As technology continues to evolve, consumer behavior does too — and retailers need to stay ahead of the curve. And using data is one way to make sure you stay ahead of trends and give customers products that solve their problems.
With the Internet creating more data than ever before, big data has become an industry buzzword. Big data typically refers to “data sets that are so voluminous and complex that traditional data processing application software are inadequate to deal with them,” and which used to be impossible for any but the largest, most technically savvy company to collect and analyze.
But simply gathering vast amounts of data alone isn’t particularly helpful. What is valuable is digging through all those numbers to find significant insights about trends, customer preferences, and even future predictions.
So, how are retailers actually utilizing the insights and data they gather? We’ve rounded up six examples of well-known brands that are using big data, artificial intelligence (AI), and machine learning to optimize their processes, anticipate their customer needs, and — in the case of one brand — even identify the early stages of pregnancy.
But before we get ahead of ourselves, let’s take a deeper look at the ins and outs of machine learning specifically and how retailers can leverage it in their business.
What is Machine Learning?
Machine learning is one of the key technologies that’s increasingly valuable for retailers as more and more businesses take advantage of big data.
Before going any further, it’s important to understand what is meant by the term machine learning — it is related to artificial intelligence (AI) but the two aren’t the same. AI, broadly speaking, refers to a computer’s ability to make decisions in a way that imitates human logic.
However, machine learning is the way in which a computer can “learn” these logical rules without simply being programmed to do things a certain way. In other words, machine learning allows a computer to continuously update its understanding of the rules as it sees more examples of how humans react to various external factors.
This type of technology has become more widespread as hardware improvements make it possible to handle the sheer volume of data and run complex algorithms. Basically, machine learning is much easier to use these days because of technology has evolved to make it easier for retailers and consumers to use.
The most well-known example of machine learning in action is the Google search engine (yeah, the website you use every day). Google uses each query (i.e. the phrases you enter into the search bar) that a person runs as a data point for teaching the algorithm about humans’ search behavior and intent. The more the Google search engine learns, the better it is at answering questions and offering relevant sites for your searches.
But machine learning isn’t just for multinational tech companies — it’s possible to use it in a retail context as well.
The Benefits of Machine Learning for Retail
Retailers can apply insights retailers from big data and machine learning in a number of ways, especially for optimizing the supply chain, product sourcing operations, and for marketing and customer acquisition.
For example, a recent McKinsey study discovered that “U.S. retailer supply chain operations who have adopted data and analytics have seen up to a 19% increase in operating margin over the last five years.”
The same study also noted that the biggest challenge U.S. retailers currently face is simply the lack of analytical talent and shared data across their company; the opportunity is there for those who can bridge this gap.
Some key ways you can use machine learning in a retail context include:
- Offering highly personalized product recommendations for advertising and promotions (for automated upselling and tailored, complementary product suggestions based on previous purchases).
- Optimizing your pricing strategy with real-time, dynamic prices. An algorithm can take key pricing variables into account, including supply, seasonality, and demand and offer you insights on how to adjust your prices accordingly.
- Optimizing inventory planning and predictive maintenance. Systems can detect “freshness” of perishables and wear and tear on machinery, and predict demand in advance for ordering stock.
- Optimizing routes for more efficient deliveries according to past data and behavior.
- Sales and customer service forecasting systems to predict customer behavior and allow retailers to deploy sales and customer service staff where they will be most effective.
- Website content customization: Personalize the online experience based on an individual’s location, purchasing history, demographic, and more.
- Segment your prospective customers based on previous behavior rather than self-identification.
The greatest value of machine learning is its predictive nature — it allows companies to use past and present customer and operations data to predict future behavior and trends. For example, let’s look at a customer who is normally a modest spender but has bought expensive planning materials around the same time each year for the past three years. Machine learning models could predict the most relevant time to offer these products again rather than wasting ad dollars at a time when the customer isn’t likely to make a purchase. Or, for high volume shopping periods like Black Friday Cyber Monday, machine learning can help retailers estimate how much inventory to stock compared to the rest of the year.
As a result, machine learning models help cut back on typical waste (such as unnecessary advertising costs and spoiled inventory) while optimizing marketing efforts to anticipate customer needs — leading to increased revenue and higher profit margins. For example, Target Corp. (one of the brands featured in this article) saw 15-30% revenue growth through their use of predictive models based on machine learning.
Examples of Machine Learning in Retail
Here are six examples of machine learning in a retail setting, illustrating the variety of use cases in which this technology can provide value.
Target: Predicting Pregnancy
As a “one-stop shop” for everything from clothes to groceries to household items, Target wanted to encourage shoppers to buy a wider variety of items from them rather than their competitors. Research has shown that the most typical time for a shopper to alter their store of choice is during a big life change: graduation, marriage, childbirth.
