How to Use Machine Learning to Further Retail Analytic Capacity

Print Friendly, PDF & Email

kerry-liuIn this special guest feature, Kerry Liu, CEO and founder of Rubikloud, discusses the importance of machine learning in maximizing the use of customer behavior data in the retail sector. Kerry leads three important functions at Rubikloud: people, sales, and technology disruption. Rubikloud helps retailers monetize their data to power personalized campaigns through the use of advanced machine learning techniques, effectively bridging the gap between marketing automation and retail management consultants. In his role, Kerry works to manage and maintain a thriving company culture that recruits the best and brightest in the industry, while also maintaining  relationships with global retailers and investors. He is passionate about machine learning and big data, and enjoys providing enterprise retailers with the tools and knowledge needed to enhance their overall business goals.

If you’re only using big data to target consumers based on demographics or purchase history, you’re making a mistake that will cost your company a significant amount of revenue.

Big data can provide retailers with a wealth of highly personalized information about their customers, informing meaningful customer interactions. But most retailers are only using the data to make very simple assumptions about their customers. As a result, shoppers are given cookie-cutter personas that do not evolve along with their shopping behaviors.

Targeting a consumer based on her age or purchase history certainly makes sense in theory. However, consumers are much more complicated, and make purchase decisions based on a complex set of factors that are constantly changing and evolving. Retailers are missing out on a significant opportunity to drive revenue and loyalty by failing to truly understand their customers with more advanced data analytics.

So what’s the answer? Advanced machine learning platforms. These tools have the ability to grow and change a retailer’s understanding of customers over time based on evolving data.

What you need to know about customers

To better know their customers, retailers must understand their buying behavior both online and off. And customers are much more sophisticated than their basic purchase history information.

Does the customer always stay within a certain price range? Will she spend more on a product if given discounts and promotions? Does this customer respond to small promotions or will she only react to a significant discount? How often does she purchase the product? Is she brand loyal? If not, why? These behaviors matter more than tracking the items customers have purchased in the past because they help retailers plan for the future and fine tune the content that customer receives.

Additionally, retailers must pay close attention to how every touchpoint influences the customer. Some buyers will only click through email promotions offering a discount, while others might be interested in promotions for new items or retargeting efforts based on search history online or previous in-store purchases. Each consumer responds to promotions and targeted marketing efforts differently, and understanding this requires more than just analyzing open and click-through rates.

Why traditional methods don’t work

Traditionally, retailers use a standard way to segment consumers based on recency, frequency of purchase and monetary value of purchases (RFM score). Retailers give each customer a score based on these three factors and forever market to this customer based on their initial score.

This traditional method, however, is far too simple. Consumers often make purchases that do not align with their predefined buying persona and, as previously mentioned, are prone to change their buying habits over time.

For example, a consumer who would regularly purchase discounted or inexpensive cosmetics might make a high end purchase at a new retailer for a gift. Using traditional methods, this retailer would continue to promote expensive, high end products to this consumer, even if the consumer later made their normal discounted purchases. This strategy would be highly ineffective given this consumer’s typical purchasing habits.

Or, this consumer might have purchased a high end product only because she was given a significant discount via email, but still regularly purchases  less expensive items in store. It would be an inefficient use of resources to continue to target this consumer with marketing for high end products as it’s unlikely she’ll ever purchase them at full price.

Consumers also evolve over time. If a consumer is making infrequent, inexpensive purchases in her early 20s her buying behavior might change has she advances in her career and becomes more financially stable. Similarly, a consumer buying baby items now (diapers, bottles, etc.) will not need to purchase these things forever. But, traditional methods will likely always target her as a new mother.

Relying on basic data does not take into account situations like these. Retailers need to use the big data at their disposal to analyze purchase behavior and target consumers effectively.

Why machine learning and big data are the answer

Consumers regularly make purchases that do not align with their typical buying behavior, making it dangerous for retailers to rely on strict personas when targeting a shopper. It’s important for retailers to have a dynamic view of the customer to more accurately predict the types of promotions and marketing efforts to which a consumer will respond.

Machine learning platforms have the ability to gain a deeper understanding of a customer over time without being explicitly programmed. What’s more, many of these platforms can analyze customer data in conjunction with larger market trends to better predict purchasing behavior and optimize pricing. Ultimately machine learning’s automation makes it the most effective way to offer highly accurate individualized promotions.

In the cosmetics example above, an advanced machine learning platform would consider the shopper’s entire purchase history and recognize that the high end purchase may have been an anomaly. To motivate her to make more purchases, the machine learning platform should then recommend that this retailer promote lower end, inexpensive products and offer unique promotions through this customer’s preferred channel to push her into a higher price range.

Machine learning would also recognize that purchases for a new baby will need to evolve as the child grows. Or, the program can recognize that if a consumer recently purchased a bottle of baby shampoo, it’s not efficient to continue targeting her with promotions for shampoo as she’s not likely to need another bottle immediately. Instead, machine learning can recommend products that compliment her recent purchases.

With all the data at their disposal, simply targeting consumers based on basic data such as purchase history or demographics is not enough in today’s complex world of omnichannel commerce. Machine learning is the best way for retailers to predict customer buying habits and adapt marketing strategies as the customer evolves over time. Consumers have high expectations for personalization, and will remain loyal to the retailers that truly understand their needs.


Sign up for the free insideBIGDATA newsletter.


Speak Your Mind



  1. Can we cross purchase behaviour with instore and online behaviour client by client? Have you use online and offline data using Seeketing technology in your ML models?

  2. Yes we have found a way to use online data to drive offline attribution which is valuable for everything from promotions to even how you spend your advertising budget.