Over the last decade, many facets of our lives have been impacted by artificial intelligence and machine learning. Virtually anything that has to do with data processing is a place for machine learning to shine. Deep learning now allows drivers to go hands-free on highways, helps biology researchers to create new molecules, predicts stock market trends, and provides farmers with soil restoration advice.
From banking to real estate to healthcare, organizations in various sectors are already reaping the benefits of this technology. Considering the evident power of machine learning to transform industries, one may wonder to what extent this technology can impact digital marketing. In this article, we examine how the world's leading companies use machine learning to enable marketing automation, enhance marketing campaigns, and increase revenue, based on Itransition’s extensive experience in machine learning consulting.
What is machine learning in marketing?
In a nutshell, machine learning enables marketers to significantly improve their decision-making by analyzing huge data sets and generating granular insights about the industry, market, societal trends, and customer profiles.
With the ever-growing processing power of computing systems and the increasing sophistication of machine learning algorithms for marketing, organizations can deliver hyper-personalized offers, content, products, and services.
Benefits of machine learning in marketing
Here are some of the benefits of machine learning for marketing:
With ML-enabled marketing automation, companies can drive down costs and free marketers’ time for more value-added tasks.
Machine learning allows companies to have a much better understanding of consumer profiles, create personalized offerings at scale, and increase customer lifetime.
With advancements in natural language processing and generation, machine learning enables companies to deliver content that resonates with customers and increases their engagement.
With the help of machine learning in digital marketing, organizations can automate customer segmentation and discover new, more potent customer groups.
1. Marketing automation
Marketers need accurate information to make data-driven decisions. However, with inordinate amounts of data available to marketers today, it has become increasingly daunting to process and analyze all of it manually. This is where machine learning can make a difference.
For example, customer segmentation, one of the most essential marketing practices, involves grouping customers based on various characteristics including age, sex, income, etc. Machine learning can not only automatically group these customers at lightning-fast speeds but also discover new customer segments based on a combination of characteristics that are not apparent to humans. Salesforce Einstein AI, for example, can analyze massive amounts of both customer and industry data and automate a range of marketing activities including customer segmentation and reporting.
Automation enabled by machine learning can also help marketing departments to make their back-office operations more efficient. For example, Zoom, a global leader in video conferencing, has partnered with conversational AI platform Ada to streamline the handling of inquiries.
Example: Ada and Zoom
One of Zoom's marketing departments, which is responsible for handling inbound sales interactions, was processing too many interactions that weren’t related to sales. Employees had to manually transfer these calls to the right department, which led to longer wait times and poor customer service.
So, the company wanted to automatically route customer inquiries to the correct department and save employees’ time without compromising the quality of customer service. By integrating Ada’s conversational machine learning model as the first point of contact across all Zoom’s interaction channels, the company managed to reduce live chat handle time by 33%. Now, when a customer starts a live chat with Zoom, Ada’s technology automatically identifies if it’s an existing user with a support inquiry or a potential lead with a sales-related question, and transfers the call to the appropriate department.
To enable a more granular understanding of customer profiles, Zoom's IT team also embedded Ada's code on the Zoom website to allow Ada to take into account user subscription models and past interactions. As a result, Ada and Zoom achieved a 25% reduction in time spent qualifying leads and automated 70% of sales interactions.
2. Marketing analytics
If there’s ever a need in the marketing realm for a superhero, it’s for reading human emotions. However, machine learning for emotion recognition is already extensively used in many industries, with companies like BMW using it to assess driver alertness and Disney for measuring viewers’ reactions to its movies. Marketers can use machine learning and emotion recognition to assess how consumers respond to advertisements and correlate these emotions to purchasing intentions.
Affectiva, a company that specializes in building machine learning models that understand human emotions and cognitive states, provides its Media Analytics solution to 28% of the Fortune Global 500 companies including Mars, Kellogg's, and CBS.
