CUSTOMER-APPROVED INVENTORY

Get out your thinking caps this week for the 8th Edition of HOW PREDICTABLE. This week you will learn to construct a matrix of colours and sizes and will best represent the product line you are displaying; tested by what historical level of inventory precipitated what purchases. Yes, it is the historical level of inventory you had at the very moment you sold the most that you want to discover and set as your model, in effect using your customer's own behavior to set your floor.

The Portable Demand Stream

The traditional issue we have had with data is that is often contaminated with lost sales and transferred sales when some products were not available. Having two Medium T-shirts in the back room and not displayed certainly will happen. The problem is most forecasting algorithms cannot tell the difference between your first choice that is not displayed and the second choice that you settled for. In clothing, that will likely be of the same SIZE but different colour, as your fit is pretty constant. Add onto that different assortments in different stores and different availability at different times, you get a bit of a complex matrix of what is available where for each customer.

The Weekly View

In previous editions, you managed to capture your historical inventory position by item/store by week, and you will need to leverage this data once again. Combine this again with the weekly sales and what you will get is a timeline of sales vs Inventory by size/colour for a given location in a given timeframe. You may want to do this exercise quarterly to start, but use the entire season of historical data to model.

Logistic Regression Testing

This is where your magic is going to manifest itself. Your LR model will start to be able to investigate the levels of INV you had at times your sales were at their optimal. By now, if you have been doing your homework, you have discovered that you likely have much more inventory than you actually needed to sell, but you now have something even more powerful. You can now start to measure the impact of NOT HAVING one colour and how that transfers to the secondary and tertiary items in the customer's purchase path. The Logistic Regression builds for you how much more you would need in each of the remaining size/colours to account for the shift in demand that will occur.

Now, let's presume you have a simple Available To Promise date calculator in your supply chain that can show you when the store can best expect its next shipment of a missing colour. The matrix can recalculate to see if it is optimal to increase the most likely second colour/size temporarily or merely wait out the shortage.

The Availability Your Customer's Demand

What you now have is a very easy-to-manage model stock for size and colour that has been proven to drive the most positive purchase trigger for your customers. It can be teamed with a traffic counter or high-level sales forecast to adjust up or down the model of planned sales. It accounts for anticipated shortages in supply and counters them with what customers have shown to be their traditional second choice BY LOCATION. You also have a tool that lets you play with dropping various colours or sizes from certain stores and maintaining the optimal number of each of the remaining to maximize sales productivity.

The application

The example I used is T-shirts but can be used in any varied assortment business. The process creates the spreadsheet below; a clear model by size and colour that can recalculate based on what you hope to sell and what you have available. Simply program your replenishment system to this model stock and let it do the work it needs to maintain it. You no longer need to worry about individual sku/store combinations, but rather what does the customer see as a wholistic view of the shop. Priority routing can be assigned by stock-out costs or cost of alternatives. You costs are being driven by a model that is going to work for the customer, as they have OK'ed that level historically.

Big data, AVAILABLE TODAY, that can get you Customer-Approved Stock Levels sounds pretty enticing. Reach out if you need some assistance setting up the tasks. Better answers come quickly as your focus shifts to better forecasting.

LIKE, SHARE, and SUBSCRIBE to spread the evangelism of HOW PREDICTABLE.

Jerry Marion