Personalization based on data has become critically important in digital marketing, but you wouldn’t know it based on how marketers run their personalization initiatives today. Many still use outdated methods, with business rules created through offline analysis of collected data.
These rules are implemented in an automated way via a business rules engine, but over time the engine becomes bloated with hundreds of different rules. This murky collection includes old efforts that no longer work and highly complex rules that are hard to understand, and it’s incredibly inefficient when you’re trying to drive customer acquisition, conversion and retention effectively.
The approach used by the most successful digital companies (Google, Netflix, Instagram, Twitter and so on) is machine learning. Self-learning sense-and-respond systems do the hard work of analyzing data and writing business rules in real time. Once these rules are applied, the data analytics system can start to collect its own data instantly, evaluate the efficacy of the rules based on success or failure, and adjust accordingly. This way, the system doesn’t just build a collection of rules that keeps expanding like an out-of-control menu. Instead, the rules evolve and improve over time, continually advancing business outcomes by delivering insights for marketers so they can make decisions in moments.
Let’s take a look at several use cases that put machine learning for personalization into context.
Using historical consumer purchasing data, machine learning algorithms can predict when certain customers will likely need to purchase certain products. For example, consider a customer who is identified as physically active and who purchases new sneakers at approximately the same time every year. This rule is identified and validated with other similar customers, and companies can send all these individuals related content or offers for new sneaker models.
Product recommendations based on customer interactions
Say there’s a customer who performs multiple interactions across different channels—online search, mobile app, location data indicating store visits and so on. Based on this interaction history, a machine learning algorithm can analyze the customer’s patterns and map them to the behavior of similar customers. It can then predict the next action this customer will take and provide the right product to that person on the right channel.
Discount offers on abandoned searches
Lots of marketers create offers around abandoned searches. For instance, say a customer searches for a hotel in London—but rather than make her reservation online, she calls the hotel to book a stay. Then, although she has already made her choice, she gets unwanted content and offers for London hotels for the next several weeks. In this case, machine learning can identify her true context and provide more customized and relevant content and offers.
Customer purchasing path
The traditional way to determine that consumers who purchase Product A are also likely to purchase Product B is via product correlations stored in e-commerce systems. But with machine learning, this same information can be derived from customer search histories and purchasing behavior. Product correlations can be derived by algorithms reacting to data in real time, then applied to entire segments of customers.
Some of the most effective business rules are based on not only internal data, but also on a combination of third-party data and customer behavior data. Correlations between these types of data can reveal patterns that target content, products or offers to customers based on their specific or current context. For example, during a particularly long string of winter snowstorms on the East Coast of the United States, home improvement companies can launch a campaign for snowblowers targeted at homeowners.
Benefits and stages of implementation
Machine learning–based personalization offers clear benefits:
- Increased revenue through more effective personalization leading to customer acquisition, conversion and retention
- Better service and higher customer satisfaction as a result of personalized engagements
- Lower costs due to automated maintenance of machine learning systems
The road to personalization based on machine learning starts with collecting and integrating data from multiple channels—online, offline and third party—and preparing it for analysis. Step two includes implementing machine learning models for analyzing data, creating the algorithm or model that represents the data, and training the model to derive the best next actions. Step three consists of real-time integration of these models, then delivery of contextual offers and content to customers at the right times on the right channels.
The bottom line is that businesses need a data-science system that offers timely insights to help them make better and more informed decisions. That system exists—at Mindtree, we call it Decision Moments.
Contact us to schedule a detailed conversation about understanding your company’s data analytics initiatives to drive personalization.
Need help moving your digital initiative forward? Discover the four steps to personalization in our latest e-book by completing the form below.