If it wasn’t for justifiably strict regulations around the storage and analysis of biometric information, organizations would be all over such machine learning retail technology as emotion recognition. However, given very strict regulations, companies can use Affectiva’s tech only on volunteers. As viewers watch advertisements, Media Analytics captures their facial emotions, analyzes them, and presents findings in an easy-to-use dashboard.
What's even more fascinating is that engineers at Affectiva managed to correlate facial emotions with crucial marketing indicators like a person's purchase intent or brand recall. These findings can be used to fine-tune advertising in numerous ways. For example, a lack of engagement identified at the end of a video ad can prompt advertisers to change the ad's climax. Detecting the most emotionally engaging moments can help marketers to choose the best parts for an ad's shorter versions.
3. Ad personalization
With an avalanche of ads that consumers view daily, personalization becomes paramount for achieving success. Today’s marketers are aware of that and often utilize automated systems for keyword generation and other related tasks. The problem is that these tools are rule-based and don’t really ‘understand’ a particular customer’s context.
Example: CommonWealth and Appier
This was exactly the problem of CommonWealth Group, one of the biggest media outlets in Taiwan. The company publishes four magazines dedicated to financial management, lifestyle, career, and parenting and gains profit from reader subscriptions and advertisements placed in its magazines. CommonWealth’s marketing team had very basic data on its audience and struggled to personalize advertisements, so the publisher showed the same ad to all readers.
CommonWealth turned to Appier, a company that provides an AI-powered platform to brands to help increase customer engagement. Appier’s custom machine learning model was able to identify reader profiles in detail by analyzing how customers interact with the mobile app and website, and interpreting data from a CRM system. This allows CommonWealth to identify reader profiles in real-time and dynamically deliver personalized ads.
To show the power of machine learning in marketing, let’s turn to a case study provided by Appier itself. When CommonWealth marketers were tasked with setting ads for an article named ‘Never Think Your Current Job is Menial if You Believe You are Good Enough’, their choice of keywords was rather predictable and understandable – ‘job seeking, human resources, and self-improvement’. Appier’s model, however, had a very different opinion, selecting 'interviews, advertising, and rock climbing as the main keywords. After switching keywords, the ad’s CTR increased sixfold.
4. Content creation
Traditionally, copywriting and marketing content creation have always been associated with creativity and emotional awareness that cold-hearted algorithms certainly lack. As it turns out, consumer sentiment about marketing messages can be interpreted as data in the digital realm. Insights generated from that data can be used to generate the exact language that encourages a particular consumer to take action.
This is exactly how many industry giants are using the Persado Motivation AI Platform to refine the language they use across marketing messages. A combination of proprietary machine learning and deep learning models allows Persado’s system to learn response patterns for a particular consumer, deliver hyper-personalized messages, and improve content marketing overall.
Persado’s natural language processing model assesses a brand’s content to determine its tone and voice. Then, it analyzes how a certain customer responded to various marketing messages and creates an emotional profile. Afterward, the platform breaks down marketing creatives, tests thousands of potential message combinations using a machine learning model, and creates a hyper-personalized copy that will resonate the most with a particular customer.
Example: Vanguard and Persado
Vanguard is one of the largest investment organizations in the world. One of its divisions struggled to personalize messages to customers in search of a retirement plan sponsor. Given that this sector is highly regulated, the company has a very limited number of channels to reach potential customers. In fact, LinkedIn is the only social media platform that the company uses to reach prospects. To stand out from the competition, the company turned to Persado to make marketing messages more personalized at scale.
Persado’s platform was used to craft exact phrases, choose the right formatting, and deliver a meaningful CTA to individual customers. As a result, Vanguard has managed to increase the conversion rates of its division by 15%.
5. Weather-triggered marketing
As much as we try to influence every part of our lives, for better or worse, humans haven’t been able to learn how to control weather conditions. However, with machine learning and predictive analytics software, we can not only forecast weather but also predict how changes in weather conditions impact consumer behavior.
