Sales want products in stock.
Finance wants less cash tied up in inventory.
And you sit in the middle with one inventory budget and thousands of SKUs.
It can look like a negotiation where one side has to give in. But finding the balance between service levels and working capital starts with a calculation you can do item by item.
The short answer: all else being equal, a higher service level requires more safety stock, which ties up more working capital. But in practice you can also improve service levels in other ways. More reliable deliveries, for example, can raise the service level without extra inventory.
The right balance depends on the cost of holding extra inventory compared with the cost of running out. That is why service level targets belong at item or segment level, not across the entire inventory.
This guide shows why the last few percentage points of service level are so expensive, how to find the optimal service level for different items, and where your working capital is actually tied up. We use one hypothetical example throughout, so you can check the calculations yourself.
Picture a component that sells 400 units a month on average. Demand varies from month to month. The standard deviation is 120 units, which is a measure of how much demand varies around the average. The supplier has a fixed lead time of one month, and each unit costs €35. All figures are constructed for this example.
There are several ways to calculate the safety stock needed to cover demand fluctuations during lead time. Here we use the classic normal distribution formula, which Peter L. King covers in his APICS article "Crack the Code", among others.
This formula uses the cycle service level: the probability of not running out during a replenishment cycle. That is not the same as delivering 95% of the quantity demanded, and we come back to the difference below.
Safety stock = z × standard deviation of demand × √lead time
The z-score corresponds to your chosen cycle service level. In Excel you find it with =NORM.S.INV(0.95), which returns about 1.65 for 95%. The formula assumes that demand is roughly normally distributed, that lead time is fixed, and that the standard deviation and lead time use the same time unit (months here, so √1 = 1).
Because the lead time is fixed, only the variation in demand determines the safety stock here. Average demand does not enter the formula, but you still need it to calculate the reorder point. For items with few and irregular sales, the normal distribution assumption breaks down and the formula gives a misleading result.
Here is what that means for the component:
| Cycle service level | z | Safety stock | Working capital in safety stock |
|---|---|---|---|
| 90% | 1.28 | approx. 155 units | approx. €5,400 |
| 95% | 1.65 | approx. 195 units | approx. €6,900 |
| 98% | 2.05 | approx. 245 units | approx. €8,600 |
| 99% | 2.33 | approx. 280 units | approx. €9,800 |
| 99.9% | 3.09 | approx. 370 units | approx. €13,000 |
Moving from 90% to 95% ties up about €1,500 more. Moving from 95% to 99% ties up about €2,900 more. The last 0.9 percentage points up to 99.9% tie up about €3,200 on their own. Each percentage point costs more than the one before, and the formula cannot give you 100% at all.
The component is one SKU, and you cannot simply multiply the result. Each SKU has its own demand, its own variability and its own price. But the mechanism applies to all of them, and when you set service levels across thousands of SKUs, the difference between 95% and 99% is no longer a detail. It is a management decision.
Capital tied up in inventory is not the same as a cost.
Inventory carrying costs are the annual costs of holding inventory, including the cost of capital, storage, handling, insurance and the risk of obsolescence.
A post on Dansk Lean Forum (in Danish) cites estimates of inventory carrying costs between 15 and 40% of inventory value, depending on industry and company.
In our example, we assume annual carrying costs of 20% of inventory value. On that basis, moving the component from 95% to 99% costs about €570 a year.
So the question is whether it is worth €570 a year to reduce the risk of running out. It depends on what it costs to run out of that specific item. If the customer is willing to wait a week, the cost may be relatively low. If the customer buys from a competitor, you lose the margin and possibly more. If a production line stops because the component is part of a bill of materials, running out once can cost more than a full year of carrying cost.
That is why there is no single right service level. There is a right service level for each type of item.
Before you set a target, you need to know what the number covers. There are several recognised ways to calculate service level, and they give different results for the same item:
This matters in practice. If you enter 95% in the safety stock formula, it means you run out in 5% of replenishment cycles. That is not the same as 95% in your KPI report if the report measures fill rate or OTIF.
The discussion about service level and working capital goes in circles when sales, purchasing and finance are looking at three different numbers. Agree on the measure before you negotiate the level.
The optimal service level is the point where the cost of holding more inventory is balanced against the cost of running out.
The right balance will differ from one item to another. But the cost of running out can rarely be calculated precisely. That is why the decision also requires a business judgement.
Inact uses a double ABC analysis as the starting point. Instead of ranking items by revenue alone, you classify them on two dimensions at once, for example revenue and number of order lines. This gives a matrix of nine segments, from AA to CC.
An AA item has both high revenue and many order lines, so poor availability can affect many customer orders.
An AC item has high revenue but few order lines, so each order line represents a relatively large share of revenue. Whether it needs safety stock, and how much, depends on demand, lead time and the consequences of running out.
Double ABC shows you where to focus, but it does not calculate safety stock.
A CC item has low revenue and few order lines. Here you can consider a lower service level, ordering the item only when a customer asks for it, or phasing it out.
The matrix lets you set service level targets, safety stock and reorder rules by segment instead of one target for the whole inventory. An XYZ analysis can add how stable demand is. When demand is predictable, reaching a high service level requires less safety stock.
But the segment is a starting point, not the final answer. Ask these questions before you lock in the level:
Even a CC item may need a high service level if one missing screw can hold up an important customer's entire order.
When inventory levels are too high, it looks like a service level problem. But the real cause may lie elsewhere.
Take the component again, but let the lead time vary. The supplier delivers in one month on average, with a standard deviation of about a week (0.25 months).
When variation in demand and lead time are independent of each other, the formula can be extended like this. It is one of several recognised models, and it still assumes roughly normally distributed variation:
Safety stock = z × √(lead time × σdemand² + (average demand × σlead time)²)
Under the square root, demand contributes 1 × 120² = 14,400 and lead time contributes (400 × 0.25)² = 10,000. At 98%, safety stock in the model rises from about 245 to about 320 units. That is about 75 extra units and €2,600 more in safety stock, simply to maintain the same service level target when lead times become less predictable.
You are compensating for supplier uncertainty rather than raising your service level.
Compare that with the table. Moving from 95% to 99% tied up about €2,900 more. In the example, an unstable supplier ties up almost the same amount before you have raised the service level at all.
That is why supplier performance belongs in the discussion about inventory balance. A supplier portfolio analysis can show where lead times vary, and where that uncertainty may increase the need for safety stock.
And then there is inventory driven by something other than the service level. Extra inventory held because of a supplier's minimum order quantity is not automatically safety stock. And items without sales are not necessarily surplus.
A spare part can sit for years and still be critical the day a machine breaks down. But inventory that has stopped selling and serves no clear purpose is not protecting your customers. If you cut service levels to hit a working capital target while that inventory stays put, you risk hurting your customers without solving the problem.
The calculation for one component is relatively simple. The hard part is doing it for thousands of SKUs with actual lead times, and keeping it up to date as demand changes. With Inact's inventory optimisation, you categorise your products and adjust service levels, safety stock and reorder points to the category each one falls into. That makes it easier to see where inventory supports customer service and where it ties up working capital without adding value.