These past few weeks, I have been speaking to audiences on the topic of how predictive analytics adoption in supply chain planning and execution processes will dramatically improve decision-making.  In these talks I like to further provide audiences current day examples of how predictive analytics capabilities can be applied to today’s supply chain planning and decision-making challenges.

Upon reading today’s edition of the Wall Street Journal, I came upon a report that I believe is a good current day example of how predictive analytics can be applied in the forecasting of supply chain resources for the upcoming holiday buying season, which is so crucial to the retail industry.

The article, Forecast Envisions a Weak Holiday Season, (paid subscription or free metered view) reported that ShopperTrak, which measures store traffic in 60,000 worldwide retail locations expects retail sales in the upcoming November and December period to rise by a mere 2.4 percent from a year earlier. Holiday retail sales grew 3 percent in 2012, and 4 percent in 2011. The current 2.4 percent forecast happens to be the lowest growth since 2009, and the byline of the report was that this will mirror concerns that retailers will have to step-up discounts and push earlier promotions for what is feared to be a dismal holiday sales period.  

However, the WSJ happens to note that early forecasts are often well off the mark, and that ShopperTrak’s forecasts have tended to undershoot actual Commerce Department retail sales data. In 2012, the firm forecasted a similar 2.5 percent increase in retail sales while the actual level turned out to be 3 percent.  ShopperTrak’s forecasts tend to be weighted towards the sales of general merchandise, apparel and accessory items.

WSJ further notes that Alix Partners expects November-December retail sales to increase between 4.1 – 4.9 percent from last year.  Alix obviously utilizes different forecast methodology and factors more detailed forecast categories that are related to apparel, gasoline, food, and e-commerce sales categories. To add more consideration, the National Retail Federation will release its own forecasts numbers in October. If S&OP teams were to respond to these current market forecasts, supply chain along with sales and marketing would be debating the earlier timing of sales promotion and discounting activities.

The most important insight in the article however, comes from a quote from a managing partner of Longboard Capital Advisors, an investor in retail stocks: “I don’t lend a lot of credence to these estimates because they weigh heavily on historic numbers and I think it’s more important to look at what trends are in-place going forward.

This is exactly what differentiates predictive analytics which is not to be completely grounded in forecasting based on transactional history or broad based categories, but to be more grounded in specific market trends that portend what product categories consumers are showing buying interest in, either in web-based research, most recent retail sales, or store visits. For example, these past few months, consumers have opted towards purchases related to replacing automobiles and investing in home improvement. Back-to-school retail sales in clothing and accessory goods were reported to be lackluster. Conceivably, certain products may trigger high consumer buying interest, while other products will have more subdued buying interest.  Knowing which items are the key to meeting the revenue and profitability plan for the most crucial quarter.

An analytics-based tool can predict demand by analyzing multiple historic and forward-oriented data points at the item level.  It would further weight external data related to consumer discretionary budget indices, consumer optimism and indicators of high or low consumer interest.

Advances in information technology, namely in-memory computing, support for analytics and data visualization provides industry supply chain and S&OP teams more affordable and user friendly predictive capabilities.  Rather than rely on a summary level or individual industry sales forecasts based on historic data, teams will have the ability to rapidly perform far more detailed analysis of expected product demand in individual item categories, and confidence weighting.

Rather than reliance on any singular historic based forecast, predictive analytics has the ability to factor multiple indicators of buyer activity that are forward-looking and grounded in current day market dynamics.

Bob Ferrari