Deeper Data: How Reinforcement Learning Boosts Marketing Performance
In this article, Piero Pavone, Co-founder and Chief Operating Officer at MainAd, explains how the adoption of advanced machine learning technologies like reinforcement learning (RL) can drive performance marketing by creating impressive results for marketers, driving sales and conversion rates. He shows how it can also help them strengthen brand reputation by creating a better user experience for consumers
Artificial intelligence (AI) is now smart enough to analyze other machines; with computer scientists recently creating an AI that can probe the ‘minds’ of computers and predict their actions.
For marketers though, as they harness the power of AI – and its applications through machine learning (ML) – predicting the actions of the consumer is still the main goal. And, thanks to advancements in ML, especially reinforcement learning (RL), it is now possible to refine and drive marketing performance by leveraging data about past events and current context to plot the best course for campaigns.
What is reinforcement learning (RL)?
Fundamentally, machine learning is the process of getting computers to learn like humans do. The ideal concept being that machines use algorithms to autonomously assess data in the form of observations and real-world interactions and learn from it with increasing precision over time; making predictions or solving problems.
Reinforcement learning takes ML to the next level. Just as the human brain makes choices based on the good or bad effects of previous decisions, so does RL – but with greater speed, accuracy, and scale. Plus, unlike traditional ML, it is not restricted by set rules for a finite range of scenarios. RL algorithms work on a case-by-case basis, evaluating data about previous activity and the rewards offered by different options, before determining what should be done to achieve the best long-term result. This creates a more fluid process.
How does reinforcement learning help marketers?
According to research by Business Insider, AI is the fastest growing marketing technology, expected to increase by 53% over the next year. The study also stated one of AI’s main advantages is that it can turn marketers from “reactive to proactive planners”, allowing them to plan campaigns more efficiently; particularly thinking about segmentation, tracking, and keyword tagging.
In other words, marketers can transform complex data generated by varied digital transactions into granular, real-time, usable insight. This detailed information provides the ideal basis for building impacting data-driven campaigns that drive strong results.
Reinforcement learning adds the personal touch
To compete with the array of digital content and buying options available to today’s consumers, marketers must ensure messaging resonates on a personal level and forms part of an engaging, streamlined online journey.
Research predicts that by 2020, customer experience will overtake price and product as the differentiating key factor for brands. Along with this shift will come three core challenges: the rising number of platforms and devices customers use, growing expectations around consistent and personalized engagement, and accelerating demand for new products and experiences.
To meet these challenges, companies must become more data-savvy, with a further study finding that data-driven organizations are “23 times more likely to acquire customers, six times as likely to retain customers, and 19 times as likely to be profitable as a result”. And, one of the ways they can do so is by utilizing RL. Armed with advanced behavioral insights and predictive ability, marketers can achieve hyper-personalization; where messaging is not only tailored to match a consumer’s current position in their unique journey, but also what is most likely to interest them right now. For example, if a consumer was browsing online with the intent to purchase a new cell phone, RL would allow marketers to analyze their previous behavior to predict when would be the best time to serve a discount offer and get the most positive response from that consumer.
Avoiding ad overload using reinforcement learning for both consumers and brands
The growth of automated, programmatic advertising can sometimes mean there is little control over the frequency of ads, and so it is not surprising that as a result of seeing the same ads too often consumers have resorted to installing ad blocking software.
RL algorithms can assess reactions to messaging and determine the ideal frequency for consumers. They can also inform real-time bidding activity in the programmatic marketplace, using predictions about consumer behavior to calculate which display ads to buy. This means advertisers benefit from a more efficient process, and hopefully as a result, a growth in online conversions – turning browsing into sales.
A practical application of reinforcement learning
To harness the speed and scale of RL, MainAd launched Logico, which combines smart ML technology with programmatic real-time bidding. Using behavioral data like social media likes, browsing history, and previous purchases, the platform can predict which channels and messaging will elicit the most positive response among defined audiences. Plus, by continually capturing data and tracking patterns, its performance improves with experience: greater insight into what has or hasn’t worked before better informs the targeting decisions of today.
Recently migrated to Google’s Cloud Platform, Logico has improved bidding accuracy by 75%, and optimized its processing speed to maximize ad inventory and run complex algorithms in less than 100 milliseconds. The technology can serve up to 50,000 requests per second and allows MainAd to build custom solutions for specific brands, offering increased performance that is more cost efficient.
One company which has benefited from the RL algorithms of the Logico technology is Asia’s leading baby and children’s wear company FirstCry, which saw its sales figures increase tenfold from January to December 2017. Following its adoption of Logico, it also saw new customer rates increase by 10% during the highest performing months.
As the example above illustrates, the adoption of RL is already showing impressive results for digital marketing campaigns. This can only be a good thing for brands wanting to move with consumers through the increasingly complex digital journey and offer more positive experiences. By taking advantage of machines that think like humans, marketers can use theory of mind to deliver campaigns that not only boost brand reputation and sales, but also consistently give consumers what they want.