Let me open with a ledger. In August 2025, a small-appliance exporter in Shenzhen asked me to look at their warehouse. Their hero SKU — a portable blender cup — moved 260 units a day, with a 42-day procurement lead time by ocean freight. The warehouse manager had eyeballed the reorder point at 9,500 units. "Close enough," he said. September peak season hit, daily sales spiked to 430, the vessel slipped, lead time drifted to 48 days, and the SKU stocked out for 7 straight days. About 1,900 units never shipped; at $2 margin per unit (roughly 14 yuan), that's $3,800 in lost margin — plus two emergency air-freight shipments that burned another $2,500. Seven days of stockout, $6,300 gone — and the entire disaster traced back to one wrong number: the reorder point.
I've seen this in nine out of ten warehouses: the reorder point is set by gut feel. Some double the lead-time demand. Some use last month's peak. Some just pick a round number. Coming from a finance background, I can't stand it — an inventory waterline is a bill you can compute to the unit. Here's the full calculation: reorder point = average daily demand × procurement lead time + safety stock. Three parameters, and not one of them tolerates guesswork.
The Reorder Point Formula: Three Parameters, Zero Hand-Waving
The formula first:
Reorder Point = Average Daily Demand × Lead Time + Safety Stock
The first term is straightforward: at 260 units a day and 42 days in transit, you need 260 × 42 = 10,920 units on hand just to survive until the next batch lands. The Shenzhen warehouse's 9,500 didn't even cover expected consumption. The stockout wasn't bad luck — it was arithmetic.
The second term, safety stock, is insurance: it absorbs demand spikes and late arrivals. The standard method:
Safety Stock = Z × Daily Demand Standard Deviation × √(Lead Time)
Z is the service-level factor from the standard normal distribution. How confident you want to be against stockouts is literally a question of how much you're willing to pay for Z:
| Target Service Level | Z | Meaning |
|---|---|---|
| 90% | 1.28 | 1 stockout per 10 replenishment cycles |
| 95% | 1.65 | 1 stockout per 20 cycles |
| 97.5% | 1.96 | 1 stockout per 40 cycles |
| 99% | 2.33 | 1 stockout per 100 cycles |
| 99.9% | 3.09 | Practically never |
Run the blender-cup numbers: average daily demand 260, standard deviation about 45 (peak-season volatility), 42-day lead time, 95% service level (Z = 1.65). Safety stock = 1.65 × 45 × √42 ≈ 1.65 × 45 × 6.48 ≈ 481 units. Reorder point = 10,920 + 481 = 11,401 units. Their 9,500 didn't even reach the formula's floor. That, to me, is money spent stupidly — the most expensive kind of wrong.
The Safety Stock Bill: Higher Service, Pricier Insurance
Now let's break down what safety stock actually costs. Many operators believe "more safety stock can't hurt." Finance disagrees: every unit of safety stock is cash sitting on a rack, and the industry-standard inventory carrying cost runs 20–30% per year (capital, rent, insurance, shrinkage — all in).
The blender cup again: 481 units of safety stock at $5.30 unit cost (about 38 yuan) ties up roughly $2,550; at a 25% carrying rate, that's $640 a year in holding cost. That's the annual "insurance premium" for a 95% service level.
Raise it to 99%? Z jumps from 1.65 to 2.33, safety stock from 481 to 680 units, and the yearly premium rises to about $900. Moving from 95% to 99% service level costs 41% more safety stock, while the stockout probability only drops from 5% to 1%. The last few percentage points of insurance are brutally expensive. That's my first counterintuitive verdict: safety stock is not a case of "more is better" — it's a trade with steeply diminishing returns.
So I tier it. No one-size-fits-all:
| SKU Tier | Service Level | Z | Logic |
|---|---|---|---|
| A (hero / high-margin) | 99% | 2.33 | A stockout costs a fortune; the premium is worth it |
| B (regular) | 95% | 1.65 | The industry-standard waterline |
| C (long tail / low-margin) | 90% | 1.28 | A stockout barely stings; don't lock up cash |
Lead-Time Variability: The Insurance Most People Buy for the Wrong Risk
There's a subtler trap, and it kills more plans than demand volatility. The safety-stock formula above only covers "daily sales are unstable." It ignores "transit takes an unstable number of days." In practice, lead-time wobble is often deadlier than demand wobble.
November 2024, a pet-supplies distributor in Dongguan. The owner, terrified of stockouts, set reorder points at "1.5× lead-time demand" — pure gut feel. One SKU moved 90 units a day with a stable 25-day domestic lead time, and he'd set the reorder point at 4,200 units. My calculation: daily demand std dev 18, lead time rock-steady (std dev about 2 days), Z = 1.65. Safety stock = 1.65 × 18 × √25 = 149 units. Reorder point = 90 × 25 + 149 = 2,399 units. He was holding 1,801 units too many. At $4.50 unit cost (about 32 yuan), that's roughly $8,100 of excess cash; at a 25% carrying rate, over $2,000 a year burned — a tax paid purely for peace of mind.
But flip it: if the supplier's lead time is wildly unstable (ocean freight with a 6-day std dev), the formula upgrades:
Safety Stock = Z × √(σ_demand² × Lead Time + Avg Demand² × σ_leadtime²)
That lead-time term (average demand squared times lead-time variance) frequently dwarfs the demand term. Here's the second counterintuitive point: most operators pile on safety stock to hedge lead-time risk, yet compute it with the demand-volatility formula — the money lands in the wrong place. With an unreliable supplier, the right fix is lead-time scorecards, dual sourcing, and squeezing the lead time itself — not stacking insurance in the warehouse.
The Cheap Lever: Shorten Lead Time Before You Add Stock
Of the three parameters, which is the money-saving lever? Do the math.
Back to the blender cup: at a 42-day lead time, the reorder point was 11,401. Negotiate production plus sailing down to 35 days: expected consumption drops to 260 × 35 = 9,100; safety stock becomes 1.65 × 45 × √35 ≈ 439; the reorder point falls to 9,539 units. Seven fewer lead-time days cut 1,862 units off the reorder point and free about $9,900 in cash. One hard week of supplier negotiation beats whatever safety-stock budget finance would have approved.
Ranked by return on effort:
- Squeeze lead time. Each day shaved off cuts the reorder point by a full day of average demand, and safety stock shrinks too. Costs nothing but negotiation and management.
- Tier your service levels. 99% for A SKUs, 95% for B, 90% for C. Stop buying luxury insurance for long-tail items.
- Add safety stock last. Only where volatility is real and no other lever works.
The Calculable Conclusion: Five Steps You Can Run Tonight
No more gut feel. Take any SKU and walk these five steps:
- Pull 90 days of daily shipments: average and standard deviation (your outbound system has both).
- Record this supplier's actual lead times over the past 6 months: mean and standard deviation.
- Set the service level by ABC tier and look up Z.
- Plug into the formula for safety stock and the reorder point.
- Set the system alert: when on-hand inventory falls below the reorder point, purchasing triggers automatically.
One full worked example so you can copy it directly. An A-tier SKU: average demand 120/day, std dev 30; lead time 20 days, std dev 3 days; Z = 2.33.
Safety stock = 2.33 × √(30² × 20 + 120² × 3²) = 2.33 × √(18,000 + 129,600) = 2.33 × 384.2 ≈ 895 units. Reorder point = 120 × 20 + 895 = 3,295 units.
Remember this: a reorder point set too high saves you the stockout money and burns you the cash; too low does the reverse. The most expensive kind is the one set by gut feel — it wins neither side, while letting you believe you did the math.
By Lao Zhou, Logistics Cost Analyst
