While one of the basic processes in successful inventory management is inventory classification, as we discussed in our Inventory Classification 101 article, there are some additional considerations for practitioners of best-in-class inventory management that can help make sure optimizing your inventory investment.
Inventory Classification Fundamentals
The fundamentals of inventory classification are always helpful, whether or not you are using them to help determine service levels, and subsequently your safety stock levels as well (which we highly recommend you do). As a brief review, the best way to do this is:
- Gather up information on all your parts by some grouping criteria, usually by site or warehouse, but occasionally by product group or some other delineation.
- Sort and rank the items within each group according to a measure or measures you define, such as hit frequency, quantity sold, margin, etc.
- Divide the resulting list up into A/B/C classes by accumulated totals of your chosen measure. Usually the items making the top 80% of your measure should be A’s, the next 15%, and so on, but these are not hard-and-fast limits.
Once you’ve done this, you will have a list of items that are each assigned a class such as A,B,C, etc, including “X” for items that have no activity at all and may need to be placed on an obsolete or dead list.
But that’s not where the story ends. With apologies to George Orwell, all A’s are equal, but some A’s are more equal than others.
Beyond the Service Level Target
Certainly it is true that if one A item has qualified to be in your top 80%, it is like the others in that most times this means it will have the same service level target assigned to it as any other A item. But that doesn’t mean they’re equal in every other way. While planning software such as StockIQ will attempt to achieve that target service level, different “A” items that show different patterns of demand history can require drastically different stocking and ordering strategies.
Take for example one of your best hypothetical A items: It sells 1,000 units every month, regularly like clockwork. Forecast error is minimal, so safety stock required is low, it can be purchased regularly and predictably, and as a result, this item requires very little attention, costs very little stock and sell, and should therefore be a great moneymaker. We should all be so lucky.
But across the aisle on another shelf is a second A item. This item has the exact same average demand as the first item, 1,000 units per month. But, one month it sells 5,000 units, the next only 100, the following 3,000, and so on. Same average level of demand, but a much different pattern of behavior. Most standard safety stock methods will fail here and lead you to realize much lower than your target service level without special consideration, and will likely lead to stock outs as well as higher carrying costs.
Two Dimensional Classing
It is for this reason that we recommend calculating a second dimension in you inventory stratification - what we call the XYZ analysis. Rather than a measure of the part’s volume, based on a measure like hits, quantity sold, revenue, and so on, this is a measure of the item’s volatility, variability, or predictability.
The most simplest approach to this is by measuring the coefficient of variation (cv) or standard deviation of the demand history of the part over a representative period, such as the last 12 months. Low cv means the part’s demand is more predictable, whereas a high value above 1 or more usually means its demand is all over the map.
The next level up, such as if you have a system like StockIQ, is that in which you can also choose to measure your XYZ percentages based on the forecast error instead of the pure variance; If the part is highly forecastable, despite being highly variable, then for your purposes, it is a predictable item. Additional approaches such as classifying by a discretely calculated usage pattern (e.g. recurring vs sporadic) are very valid as well.
Once you have calculated these values for all of your products, you then you can assign them an X (predictable), Y, or Z (highly unpredictable) class based on a score, and breakpoints that you determine.
After you’ve gone through this process, each one of your parts will have a pair of class assignments, such as “AX” or “CZ” to help you decide not only what your target service level should be for an item, but also if and how to handle making sure that safety stock and order policies are in place to achieve that target goal based on the item’s XYZ class.
Having added this additional informational arrow to your inventory stratification quiver, you will find it helps provide additional insight and clarity into achieving your inventory service level goals, as well as providing guidance on how and why if you do happen to miss those goals.