Welcome to another addition of the Numerical Insights newsletter. This month's newsletter is dedicated to the many challenges of maintaining inventory.

I was recently approached by a company who has inventory challenges. The demand for their products is slowly on the rise, but they find they are continually running out of component parts in the inventory. These inventory component parts are used to assemble their products. The bottom line is: the orders are coming in but the product is not going out the door at the same pace. Orders are being delayed for lack of component parts in their inventory.

The first challenge in a study of this nature is to determine what data is needed. The company has inventory stocking data and this is an inventory study, so inventory data is what we need, right? Not quite. Studying the inventory data at this point would show us what we already know. The company runs out of parts and doesn't know why. Instead, we'll look at the customer demand data to try to understand customer order behaviour.

The first challenge is to get the data. The customer's system can only export about 65,000 lines of data at a time and only keeps 24 months worth of demand records. The data was exported in 4 pieces, cleaned and reassembled into a Microsoft database with 145,000 records. Now we can do some analysis.

A high level query of the data reveals that the customer offers just over 7,000 products. A further query revealed that over a 2 year period (the time-frame of the exported data), 34% of these products were only ordered once! Think of the amount of support and inventory involved to stock these products for the odd chance that someone will order one of these products sometime in the next 2 years. It was recommended to the company to look into having the suppliers of these components stock the parts on consignment. It was further recommended that the company look into these products and see of any of their other existing products could cover the application, thus allowing for elimination of product numbers.

For the remaining 66% of the products that are ordered more often, a study over the 2-year time-frame was conducted to show the minimum and maximum monthly order levels. This provides the company with an idea of which products are ordered on a steady basis (in which case recommended inventory levels are fairly steady) and which ones swing up and down substantially in demand.

The final analysis conducted was to include confidence intervals at a 90% level for monthly demand. This provides a guideline on stocking inventory levels as it determines the inventory level for which 90% of the historical demand levels for each part fall within that inventory level or below. The company will now evaluate how this will impact their current inventory levels to see if a 90% level is acceptable.

In the future, for the highest moving products, a full seasonal forecast may be developed. This newsletter describes the first level of analysis conducted in order to gain useful numerical insights into a large data set since it would not be practical to do full seasonal models on 7,000 products.

I hope this edition of the Numerical Insights newsletter has provided yet another example of how you can use the Power of Information to make data-driven decisions for business. As always, I welcome input from my readers and requests for information.

Until next time,
                          Tracey Smith
                          www.NumericalInsights.com