How Retailers Can Scare Up Profit this Halloween with Automated Machine Learning

Americans love Halloween and will spend around $9 billion for the holiday season -- $2.7 billion on candy alone, according to the National Retail Federation (NRF). This is a huge opportunity for consumer packaged goods (CPG) companies that produce candy and for many retailers to drive direct business and residual shopping traffic.

The Halloween season sounds great for retailers, but they could end up turning into pumpkins at midnight, leaving behind a trail of missed opportunities. If they hold too much candy in inventory, then it mummifies into day-after discounts, resulting in a loss. If they have too little candy in-stock, they can miss opportunities to sell more. If they price too high, customers will spend online or at other retailers. If they price too low, customers pocket money that would otherwise end up in retailers’ coffin -- err -- coffers.

What types of insights do retailers want to know? Today’s retailers that are successful look for insights to guide them in their strategy and execution, such as:

  • Demand forecasting for getting the right types of candy to the right stores at the right time
  • Pricing optimization by SKU, brand, or category
  • Assortment optimization across the candy shelf to determine the right mix of products, package sizes, and flavors

 

How is the retail market polarizing?

Many consumers are looking for low-priced candy, with 42% of Halloween shoppers going to a discount store, per NRF. This means that the discount channel will be a big player in the competitive landscape. While hypermarkets and superstores remain important, the entire channel composition has been changing, with discount, convenience, and online stores being the fastest-growing grocery channels. E-commerce is swiftly reshuffling also, with 25% of Halloween shoppers looking online. Also, social media now influences about 15% of Halloween purchases, with Pinterest having the most influence (18%) and Instagram on the rise (from 7% to 14% over the past four years). 

It’s hard to get predictive insights. Companies often take their insights from business analytics, pulling descriptive and historical sales. However, predictive analytics provides better understanding of the complexities of shopper behavior, channel dynamics, pricing elasticities, and other market conditions. Predictive analytics are forward-looking, and leverage machine learning and artificial intelligence to provide insights that, if acted upon, increase profits while lowering risks.

I have a headache! Finding and retaining data scientists with proven track records is very difficult and will continue to be for a long time, especially for companies that want to inject AI into their business decision processes. Within the next five years, the shortfall of data scientists in the U.S. alone is projected to reach 250,000.  And when you have data scientists, building predictive models is a very slow process, taking weeks or months to complete, especially retailers who need to move at the speed of the consumer, meaning responding in hours or days. Accuracy is yet another challenge. With hundreds of models that would need to be hand-coded, data scientists can only work so fast, often leaving behind missed modeling opportunities.

 

We need a hero!

If only there was a solution that could inject AI into the forecasting of Halloween candy for making clear and actionable recommendations. If only retailers could know how much to stock based on region, time, SKU, brand, sub-category, pack sizes, pricing, and discounting. Well, we’re in luck. Automated machine learning is the solution that can help with the pricing, assortment, and marketing of candy. And, for the supply chain, it adds in a time series component for predictive demand forecasting. As demand-based forecasting and predictive analytics become commonplace, the retail industry will rely heavily on artificial intelligence to set prices, manage inventory, and manage staffing.

 

Trick or Treat?

Will retailers learn the tricks of Enterprise AI and receive their treat of increased sales and profit with better forecasting? Interested in more retail analytics? DataRobot works with retailers and CPG companies globally such as Kroger, Walmart Canada, Carrefour. Visit our Retail solutions page for more content and insights.

 

New call-to-action

 

About the Author:

Ari Kaplan is Director of Industry Marketing at DataRobot. He works with DataRobot customers across retail, CPG, and sports. He has been working as a data scientist and leading teams of data scientists for decades in a wide variety of domains from Nielsen and IRI to Major League Baseball. He was president of the worldwide Oracle users group during the acquisition of MySQL, Java, and Peoplesoft. Ari graduated from the California Institute of Technology and received their Alumni of the Decade award.