Service level is one of the most important but least understood supply chain concepts. Service level plays two important roles in inventory management:

The problem is that the metrics used for the first are usually not the metrics used for the second. And although they might sound similar, they are often measuring something very different.

What impact does this have on your supply chain and what should you do about it?

Here are some of the most common supply chain service level metrics:

Different service level metrics suit different circumstances. For instance, OTIF might be the favoured measure for customer service, but as OTIF is dependent on transportation as well as stock availability, it is not ideal for measuring inventory performance. Or, to take another example, ready rate is a good measure for spare parts.

Choice of metrics is often driven by what is easy to measure and what is familiar. While these are valid considerations, you may be missing opportunities to add value by using other metrics. Cover is in very common use, despite having a very limited and questionable value for service levels, whereas back orders and out of stock time are less commonly tracked, despite being very valuable metrics in inventory optimization.

Whatever metrics you use, it is essential to define them precisely and uniformly across your organisation.

Whenever you talk about, say, back orders, then everyone should measure and understand them in the same way. (Do back orders include items that are on order but not yet late, or only overdue items? Are you measuring back orders at the end of a cycle, at the beginning of a cycle or at regular time intervals not aligned with the cycle? And so on.)

The most appropriate metrics for your organisation will depend on a number of factors. However, it is always good to review what you use. New metrics might provide additional insights or drive further improvement.

If you only use service levels to measure the performance of processes then it is enough to use those service measures carefully and accurately.

However, if you want to optimize your inventories then you also need service level as an input to planning decisions such as setting re-order points or safety stock levels. Planning technology usually either assumes a service level, or more likely requires a service level as an input. It is very important to understand exactly what service level your planning tools use and how it relates to your operational KPI’s.

Most tools use metrics like fill rate or cycle service level (or variations of these) that are usually not in use as operational KPI’s. This means that you need to be able to translate between the metrics you use to measure your performance and the metrics used for planning, or there is a major risk of sub-optimal outcomes.

All of the service level metrics listed above are interrelated and can be calculated from each other if all the right data is present. But their interrelationship is not simple. Confusing them can cause serious problems.

Let us look at three service level metrics to illustrate the differences: OTIF, fill rate and cycle service level.

Let’s say you want to consistently achieve an OTIF rate of 99% to your customers.

You might be able to deliver that 99% OTIF rate with a finished goods fill rate of 0%.

Let’s also say that you want to have a fill rate of 99% for the raw materials needed. You might require a cycle service level of only 30% to achieve that.

The first thing that should be apparent is that OTIF, fill rate and cycle service level are measuring different things. The relationship between them is neither linear nor intuitive. So what is going on in our example?

Fill rate sounds like OTIF, but it’s measuring something different. In our example, the production lead time is shorter than the customer promise time, and you choose to make the product to order. You have a fill rate of 0%, since you never hold it on stock but make it when needed and dispatch it as soon as it is ready. But the OTIF is 99%, since your customers almost always get it on time.

You aim to keep the raw materials on hand for whenever your customers place an order. The fill rate for the raw materials is 99%. You order large batches of raw materials at a time. With a long order cycle, you only need a cycle service level of 30% to deliver a fill rate of 99%. (The exact numbers are an example, not a rule!)

You will notice that in this example we are actually measuring different points in the supply chain: OTIF is measured at the customers’ premises, fill rate is measuring both finished goods and raw materials, cycle service level is measuring raw materials.

Let’s dig into the three metrics a little deeper.

OTIF is a very common performance measure for deliveries to customers (as well as deliveries from suppliers). While the example we gave is extreme, there are plenty of instances imaginable where OTIF is quite different from fill rate. We worked with a retailer who had problems with on shelf availability that were almost entirely down to picking and last mile logistics: fill rate at the distribution centre was consistently well above 99% but only 85% made it to the retail stores on time and in full.

OTIF is an important supply chain performance indicator, but has limited value in inventory management.

The cycle service level is used in many applications, especially for setting safety stock levels, and is easy to calculate. Equations such as Z×σ(demand) ×√LT and its derivations use cycle service level.

Cycle service levels are not comparable between items with different order cycle lengths. If you order an item daily, then a cycle service level of 99% means on average you will be out of stock once every 100 days. If you order an item every 12 months, then the same 99% cycle service level would mean you are out of stock only once every 100 years!

Note also that cycle service level does not measure how great any stock out may be, just how likely you are to stock out in a cycle.

Cycle service level percentages therefore tend to be quite unintuitive for most people: deliberately targeting a percentage as low as 30% just feels wrong.

Fill rate is much closer to the percentages people have in mind when they think about supply chain performance. A fill rate of 99% means that you can fulfil 99% of demand directly from stock, regardless of the cycle time. It is useful for comparing performance across items and it is sensitive to the size of any stock outs.

The only real downside of fill rate is that it requires much more work mathematically than cycle service level.

A major pitfall arises if target fill rates are used in applications that require cycle service levels. The fill rate will almost always be higher than the cycle service level and sometimes substantially higher.