For example, on a surface level, rainy days cause car washes to be empty, early snow ensures a fruitful season for ski resorts, and hot summer days make ice cream sales skyrocket. While businesses have been long aware of these correlations, machine learning allows for a much more reactive approach to digital marketing and advertising.
For example, IBM Watson Advertising Weather Targeting allows companies to tap into local weather’s influence on consumer preferences and actions. IBM's machine learning model automatically adjusts advertising copies and creative elements based on weather conditions in real-time.
Example: Walgreens, Clinch and IBM
Walgreens, the second-largest pharmacy store chain in the US, joined forces with AI-powered ad personalization platform Clinch to drive traffic during allergy season. Given that the weather and pollen counts are one of the main triggers of allergies, Walgreens wanted to advertise its products just at the time when customers needed them the most. Based on users’ locations and weather data from IBM, consumers in a specific area were targeted with a dynamic ad and provided a coupon for allergy medicine.
Based on local time, weather, user data, and other parameters, Clinch’s machine learning-powered platform personalized 160 creatives by location. As a result, Walgreens saw a 276% increase in CTR and a 64% decrease in cost per click.
6. Contextual advertising
In 1994, AT&T’s first online ad displayed on HotWired had a 44% click-through rate (CTR). Today, 0.3% is an average CTR for an online banner. With an avalanche of online ads displayed to a modern internet user daily, it has become increasingly hard for marketers to create ads that truly resonate with potential clients. With strengthening regulations around cookies, content marketing has become even harder, which also causes companies to rely more on contextual advertising.
In the marketing context, contextual advertising means placing ads that target specific audiences on relevant websites. For example, a news website dedicated to consumer electronics would potentially be a great place to display ads for a new phone. However, figuring out the best web pages to show ads and formulating the right message for a targeted audience is a monumental task if done manually.
Example: MINI and GumGum
This is why MINI, an automotive company owned by BMW, turned to GumGum, a company that focuses on using artificial intelligence and machine learning for analyzing digital content. MINI wanted to ensure that their new hybrid SUV Countryman would be perceived as a serious competitor in the smaller SUV market. MINI had a clearly defined target audience – young families passionate about the outdoors and eco-conscious adventurers.
With the help of proprietary image recognition and text-mining technology, GumGum’s system can identify what pages a defined target audience will naturally gravitate towards. In the case of the MINI Countryman, it was web pages about outdoor lifestyles, sustainability, eco-friendliness, parenting, and SUVs of course. On top of that, the system targeted prospects with positive reviews of MINI Countryman and less positive reviews of other SUVs that compete with MINI. What’s more, GumGum also partnered with Lumen Research to utilize their machine learning-enabled eye-tracking technology to measure consumer engagement during ad display.
Augment your marketing with machine learning
7. Identification of marketing opportunities
While we often attribute machine learning's value to improving the speed and accuracy of the data-driven decision-making process, its truly disruptive potential lies in its capability to make sense of loads of unstructured data. As much as it's instrumental to rely on well-known metrics to craft marketing campaigns, great marketers also try to dig deeper and understand what societal trends and cultural nuances drive consumer demand. Books, movies, music, and a myriad of other media sources influence our thinking, and, consequently, our buying behaviors.
Stan Sthanunathan, head of insights at Unilever, understood that and pioneered a machine learning-powered solution for identifying insights about societal trends at his company. By analyzing unstructured data and interpreting metaphors from movies and songs, Unilever gained invaluable insights into consumers’ hidden desires.
For example, the model revealed that there are at least 50 songs in the public domain with lyrics about ice cream for breakfast. Eventually, after a whole lot of data crunching, Unilever discovered that there is an unsatisfied need for ice cream for breakfast, and launched a respective product under its Ben & Jerry’s brand. While official sales reports of Ben & Jerry's new product haven't been disclosed, the majority of the brand's competitors have launched similar products just after 2 years of Ben & Jerry’s new product launch.
