Home / Blog / Supply Chain / Service Level vs Fill Rate
Service Level vs Fill Rate: Which Metric Should Drive Your Safety Stock Decision?
By Zeeshan Abbas . Reviewed by Rimsha Nadeem Anwar (Six Sigma Black Belt) . September 2026
In short: Cycle service level (CSL) measures the probability of not stocking out during a replenishment cycle. Fill rate measures the fraction of demand fulfilled from available stock. The two metrics reward different behaviors: a 95% CSL tolerates one stockout in twenty cycles regardless of how much demand is lost, while a 95% fill rate guarantees that 95 cents of every dollar of demand ships. For most distribution environments, fill rate is the better business metric, but CSL is simpler to compute and more common in textbooks. For the DC item used throughout this series, a 95% CSL gives Z=1.65 and safety stock of 172 units; reaching 95% fill rate on the same item typically requires slightly less safety stock because fill rate accounts for the shape of demand shortfalls, not just whether they occur.
Two planners at the same company can each claim a 95% service level and be measuring completely different things. One counts cycles without a stockout. The other counts units shipped versus units ordered. The gap between those measurements can be enormous, and the safety stock each requires differs in ways that matter for capital planning.
Understanding which metric you are actually using, what each one requires in safety stock, and when the gap between them becomes large enough to change decisions is practical knowledge, not academic. This guide covers the mechanics, the formulas, the tradeoffs, and how to connect either metric back to the safety stock and reorder point numbers developed earlier in this series.
What cycle service level measures
Cycle service level, often abbreviated CSL or alpha-service, answers one binary question for each replenishment cycle: did you run out of stock or not? A 95% CSL means that in 95 out of every 100 replenishment cycles you complete, the item stays in stock from the moment a new order is placed until the new order arrives.
CSL makes no distinction between a stockout that loses one unit of demand and one that loses five hundred. A cycle that runs out of stock with one day left until replenishment counts the same as a cycle that runs dry on day two of a four-day lead time. The metric cares only whether a stockout event occurred, not how large it was.
The Z-score approach to safety stock targets CSL directly. When you set Z=1.65 and compute safety stock as Z times the standard deviation of demand during lead time, you are targeting approximately a 95% CSL. The formula assumes normally distributed demand, and the probability that demand exceeds safety stock plus expected demand is the complement of the CSL.
CSL = P(demand during lead time <= reorder point) = Phi(Z)
For the DC item in this series: demand during lead time averages 400 units, sigma_dLT = 104.4 units, Z = 1.65, safety stock = 172 units, reorder point = 572 units. The probability that demand during lead time exceeds 572 is about 5%, so CSL = 95%.
What fill rate measures
Fill rate, often called beta-service or FR, measures the fraction of total demand served directly from available stock without delay. A 95% fill rate means that for every 100 units ordered by customers, 95 ship immediately from stock and 5 are either backordered or lost.
Fill rate integrates the size of shortfalls, not just whether they happen. A cycle that runs out by 10 units counts worse than one that runs out by 1 unit. This makes fill rate a better representation of customer experience and revenue impact than CSL, because what customers feel is not the probability of a stockout but the fraction of their orders fulfilled on time.
Computing fill rate from safety stock requires the unit normal loss function, denoted L(z) or G(z), which captures the expected shortage per cycle given a normal demand distribution.
Fill rate = 1 – (sigma_dLT x L(Z)) / Q
Where Q is the order quantity (the EOQ), sigma_dLT is the standard deviation of demand during lead time, and L(Z) is the unit normal loss function evaluated at the chosen Z. For the DC item: sigma_dLT = 104.4, Q = 632 (EOQ from post one of this series), and for Z = 1.65, L(1.65) is approximately 0.0206.
Fill rate = 1 – (104.4 x 0.0206) / 632 = 1 – 2.15 / 632 = 1 – 0.0034 = 99.66%
That result reveals a common misconception: a 95% CSL often delivers a much higher fill rate. On this item, targeting 95% CSL gives roughly 99.7% fill rate. If you are targeting 95% fill rate and using a 95% CSL formula to get there, you are carrying significantly more safety stock than the fill rate target requires.
The unit normal loss function
The unit normal loss function L(Z) is the key to connecting Z-scores to fill rates. It equals the expected number of standard deviations of demand that will go unfilled per cycle, assuming demand is normally distributed and the item is managed with a reorder point system.
| Z | CSL | L(Z) | Fill rate (Q=632, sigma=104.4) |
|---|---|---|---|
| 0.00 | 50% | 0.3989 | 93.4% |
| 0.84 | 80% | 0.1120 | 98.2% |
| 1.28 | 90% | 0.0470 | 99.2% |
| 1.65 | 95% | 0.0206 | 99.7% |
| 2.05 | 98% | 0.0068 | 99.9% |
| 2.33 | 99% | 0.0027 | 99.99% |
The table shows that for the DC item, even Z=0 (zero safety stock) delivers 93.4% fill rate. This happens because the order quantity is large relative to the demand variability. When Q is large, individual stockout events are small relative to total throughput, so fill rate stays high even at low Z values.