It is very important to understand the differences between metrics. The difference between OTIF and fill rate is normally quite well understood. What is not always so well understood, is what service level metrics planning tools use and what service levels should be set at different points in your supply chain to deliver the desired performance without overstocks.

So what should you do?

The first step is to look carefully at your planning tools and methods and make sure you understand exactly what service levels are being used. (We have found, unfortunately, that documentation is often unclear.)

The second step is to define what good looks like. This may already be clear from your existing operational KPI’s, but you may want to update the metrics you use.

The third step is to build a bridge between the first two. Let’s say you have defined success in terms of out of stock time and that your planning tools use fill rates. You can calculate one from the other, although, it should be admitted, not at all easily. The same is true of all the metrics listed above. They are all related, but you need to be careful in translating one to another.

It is quite easy to get bogged down in technical differences, or dispirited by them, but as with all things related to inventory optimization, we like to say that being approximately right is better than being precisely wrong. All “optimal” levels are approximations since variability is unknowable in advance. But it is essential to understand the big differences, such as that between cycle service level and fill rate. And it is therefore vital to understand what service levels are being used by your tools.

Supply chain is complex enough without looking for further complications, but service levels are a prime example of something seemingly simple hiding a number of potential pitfalls. Metrics are very important in how organisations operate. Even if targets aren’t always met, metrics influence how you work, what you prioritise, even how you see the world.

If your inventory metrics are limited to cover, inventory turns and OTIF, however, then you are missing out on a huge potential to drive improvement. Service level, in its most generic sense, is almost always prioritised over leanness of inventories as the primary measure by which inventory performance is judged. But service level can refer to a broad range of actual metrics, each of which has a time and a place. To get full value from service levels, they need to be well understood and applied with care.

Safety stocks represent one of the biggest opportunities to optimize inventories. Service level is an essential concept to understand if you want to optimize your safety stocks. See also our guide to safety stocks.

In the absence of a method factoring in service levels, safety stocks might be set based on some kind of rough estimate. For example, your policy might be to maintain 2 weeks’ average sales as safety stock for finished goods. The exact size of the safety stock might be grounded in some kind of consideration, such as usual lead times, maximum experienced demand, or similar. This type of approach has the advantages of simplicity and comprehensibility. But if you don’t factor variability properly into the equation, you will have a very substantial opportunity to further optimize your inventories.

The next level of maturity is to factor variability into your calculations. Safety stock is by definition there to protect you against various types of variability. This is a positive step, but needs to be treated carefully.

Cycle service level is normally a good place to start, since it is mathematically simplest. Spreadsheet-led approaches often use cycle service level since the equations are reasonably well known. There are some important limitations to understand, in particular that cycle service level approaches only work accurately with normally distributed demand.

Normal distribution may frequently be assumed in theoretical inventory models, but is encountered less frequently in reality. It is a good assumption for heuristics covering overall inventory, thanks to the central limit theorem. It is also not a bad approximation for most individual items, unless they display very variable and/or sporadic demand.

While it is in many respects inferior to fill rate, a lot of progress can be made using cycle service level, as long as some allowances are made for the normality or otherwise of distribution. We recommend an incremental approach. If your cycle service level calculation is suggesting safety stocks 50% lower than your current ones, don’t halve them straight away, but try lowering them by steps. And in a first instance leave aside items with sporadic demand.

The other hurdle you have to overcome is working out what cycle service level percentage you need. Especially for items with a long order cycle (although this is not the only consideration), the percentage might be much lower than you would imagine.

The next level of sophistication is to work with fill rates instead of cycle service level. This has some clear advantages, once the additional mathematical complexity is overcome.

Probably the biggest advantage from a practical perspective is that fill rate percentages are more intuitive and comparable across items. You could, for instance, reasonably set a target fill rate of 99% across all items, whereas the equivalent cycle service level percentages are likely to vary considerably by item.

Another advantage of fill rate is that it is not dependent on normally distributed demand, but can be made to work with various distributions, including non-parametric. (Which is not to say that it always is adjusted properly. Just because your tool works with fill rate or similar, it does not necessarily mean that it is factoring the actual distribution of demand in.)

Using fill rate will allow you to set safety stocks in a more exact and aggressive fashion. It can even work better than the cycle service level approach for the normal distribution, especially if order quantities are low. But the additional complexity of working manually with fill rate means that we would not recommend it before getting a good understanding of cycle service level.

Some planning technology gives you the option to set a service level target and then factors it into re-order point or safety stock calculations. This is a great resource to have and prevents you having to wade through the calculations yourself, but be aware of the exact type of service level used, be aware of the potential delta between whatever the technology uses and your operational KPI’s, and be aware that the tools are most likely making various assumptions, such as distribution of demand, that will affect their accuracy to a greater or lesser degree for each inventory item.

Most practitioners know from experience that their tools don’t set good levels for at least some of their inventories. Developing a deeper understanding of service levels is likely to help you understand one of the major reasons why that might be the case.