8. Recommendation system
Nowadays, recommendation systems are the backbones of successful ecommerce stores. Product recommendation engines help customers to navigate often enormous online catalogs and find items they need. Machine learning-powered recommendation systems play an invaluable role in increasing customer satisfaction and engagement, and improving a company’s bottom line.
Example: LUISAVIAROMA and Dynamic Yield
LUISAVIAROMA (LVR) is a high fashion online retailer that sells more than 600 brands across more than 150 countries. LVR wanted to optimize its recommendation strategies to increase customer retention and boost revenues. The company turned to Dynamic Yield, a company that provides ML-powered personalization solutions for online retailers.
First, the Dynamic Yield team built personalized add-to-cart recommendations to cross-sell and upsell at the point of sale, which led to a 15% boost in revenue per user. Second, the company optimized recommendation strategies on the thank you page, which led to a 14% increase in average revenue per user. Third, Dynamic Field deployed a recommendation system that notified customers about popular items that are about to go out of stock, leading to a 6% increase in revenue per user.
9. Marketing budget optimization
Aside from customer profiling and coming up with the next creatives, all marketers have to learn the art of budget optimization. Spending too little results in insufficient revenues, while spending too much hurts profitability. Especially when it comes to big enterprises with thousands of marketing campaigns going on at the same time, there is usually a dedicated team of people who decide how to allocate budgets to get the most returns on investment. Unsurprisingly, this often takes inordinate amounts of time and results can be subpar.
Given that there is no shortage of data on campaign performance and customer behavior, machine learning can be used to automate a large part of campaign bidding and increase its performance.
DoorDash, one of the leading food delivery companies in the US, developed a custom machine learning model to automate budget spending. DoorDash spends millions of dollars on thousands of marketing campaigns to attract new users and beat the competition. Realizing that they spend unjust amounts of man-hours on allocating budgets, DoorDash turned to machine learning to optimize this process.
Interestingly enough, given that the company delivers a number of campaigns across a range of marketing channels, DoorDash engineers opted for a dedicated machine learning model for each channel. Currently, based on campaigns’ historical data, machine learning models automatically adjust budgets, significantly lowering marketing costs.
10. Lead conversion
While every marketer’s goal is to generate as many leads as possible, identifying the most potent leads and converting them is an increasingly complex task, especially when there are more leads than a company’s employees can physically process. With the latest advancements in natural language processing, it is becoming possible to automatically and intelligently assess leads’ chances for conversion.
Example: Hootsuite and Conversica
Hootsuite, a Vancouver-based company that offers a full suite of social media management and analytics tools, has become so popular that it struggled to follow up on all the leads. To prevent missing opportunities and realize the full potential of its offerings, Hootsuite turned to Conversica, a company that provides conversational AI solutions.
As a result, Hootsuite significantly boosted the productivity of the sales team by offloading hundreds of leads per day to Coversica’s AI assistant. Currently, the company employs three AI assistants that cater to the nuances of different regions including North America, Latin America, and Europe.
AI assistants can provide prospects with all the relevant product information and answer questions in a human-like manner. Based on Conversica’s sophisticated machine learning algorithms, AI assistants can identify when the lead is ready to talk to the sales team. All content leads are now going through Conversica’s AI assistants, which boosted engagement rates from 0.5% to 4%.
Marketing has always been about establishing meaningful connections with customers.
In a sense, our conversations with colleagues, friends, relatives, and loved ones become meaningful when both parties are aware of each other's past experiences, can intelligently assess each other's emotions, and be empathic. While banking on technology to become more human may sound counterintuitive, it turns out that machine learning can be exceptionally good at deciphering human emotions, needs, wants, and intentions by analyzing mountains of data that companies have at hand.
Whether it’s marketing automation or analytics, machine learning is ripe for marketers to use, and it’s high time companies augmented their marketing departments with it.