Why Basic Inventory Data Isn't So Easy
There are three things you must do – and one reality you must accept – when managing inventory data…if you want to get your inventory right.
Getting the basic inventory data right is a lot of work.
Henrik Danford-Klein reminded me of this in a great conversation. Henrik is a supply chain executive who’s spent many years managing, analyzing, and trying to reduce inventory. He had great lessons and stories that matched my experience.
Here are some lessons for all of us to keep in mind:
First, you might have multiple demand streams: Inventory formulas have a single value for demand. This works fine in many cases, but not all.
For example, you may have a set of customers that consistently order daily or weekly and others with big orders monthly or quarterly. Or, you may have customers that you want to serve with a much higher fill rate.
You want to analyze and track the different demand streams in such cases. For example, you may want to set aside inventory for one group or have different promised delivery times.
Unfortunately, this will require some work to think through. For example, you may miss out on needed risk pooling if you have two (or more) independent inventory calculations. On the other hand, if you have one calculation, you need to think through the rules and when it is ok to break them.
Second, watch for the bullwhip effect: If a company sells directly and through many channels, you may need to filter to determine the actual demand.
This can happen if there is a shortage. In this case, the customer may place multiple orders through different channels. These orders can look like a lot of missed demand. But, of course, it is not all real. As soon as the customer gets the product they want, they cancel the other orders. This is one of the causes of the bullwhip effect.
You’ll want a system to filter this demand out. If you don’t, it can lead to too much inventory or, worse, unneeded investments in production capacity.
Third, watch out for calendars and time buckets. Henrik mentioned a company where inventory would spike every February. No one knew why. This was a business that didn’t have much seasonality.
At first, the spike was blamed on the automatic adjustments of lead times. There would always be delays at the end of the year. Since lead times showed an uptick, maybe this led to extra inventory in February.
Lead times weren’t to blame. Although, this does give another lesson: Don’t be too dynamic in changing parameters if you are catching more noise than signal.
Instead, the problem was that February had fewer days. The planning systems were based on monthly buckets. Then, specific demand was allocated to the days, and each day in February got too much demand.
The good news is that once you know about this, it is easy to spot, and there are fixes. The fixes are converting to just using weeks as the primary time bucket or creating your own definitions of months (and explicitly calling out the four- and five-week months).
The bad news is that it can be a headache fix. This is because we tend to think, plan, and have other important systems that work in monthly buckets.
Fourth, if you calculate values automatically, put in some safety caps. There will always be anomalies in the data. Anyone who calculates the forecast error, lead time, or lead-time variability will have stories about one of these values spiked because something funny happened with the data.
So, when you build these systems, you want to include a process to either cap the amount a value can change from period to period or only make a large change if it gets the proper approvals. Of course, you should also build in periodic audits to ensure the data stays clean.
###
