Safety stock versus cash locked in inventory

A full warehouse locks up cash and an empty one loses sales. Both cost money and neither appears in the accounts under its own name. This case puts a price on both for the same product and compares three policies.

The shop is hypothetical. The starting figures are declared and belong to no real company.

The starting figures

Two case assumptions, stated up front

This case rests on two assumptions that are not measured data and that belong on the table before any number is read.

First: how much unserved demand is genuinely lost. When a product is out of stock, not everyone walks away. Some wait, some buy something else from you and some go to a competitor and never come back. Here it is assumed that 55% of unserved demand is lost and that the remaining 45% is simply deferred. Turning all unserved demand into lost sales would overstate the cost of a stockout, and it is a very common mistake.

Second: how many stockouts each policy brings. It is assumed the aggressive policy suffers 3 stockouts a year, the balanced one 2 and the conservative one 1, each lasting 4 days. That is a plausible relationship, not a law: the real frequency depends on the product and the supplier.

Both assumptions are stress-tested below.

The order of calculation

1. Safety stock by the statistical method, for three service levels.
2. The reorder point for each.
3. The annual cost of holding that buffer, summing the four cost families.
4. The annual cost of the stockouts associated with each level.
5. Total cost and comparison.
6. A sweep of the first assumption to see whether the decision holds.

The three scenarios

Read it left to right and you can see the two forces pulling against each other. Raising the service level from 90% to 99% multiplies safety stock from 76 units to 137 units and cash locked in from €714.40 to €1,287.80. Holding cost rises from €142.88 to €257.56.

But stockout cost does the opposite: it falls from €1,030.92 to €343.64.

The decision, derived

Adding the two columns, the lowest total cost belongs to the conservative policy, at €601.20 a year against €869.64 for the balanced one and €1,173.80 for the aggressive one.

This surprises a lot of people, because intuition says less stock is always better for cash. And it is — for cash. But not for total cost, because on this product a stockout costs more than the buffer does: every point of service level bought saves more in lost sales than it spends in the warehouse.

The reason is in the starting figures. Contribution per unit is €7.10 on a cost of €9.40, so losing a sale hurts nearly as much as storing a unit for a whole year costs. On a thin-margin product the sum would flip.

But the decision is not stable

Here is the important part of the case. That conclusion depends on the first assumption, and that assumption is not measured.

Sweeping the share of demand genuinely lost, from 5% up to 95%, the winning policy changes twice: no is the answer to whether the decision survives the sweep.

• Below 10% genuinely lost, the aggressive policy wins: if almost everyone waits, the buffer does not pay for itself.
• Between 10% and 15%, the balanced one wins.
• From 15% upwards, the conservative one wins, and keeps winning to the end of the sweep.

The case assumption, 55%, falls inside the third band and comfortably away from the boundary. That gives confidence in the conclusion, but the figure genuinely worth measuring before deciding is how much unserved demand is really lost, not the service level.

That is what this case contributes: the question was never "99 or 90?". It was "how many people wait?".

How to measure the assumption that decides

• Mark the out-of-stock days in your history and look at what those customers did: how many bought the product in the days that followed.
• Compare demand in the week after a stockout with a normal week. A spike means demand was deferred.
• Check whether the product has a substitute in your own catalog. If it does, the real loss is much smaller.
• Separate new customers from returning ones: returning customers wait longer.

Independent checks

• Combined lead-time deviation: the square root of (12 × 6² + 22² × 2.5²), computed separately and compared with the engine.
• Each safety stock: z times sigma, rounded up by hand.
• Each reorder point: 22 × 12 plus the buffer, added separately.
• Cash locked in: safety stock times unit cost.
• Each stockout cost: days out of stock times daily demand times the lost share times unit contribution, multiplied separately.
• Total cost: both columns added again.

Limitations

• The statistical method assumes normal, independent demand. With strong seasonality or intermittent demand, the calculated buffer falls short exactly when it is most needed.
• The relationship between service level and stockouts per year is a case assumption, not a measurement. In reality it depends on the supplier.
• Stockout cost here captures only lost contribution. It excludes extraordinary operational costs, negative reviews and lost channel ranking, all of which exist and are valued at zero here.
• Holding cost is computed on the safety stock, not on total inventory. Cycle stock costs money too.
• One product only. A catalog also needs a prioritization rule.
• No opportunity cost of capital beyond the declared rate.

What would change the conclusion

• Unserved demand being lost less than assumed: below 10%, the aggressive policy wins.
• A much lower contribution per unit: stockouts would hurt less and the buffer would pay worse.
• A full warehouse: storage cost stops being proportional and jumps.
• A more stable supplier lead time: less deviation, less buffer for the same service.
• A second supplier: lead-time uncertainty falls and the whole table moves.

Sources and methodology

Sources

• Cumulative Distribution Function of the Standard Normal Distribution, NIST/SEMATECH e-Handbook of Statistical Methods, National Institute of Standards and Technology. Provides the critical values used (0.900 → 1.282; 0.950 → 1.645; 0.990 → 2.326). No territory applies; consulted 9 September 2026; it does not expire. The page says nothing about inventory: linking service level to a z value is a modelling decision.

• Reorder Point Formula: Inventory Management Models · A Tutorial, Supply Chain Resource Cooperative, NC State University. Supports the distinction between cycle service level — which is what is set here — and fill rate. No territory applies; consulted 9 September 2026; reviewed annually.

What is not cited. There is no source for "what percentage of customers never come back after a stockout". We looked, and no identifiable study exists with a sample, a method, a date and a population that applies to ecommerce. That is why the percentage appears here as a declared case assumption and gets swept, rather than being cited as a fact.

Methodology

Four Profyza calculation engines, run three times each, once per policy: safety stock, reorder point, inventory cost and stockout cost. The assumption sweep runs the stockout engine another nineteen times, from 5% to 95%. No figure is typed by hand and all of them are checked against an independent calculation.

Currency and market: euros, European Union.

Last reviewed: 9 September 2026.

Author: Profyza editorial team.

How the figures were produced. Each one is generated by the Profyza calculation engine for that step, using the inputs declared above. Every figure is also cross-checked against an independent arithmetic check — also automated — that recomputes it without calling the engine, which is what catches the text and the calculation drifting apart. It is not a human review, and this piece has not been reviewed by an economist, an accountant or a licensed adviser. If your situation needs that, get it.

Methodology: how all of this is calculated is set out on the methodology page.

Automated mathematical verification: Profyza calculation engines. This is not a human or professional review.