The ratio sigma_dLT / Q is the key driver of how much safety stock you need to hit a given fill rate. When this ratio is small (large EOQ relative to demand variability), fill rate targets are easy to hit and the fill rate curve is flat. When the ratio is large (small EOQ relative to variability), fill rate drops sharply with lower safety stock and CSL and fill rate diverge more.
When CSL and fill rate give different answers
CSL and fill rate diverge most when three conditions hold: the order quantity is small, demand variability is high, or the target service level is modest (below 95%). In those cases, the Z-score that satisfies a CSL target may deliver a fill rate well below the business expectation, or the Z-score needed for a fill rate target may be far lower than what a CSL formula suggests.
Consider a high-variability item with sigma_dLT = 300 and Q = 200. At Z = 1.65 (95% CSL), L(1.65) = 0.0206, giving fill rate = 1 – (300 x 0.0206) / 200 = 1 – 0.031 = 96.9%. Now suppose the target is 90% fill rate. Setting fill rate = 0.90 and solving: 0.10 = (300 x L(Z)) / 200, L(Z) = 0.0667. Looking up L(Z) = 0.0667 gives Z about 1.00, corresponding to 84% CSL. A planner using the 90% CSL formula would set Z = 1.28 and carry unnecessary safety stock, while the fill rate calculation reveals the true Z needed is only 1.00.
For the DC item in this series (sigma_dLT = 104.4, Q = 632), the ratio sigma_dLT / Q = 0.165 is small enough that CSL and fill rate move in step. The practical implication is that this item is forgiving: either metric gives a workable answer. High-variability, small-lot items are where metric choice matters most.
Which metric to use in practice
Fill rate is the better business metric for three reasons. First, it maps directly to customer experience and revenue: a 95% fill rate means 5% of demand is delayed, which is a number finance and operations can reason about together. Second, fill rate accounts for the magnitude of shortfalls, not just their occurrence. Third, for most distribution items, a fill rate target requires less safety stock than the equivalent-sounding CSL target, freeing capital.
CSL has practical advantages too. It is simpler to compute, requires no knowledge of the order quantity, and works directly with the Z-score and sigma_dLT that most inventory systems already store. CSL is the standard in most supply chain textbooks and ERP systems for this reason. If your WMS or ERP system calculates safety stock using Z x sigma_dLT and asks you to set a “service level,” it is almost certainly computing CSL.
A reasonable operating rule: use CSL for day-to-day safety stock calculations because the computation is straightforward, but validate fill rate quarterly to check whether the CSL target is delivering the business outcome you expect. If the calculated fill rate from your CSL-based safety stock is materially above the fill rate target, you may be carrying excess buffer.
Connecting to the series numbers
In this series, Z = 1.65 and safety stock = 172 units correspond to a 95% CSL on the DC item. The fill rate delivered at that safety stock level is approximately 99.7%, far above any realistic fill rate target. That gap exists because the EOQ of 632 units is large relative to sigma_dLT = 104.4 units. The sigma_dLT / Q ratio of 0.165 means stockout events, when they occur, are small relative to cycle throughput.
If this DC item were managed to a 95% fill rate target using the fill rate formula instead of the CSL formula, the required Z would be much lower than 1.65, the safety stock would be substantially less than 172 units, and the reorder point would drop well below 572. For the class A classification given in the ABC analysis post, however, a 95% CSL is a reasonable policy choice on its own merits regardless of the fill rate arithmetic.
ABC class and service level policy
Service level targets should vary by ABC class. The ABC analysis post established that this item is class A with an annual usage value of $365,000. Class A items typically carry the highest service level targets because stockouts are most costly in revenue and customer relationship terms.
| ABC class | CSL target | Fill rate target | Z | Policy rationale |
|---|---|---|---|---|
| A | 97-99% | 99.0-99.5% | 1.88-2.33 | High AUV, visible to customers |
| B | 90-95% | 95-98% | 1.28-1.65 | Moderate AUV, some flexibility |
| C | 80-90% | 85-95% | 0.84-1.28 | Low AUV, simple rules acceptable |
The DC item at Z = 1.65 sits in the middle of the class A CSL range. Raising it to Z = 1.88 (97% CSL) would increase safety stock from 172 to 196 units and reorder point from 572 to 596. The additional 24 units of safety stock is a deliberate choice for a class A item given this analysis.
Three expert tips
Calculate both metrics and report the gap
If your current safety stock policy uses CSL targets, run the fill rate formula on each item quarterly. Items where fill rate is more than five percentage points above the CSL target are likely carrying excess safety stock. Items where fill rate is below the CSL level have a sigma_dLT / Q ratio large enough that CSL is a misleading metric for that item.
Use fill rate when order quantities are small
For items with small EOQs relative to demand variability, the CSL formula overstates the safety stock needed to hit a given business service target. Any item where sigma_dLT / Q exceeds 0.5 should have its safety stock policy reviewed against a fill rate target, not just a CSL target.