Target hired a machine learning expert and statistician, Andrew Pole, to analyze shopper data and create a model which could predict which shoppers were likely to be pregnant. After cross-referencing women’s common purchases who later registered with the Target baby registry (providing their due date in the process), Pole was able to identify key patterns.
These trends not only indicated pregnancy, but could pinpoint the current trimester of a woman’s gestation period (for instance, if a woman suddenly started buying certain supplements, she was likely in her first 20 weeks of pregnancy, whereas purchasing a lot of unscented lotion indicated the start of the second trimester).
This case study also illustrates the caution with which retailers must proceed in utilizing this type of insight. Target used this data to send coupons related to pregnancy and parenting to customers whose buying patterns fit the model. That included a 16-year-old girl whose father found out about her unintended pregnancy when she received these targeted promotions. Target later adapted their strategy to mix other offers in with the pregnancy focused promotions after finding that their customers felt uncomfortable with this degree of personalization.
Walmart: Anticipating Customer Needs
Retail giant Walmart has also implemented new technologies to anticipate customer needs and optimize operations. In 2015, the company tested facial recognition software as an anti-theft mechanism.
However, the discount giant also plans to use this machine learning technology to upgrade its customer service. According to Forbes, Walmart’s patent application for the machine learning tech that customer service can “be very expensive to maintain sufficient staff to provide great customer service. It can also be difficult to establish an appropriate staffing level that will provide proper customer service without excess staffing.”
The facial recognition software has the ability to recognize the level of frustration of customers at checkout and trigger an alert for a customer service representative to speak with the frustrated customer.
North Face: Robot Sales Associates
Outdoors clothing retailer North Face has been using artificial intelligence and machine learning to offer website users a highly personalized shopping experience called “Shop with IBM Watson.”
After downloading the app, shoppers speak right into their phone to access Watson, an AI system from IBM. Similar to a human salesperson who might help you select the right option, the virtual assistant walks users through a series of questions and learns from your answers to offer you the most relevant products for your preferences and needs.
Alibaba: Making Big Data Accessible for Smaller Retailers
Alibaba, a Chinese ecommerce platform similar to Amazon, is by some accounts the world’s biggest ecommerce marketplace. Unlike Amazon, however, which has traditionally been in the business of order fulfillment, Alibaba relies much more heavily on its retailers and considers itself a “retail ecosystem.” Because of this, they have prioritized big data analysis and one of their major features is to make that data more accessible to the smaller retailers who sell through their service.
Their latest application brings big data to the offline retail world so merchants can understand the bigger sales picture. For example, shoppers can order online for delivery from the Alibaba-backed grocery store HEMA. Or they can shop in-store, scan barcodes as digital price tags update in real-time, pay via their app, and get free delivery for their in-store purchases.
This allows Alibaba to capture this “offline” shopping behavior via the mobile app, which can be analyzed alongside the online data to offer a complete picture of customer behavior.
Amazon: Personalization and Predicting Supply and Demand
Amazon has one of the most famous recommendations engines of any ecommerce retailer, and for good reason; their machine learning algorithms work so well that 55% of sales are driven by these machine learning recommendations.
But the recommended products engine serves a dual purpose. It’s not only valuable in driving additional revenue through upsells and suggested products; the insights gained by these machine learning algorithms can also help Amazon to forecast predicted demand for inventory, making seasonal and trend-based supply decisions simpler.
Netflix: Giving Viewers The Entertainment They Want
Since its inception, Netflix has been using big data and machine learning to understand how its users consume television and film content and deliver the content the viewers want. This data has informed strategic decisions such as the way in which they release full seasons all at once, auto-play the next episode, and offer recommendations for how likely you are to enjoy a related film or show (their “% match” rating is the latest example of how they offer this type of data-based recommendation engine). This data has also informed all the original content they produce.
According to estimates from Netflix executives, machine learning insights save them $1 billion per year.
Moving Forward With Machine Learning in Retail
Want to learn more about how to use machine learning in your marketing campaigns? Check out this article on How to Use Machine Learning to Enhance Your Marketing Campaigns.
Already on the big data and machine learning bandwagon? Share your advice in the comments below.
Marchine learning FAQ
What is machine learning in simple words?
What are the 4 basics of machine learning?
- Data collection and preparation: Gathering data from various sources and preparing it for use in machine learning algorithms.
- Model selection: Choosing the right type of machine learning model for the task at hand.
- Training: Training the model using the prepared data.
- Evaluation: Evaluating the trained model’s performance and adjusting parameters to optimize results.