Bart Mroz, founder and CEO, Sumo Heavy

For retailers, the saying that “knowledge is power” has taken on new meaning in the age of big data and advanced analytics. With the enormous amounts of customer data available to retailers today, those retailers that can employ advanced analytics to derive actionable insights from their data will be able to increase their sales and gain competitive advantages.

Advanced analytics includes artificial intelligence (AI) and two subsets of AI—machine learning and deep learning. Many of the analytics applications that can help retailers gain advantages—such as customer segmentation, personalization, recommendation engines, and predictive analytics—are based on machine learning.

The beauty of machine learning is its ability to perform millions of operations per second on large volumes of data to find insights and patterns. Starting with a model and a set of data, machine learning can improve its accuracy, or “learn,” by comparing predictions with actual results repeatedly.

The potential of machine learning for retailers is so great that the National Retail Federation has declared that “machine learning will revolutionize retail.”

Recommendation Engines

One of the most effective ways retailers can use machine learning to boost revenues is through product recommendations. McKinsey estimates that Amazon generates 35% of its revenues from sending emails with product recommendations driven by machine learning.

Intelligent chatbots also can be employed by retailers to make product recommendations. Back-ended by machine learning, chatbots can engage consumers in conversations while accessing their personality profiles and shopping histories in the background. The intelligence gathered can be fed back into the system to inform subsequent recommendations.

Personalization at Scale

Personalization can yield powerful results, and personalization conducted at scale can be even more powerful. HubSpot found that personalized calls to action resulted in a 42% higher conversion rate than generic calls to action, while Kibo research showed that personalization can improve average order values by 40% and conversion rates by as much as 600%.

Machine learning systems can comb through millions of records to identify the most promising customer segments and best offers to present to customers to win their business. For example, if you are a women’s clothing company selling to all age groups, machine learning can identify which products sell the most to particular age groups. This data can yield an immediate boost in sales by enabling you to target particular customer segments with the most attractive product offers. Moreover, the entire process can be automated to be conducted at scale.

Machine learning systems can scour customer records to identify the customers who are the most frequent buyers and who spend the most money. These loyal and high-value customers can then be sent personalized and customized offers based on their personality profile, buying history, and product preferences. This process also can be performed automatically and at scale.

Harley-Davidson’s New York dealership used a machine learning system to personalize advertisements for different customer segments. The system found that the potential customer base in New York included groups the dealership would never have considered. By optimizing all advertising factors—including messaging, channels, and budgets—the system increased leads by almost 3,000% in three months.

Real-Time Targeting

Machine learning systems can monitor clickstreams to identify in-the-moment opportunities such as when a consumer shows interest in a product on a product page, on a social media site, or in a search. The offers that are made can be informed by each consumer’s personality profile, product preferences, shopping behavior, and social media activities (for example, products on Pinterest they have “liked”).

Real-time event processing systems can determine whether an action should be triggered when a particular interest is detected, such as when customers indicate that they are involved in a wedding or graduation, or when shopping for a house or new car.

Using geolocation and machine learning, real-time offers can be triggered by consumer visits to stores and local geographies. Personalized emails also can be triggered in response to real-time events such as website and shopping-cart abandonment, with sales losses minimized through personalization engines and retargeting techniques.

Walmart is at the forefront of real-time merchandising, using an enormous private cloud to track millions of transactions a day, including demand, inventory levels, and competitor activity. Its system responds automatically to market changes in real time, triggering immediate actions based on machine learning insights.

Predicting Purchases, Setting Prices

Machine learning systems can analyze market data, customer buying patterns, and social media data to predict trends, prices, and customers buying intentions. Based on a customer’s profile and past behavior, machine learning systems can predict what a customer is likely to purchase next, which in turn can enable retailers to present the most attractive offers to that customer.

Machine learning systems can predict pricing by analyzing demand, product pricing history, competitor activity, and inventory levels. The systems can automatically set optimal prices in response to market changes in real-time.

By processing transactional, behavioral, and intent data, machine learning can determine who are the best customers to target, when a customer is likely to churn, and the most effective message to send at key points throughout the customer lifecycle.

To minimize churn, machine learning can identify the best incentives to offer customers at the critical moment the customer is considering leaving, a process that has proven to be effective in retaining customers.

Culture of Analytics

Research shows that to be successful, retailers must foster a culture of analytics in which organization and governance come from the top down. As IBM points out, to derive the benefits of analytics “takes hard work, effort, and a deep understanding of the organization and especially its data.”

Despite widespread recognition of the value of data, few companies have implemented modern data strategies. A lack of organizational structure, in-house expertise, as well as technological barriers are preventing most retailers from capitalizing on analytics.

For example, data remains fragmented and siloed in most retailers’ operations, even though aggregating various data types into a centralized repository is critical for creating 360-degree customer profiles for personalization.

With so many retailers failing to employ analytics, there is a significant opportunity for those retailers that make a disciplined commitment to gain competitive advantages from their data. By molding the talent, assembling the architectures, mastering the techniques, and applying analytics to their retail operations, retailers can reap major financial rewards.

Sumo Heavy is a digital commerce consulting and strategy firm.