Align the metric to how you measure stockouts
If your operation counts stockout events (did we run out this week?), CSL is the natural metric. If your operation counts units or revenue shorted, fill rate is the natural metric. The metric should match your measurement system; otherwise, you optimize for the wrong thing. Most modern WMS systems can track both, and the cost of running both calculations is low once the data infrastructure is in place.
Free supply chain calculators
The Service Level Calculator converts between Z-scores, CSL, and fill rate for a given sigma_dLT and order quantity. Use it alongside the Safety Stock Calculator to find the Z that hits your fill rate target directly, the Reorder Point Calculator to convert the result into a trigger level, and the EOQ Calculator to verify the order quantity assumption that underlies the fill rate formula. The ABC Analysis Calculator determines which service level tier each item belongs in. All tools are at the Supply Chain hub.
Frequently asked questions
What is the difference between service level and fill rate?
Service level (cycle service level or CSL) is the probability of not stocking out during a replenishment cycle. Fill rate is the fraction of total demand served from available stock without delay. CSL is a yes/no per cycle measure; fill rate integrates the size of shortfalls across all demand.
Which is better: service level or fill rate?
Fill rate is generally the better business metric because it maps directly to customer experience and revenue. CSL is simpler to compute and more common in ERP and textbook formulas. Use CSL for day-to-day operations and validate against fill rate periodically to check whether the CSL target is delivering the business outcome you intend.
Can a 95% service level give a very different fill rate?
Yes. A 95% CSL can deliver a fill rate of 99.7% or higher when the EOQ is large relative to demand variability. Conversely, when sigma_dLT / Q is large, 95% CSL may deliver a fill rate well below 95%. The unit normal loss function and order quantity determine the relationship.
What is the unit normal loss function?
The unit normal loss function L(Z) is the expected number of standard deviations of demand that go unfilled per cycle given a normal demand distribution and a Z-score safety stock policy. It is used in the fill rate formula: fill rate = 1 – (sigma_dLT x L(Z)) / Q.
How do I calculate fill rate from safety stock?
Fill rate = 1 – (sigma_dLT x L(Z)) / Q. Find Z from your safety stock (Z = safety stock / sigma_dLT), look up L(Z) from a standard table, and plug in sigma_dLT and Q. For the DC item: Z = 172 / 104.4 = 1.65, L(1.65) = 0.0206, fill rate = 1 – (104.4 x 0.0206) / 632 = 99.7%.
What Z-score do I need for a 95% fill rate?
It depends on sigma_dLT and Q. Set fill rate = 0.95 and solve: 0.05 = (sigma_dLT x L(Z)) / Q, L(Z) = 0.05 x Q / sigma_dLT. Look up the Z corresponding to that L(Z) value. For the DC item: L(Z) = 0.05 x 632 / 104.4 = 0.303, which corresponds to Z near 0, meaning zero or near-zero safety stock satisfies the 95% fill rate target for this item.
How does ABC class affect service level targets?
Class A items typically target 97-99% CSL or 99-99.5% fill rate due to high annual usage value and customer visibility. Class B targets 90-95% CSL or 95-98% fill rate. Class C accepts 80-90% CSL or 85-95% fill rate, where simple rules like fixed-days coverage or a two-bin system are acceptable.
What is the sigma_dLT to Q ratio and why does it matter?
The ratio sigma_dLT / Q determines how sensitive fill rate is to safety stock. A small ratio (below 0.2) means CSL and fill rate move closely together and stockout events are small relative to cycle volume. A large ratio (above 0.5) means fill rate drops sharply with lower safety stock and CSL becomes a misleading guide for business decisions.
Does fill rate require a specific order quantity?
Yes. The fill rate formula includes Q (the order quantity) because fill rate depends on how much total demand passes through each cycle. CSL does not require knowledge of Q, which is one reason it is simpler to compute. When using the EOQ, plug the EOQ value in for Q.
How do I target a specific fill rate in my safety stock formula?
Rearrange the fill rate formula: L(Z) = (1 – target fill rate) x Q / sigma_dLT. Solve for L(Z), look up the corresponding Z from a unit normal loss function table, then compute safety stock as Z x sigma_dLT. This gives the exact safety stock needed to hit the fill rate target, which may be lower than the CSL-based approach.
Is cycle service level the same as in-stock rate?
Approximately yes. In-stock rate is the fraction of time an item is available. CSL is the fraction of replenishment cycles without a stockout. For items with short cycles relative to the review period, the two converge. For items where cycles are long (slow movers), they can diverge because a single stockout cycle may span a large fraction of calendar time.
How often should I review service level targets?
For class A items, quarterly. For class B, semiannually. For class C, annually. Also review after any significant change in demand, lead time, or business priority. The calculation is fast; the bottleneck is usually collecting updated sigma_dLT from recent data.
CSL and fill rate measure the same underlying problem from different angles. CSL asks whether a stockout happened; fill rate asks how much demand was served. Both require safety stock to improve, but the amount differs by item, and using the wrong metric for your business can mean carrying more inventory than necessary or delivering less service than you think. Run both calculations, pick the metric that matches how your operation experiences stockouts, and revisit when inputs change.