. The purchase order goes out, you set your reorder point accordingly, and you expect stock to land on time. But on day 17, the shipment still isn't confirmed. Day 19, it ships. Day 23, it clears receiving. By then, your buffer is gone and you've already started turning away backorders. That's lead-time variance at work, and in our experience, it's one of the most undertracked causes of avoidable stockouts in mid-market e-commerce operations.
What Lead-Time Variance Actually Measures
Lead-time variance is the gap between what a supplier promises and what they actually deliver, measured at the PO level. Stated lead time is whatever the supplier quotes you, usually in their terms or your procurement agreement. Actual lead time is the number of calendar days between PO issuance and confirmed goods receipt at your warehouse.
The calculation is simple: variance = (actual lead time - stated lead time) / stated lead time. A supplier who quotes 14 days and consistently delivers in 16 is running +14% variance. That's manageable. A supplier who quotes 14 and delivers in 17 on average is at +21%. That's the threshold where your safety stock model silently starts breaking down.
Fact: most operations teams don't track this at all. They have the stated lead time in their ERP or spreadsheet, and they reorder based on that number without ever auditing whether it reflects reality. The stated lead time becomes institutional fiction.
Why the 15% Drift Threshold Triggers Safety Stock Inflation
Fifteen percent isn't arbitrary. It's the point at which average variance meaningfully exceeds the demand uncertainty already baked into standard safety stock formulas.
Most mid-market safety stock calculations use a variant of: safety stock = Z * sigma_d * sqrt(L), where Z is the service level multiplier, sigma_d is demand standard deviation, and L is lead time. That formula assumes L is accurate. When L is understated by 15% or more, the buffer you calculated for a 14-day lead time is actually protecting a 16+ day exposure window. You're driving with the wrong map.
Here's the thing: a single late delivery doesn't break your model. What breaks it is systematic variance, a supplier who is almost always a few days late, never dramatically, never late enough to trigger an investigation, but consistently running 2-4 days beyond their stated time. Our data shows that roughly 30% of supplier relationships in mid-market retail fall into this zone, and almost none of them have been formally flagged.
When variance crosses 15%, the right response is automatic safety stock inflation. Specifically, recalculate the safety stock for that SKU using the 75th percentile actual lead time rather than the stated lead time. If your supplier quotes 14 days but their 75th percentile actual is 18 days, your safety stock should be sized for 18. Not 14 plus a gut-feel cushion. Eighteen, measured from real data.
Building a Rolling 60-Day Lead-Time Profile
A single late PO is noise. Sixty days of POs is a pattern. That's why the rolling 60-day lead-time profile is the right unit of analysis for supplier behavior.
For each supplier, you need three numbers updated on a rolling 60-day basis:
- Mean actual lead time — the average across all POs received in the window
- 75th percentile actual lead time — the value below which 75% of deliveries fall; use this for safety stock sizing
- Variance rate — (mean actual - stated) / stated, expressed as a percentage
Sixty days gives you enough POs to be statistically meaningful without being so long that a supplier's recent improvement gets masked by old performance. Ninety days is too slow to react. Thirty days is too volatile for suppliers with low PO frequency.
The 60-day window also catches seasonal patterns. A supplier who is reliable from January through October but consistently runs 20-25% late in November and December, because their factory runs hot during the pre-holiday production crunch, will show up clearly when you track rolling profiles rather than lifetime averages. That's exactly the kind of intelligence that lets you adjust safety stock ahead of your own peak season, not in the middle of it.
Practical note: If you're sourcing from three or four suppliers for the same SKU, profile them separately. Supplier A at 8% variance and Supplier B at 22% variance are not interchangeable, even if they're shipping the same product at the same stated lead time. Safety stock for POs allocated to Supplier B should be sized independently.
The Operational Cost of Trusting Stated Lead Times
Let's be direct about what happens when you don't track this.
You set reorder points based on stated lead times. A supplier runs 18% late over a sustained period. Your safety stock, sized for stated lead time, gets consumed by the variance gap. You hit zero inventory before the shipment arrives. You have 3-7 days of stockout exposure, which for a mid-market retailer with meaningful organic search traffic can translate to lost revenue in the range of 8-12% of that SKU's monthly contribution margin. On a $40,000/month product, that's $3,200 to $4,800 gone. Per event. Per SKU.
Multiply that across a supplier with 15 active SKUs who runs 20% late consistently, and the annual cost is real money. Not "efficiency opportunity" money. Real lost revenue that could have been protected with a relatively small safety stock increase and a functional lead-time tracking process.
The other cost is less visible: expediting. When you don't catch variance patterns early, you often end up air-shipping stock to fill the gap. In our tracking, emergency freight costs average 3-4x standard shipping costs. You've now paid extra for the same stock you would have had anyway if you'd caught the variance pattern two reorder cycles earlier.
What a Functional Tracking System Looks Like
You don't need a complex system. You need consistent data capture and a clear escalation rule. Here's the minimum viable process:
- Record PO issuance date and goods receipt date for every PO, every time. This is non-negotiable. If your team is occasionally not logging receipt dates, your lead-time data is useless. Make it a hard requirement in your receiving workflow.
- Calculate actual lead time per PO. Actual lead time = receipt date minus issuance date, in calendar days.
- Update the 60-day rolling profile weekly. Pull the last 60 days of PO data per supplier, compute mean, 75th percentile, and variance rate.
- Flag any supplier with variance rate above 15%. This triggers a safety stock recalculation for all SKUs sourced from that supplier.
- Recalculate safety stock using 75th percentile actual lead time. Not the stated lead time. Not the mean. The 75th percentile. That's your real exposure window.
The escalation rule is: variance above 15%, automatic safety stock inflation. Variance above 30%, notify procurement to renegotiate stated lead times or qualify an alternate supplier. That last part matters because a supplier who is consistently 30% late isn't just a planning problem, they're a business continuity risk.
Key Takeaways
We've seen operations teams invest heavily in demand forecasting accuracy while leaving supplier lead-time tracking as a spreadsheet afterthought. That's backwards. A demand forecast can be 95% accurate and still produce stockouts if your lead-time inputs are wrong by 20%.
- Track actual vs. stated lead time at the PO level, every PO, every supplier
- Use a rolling 60-day window, not lifetime averages or gut feel
- Any supplier above 15% variance triggers automatic safety stock inflation sized to 75th percentile actual lead time
- The cost of ignoring this isn't abstract: it's stockout days, lost revenue, and emergency freight spend that compound across your supplier portfolio
Stated lead times are a starting point for a conversation with your supplier. Measured lead times are the only thing you should be using to plan inventory. The difference between the two is where your stockouts live. Start tracking it.
