Product selection is a common problem in selling operations. Here are the pain points. Companies, especially those in manufacturing and trade industries, typically have a large amount of assets tied up in inventory, and a huge portion of which might not sell well, and thus be unproductive. Holding such inventories is a waste of money. As they also occupy the space which could have been used to sell other more productive items. Analytics can help optimizing product offerings by identifying and replacing less productive items with more productive ones. Using the liquor and drinks at Martini as an example, statistic shows that 23 percent of restaurant failed in the first year of their operations. Because liquor and drinks are the most profitable items. A restaurant must identify the right items to sell, thus ensure its survival. A simple one-dimensional analysis can rank all items by their total gross profits from high to low, to identify those that should continue and those that should be replaced. However, this can be misleading because an item can achieve a high total profit in two ways, either by a high unit profit or by a high volume. Given the same total profit, the restaurant might prefer the former which is easy money, than the latter which is hard money. Because the latter may take more space and effort to manage. Thus, a two-dimensional analysis on both the unit profit and demand volume is better. We can identify the cash cows, that is, the items with both high unit profit and high volume, and commodities, the items with high volume, but lower unit profit. We can also identify the luxuries, the items with high unit profit, but lower volume. Finally, the laggards, items with lower volume and lower unit profit, they should be discontinued. The third typical problem which we'll discuss in sell analytics is pricing and promotion. This is a popular problem in practice. Companies often offer price discounts to customers in a hope to increase demand. The pain points are that sometimes the demand does not increase at all, and thus, the companies just lose the profit. Some other times, the demand might skyrocket. The companies may suffer stock-outs and lose sales and customers' good faith. Analytics can help the companies to make more accurate forecast on how price discounts may affect the demand. That is the price elasticity. So, the company can select the right product and timing for price promotion to increase profit and customer satisfaction. As an example, let's look at XiaoDingDang T-Mall Store. XiaoDingDand is a manufacturer and seller of children shoes in China. It has about 50 to 100 SKUs and an annual revenue of about $4 million. Intensive competition forces the firm to use price discounts from time to time to attract more customers. These are the prices and unit sold for one SKU. A boy shoe for the fall and winter. The blue dots are the prices and red dots are the units sold, which is the demand. Clearly, the price affects demand as demand went up at most price discounts. However, the price elasticity varies over time. For instance, in box A, which is Summer, deep price discounts had almost no impact on demand, but in box B, which is Fall, and Winter, a small price discount. Say 10 percent, can increase demand significantly sometimes by five to 10 times. The pattern repeated over time. The current practice just has only a few price discounts during the peak season, because doing so, may significantly increase demand and they immediately ran out of inventory. During off-peak season, they built up so much inventory and thus were forced to do frequent and deep discounts, which had almost no impact on the demand, they just lost profit. The root cause is that they cannot predict the price elasticity. If they could predict accurately, how price discounts affect demand in different seasons, they could build up inventories properly and make a lot more profit. The responses of demand to price discounts are actually hidden in the historical data. Applying analytics to data prior to 2016 we build a model that can capture the demand peak quite accurately for this product upon a 10 percent discount in the winter of 2016. Also capture the almost zero impact of price discounts on demand during the summer of 2016, which is the off-season. The overall forecast error for this product is about 30 percent. For another popular product, a girl's shoe, the model captures almost all the major demand peaks. With a forecast error of about 44 percent. To summarize, supply chain analytics can contribute to selling operations significantly. An Accenture high-performance supply chain survey in 2009 interviewed 1,500 plus executives, from 10 industries and 21 countries, and showed that Planning Leaders can achieve 10 percent better forecast accuracy and lower inventory cost than their peers, while maintaining a 99 percent fill rate.
