Marketing automation’s best opportunities come from data, and marketers will soon realise the transformative potential of machine learning technology
Google Trends shows that over the past couple of years, search interest related to machine learning (ML) has been consistently increasing. In the present day, ML is being applied to finding solutions for many real-world problems — from medical diagnosis, speech recognition to fraud prevention. Earlier this year, we also witnessed how AlphaGo (a computer program developed by Google DeepMind) played one of the ancient games of the world called ‘Go’ and defeated the world champion.
Machine learning: the past and the present
Essentially, ML combines data and computer programming to develop intelligent algorithms that can improve the predictive capability of computers as they continue to consume more and more data. That said, ML techniques are not brand new — they date back to 1950s; the most relatable examples would be spelling and grammar checking tools. These tools have been built on the fundamental principles of ML, i.e., using data sets to recognise errors.
Traditionally, the two most important hurdles in the application of ML have been lack of adequate data and significantly lower number of software libraries for software engineers. But, in the last few years the adoption of ML techniques has gained momentum with the advent of big data and proliferation of commercial grade machine learning frameworks.
Notable examples include the widely used scikit-learn Python library and well-publicised libraries like Tensorflow by Google and Azure ML Studio by Microsoft. In addition to all this, companies are recognising the value of analytics and placing it at the centre of their decision-making process.
One of the applications of ML that has directly touched almost all netizens is Google’s RankBrain system — the third-most important factor for ranking webpages. When the search engine encounters a search query that it had never seen, RankBrain interprets the keywords, matches them with phrases that might have similar meaning, and presents the best possible result page to the user. Over time, this system will learn language semantics by applying advanced mathematical processes (word vectors) and upgrade itself by recording the real meaning behind the search queries; without relying on human intervention for pre-programming.
Here is an example.
I entered “what is the name of the robot character in the movie on artificial intelligence and turing test” as a search query, which can be quite complex for a machine.
Google did a remarkable job of recognising my query and displayed the movie name. The exact name of the robot character (Ava) can be located in the fourth link. Try out the same search on Bing to spot the difference!
Of course with better accessibility to ML, larger companies are already applying it to sales and marketing.
Exponential growth in data, economies of data storage and advancement of ML tools, have set the stage for innovative application of ML in marketing automation as well.
Marketing automation
Leave aside the marketers, it’s human nature to seek convenience. When marketers had to perform repetitive tasks, and trigger certain activities based on specific events, marketing automation came into play. Now, automation applies to various facets of marketing: social media management, rule-based drip campaigns, analytics, data management, and more. Websites, apps, and all other media offer billions of opportunities to marketers to communicate with consumers every single day.
Every touch point generates a wealth of data and provides the possibility analyses in real-time to deliver value to consumers at the best possible time via the most suitable channel. As data is the most important element of an effective ML strategy, it fits perfectly with marketing automation.
Implications of machine learning in marketing automation
Marketing Automation’s best opportunities come from data, and marketers will soon realise the transformative potential of ML technology. Now, let’s discuss some of the most important applications of machine learning on marketing automation.
Automated personalisation. Marketers come up with great ideas for marketing campaigns, but it always involves set of assumptions along with testing, measurement and refinement. All of these activities consume considerable amount of time and marketers won’t be able to really scale in terms of personalisation if they have to continuously observe the test data before making decisions.
Machine learning can help in automatic personalisation by taking over the decisions related to content, holistic experience and optimisation. For example, a self-learning algorithm can personalise the experience in real-time for millions of website visitors as it’ll be able to update itself by continuously going through the past data.
Apart from providing highly personalised offers, content and customising call-to-actions as per the predictive profile of the customer, it could also help us dynamically deliver the most suitable experience to each user by controlling the finer elements of the site or app, i.e., personalising the layout, navigation and showing or hiding different components.
Another example would be high volume drip campaign tests with respect to email subject line, copy, content and CTA. In this case, multi-armed bandit algorithms can be deployed to go through all the variations and continuously move to the variation that has the highest conversion rate. This is what we call ‘continuous optimisation at scale’.
Customer Segmentation. One of the biggest challenges for any marketer is accurate customer segmentation. The idea is to create clusters of customers with similar characteristics and trigger certain actions depending on these. When marketers spot at-risk customers who are about to churn out, they react to it and execute the churn management campaign. For example, a SaaS company would consider number of sessions, time spent on the app, last login time, features used to predict churn and guide the customer to increase engagement.
Machine learning can be used to identify cohorts and dynamically create micro-segments by factoring uncertain customer behaviour. It willl be able to the predict the movement of customer from one micro-segment to another and the actions that a customer would take. Instead of a reactive approach, ML can help marketers deploy predictive approaches and automatically lead the target audience (whether a loyal customer, a prospective customer, or someone who is about to churn) in the most suitable buyer journey via sophisticated customer retention models.
Right content, right channel and right time. There has been exponential growth in content with widespread adoption of content marketing and customers have just too much to read. While each content piece seeks undivided attention of customers across all the touch points, it is imperative for marketers to accurately cater to customers’ needs.
Machine learning can be applied to content marketing to make the content truly resonate with the target persona. Here are some of the crucial components of content marketing where ML can be leveraged:
- Insight. ML coupled with text analytics can be used to process millions of content pieces at granular level, uncover the common elements that generate maximum engagement, and predict the type of content that will have laser sharp focus on audience.
- Time. Based on the level of engagement, ML will be able to automatically schedule content at the right time for maximum impact. From newsletter, in-app notification to social sharing — each content delivered at the best possible time.
- Channel. We always strive to find out the most effective channel to reach out to target audience in terms of ROI. ML can precisely tell marketers which channel will generate maximum engagement (e.g., push notification, email or any IM app), and this can be used to gain attention without cluttering all the communication channels.
Predicting customer lifetime value. Customer lifetime value (CLV) calculation requires prediction of future events by considering various types of data — demographics, actions on websites or app, previous purchases, interaction with marketing campaign, GDP of country and much more.
While applying machine learning for CLV calculation with higher accuracy, a multi-task learning approach can be used to build small, hyper-targeted and distinct ML models. It will be optimised for different types of future interactions which will act as a factor for campaigns in the customer life cycle. For example, if a customer with high lifetime value is about to churn, then the customer retention model would trigger a personal approach like one-to-one call instead of an email.
Bottom line
ML will truly transform marketing automation by moving the guiding principle of data usage from rear-view mirror to the sight of the future. It will start off by learning about consumer preferences, choices, habits, and finding patterns to predict the future action. It will not only help marketers answer questions with accuracy, but also move further by uncovering the vital questions that they might not be asking.
With advancement in technology, ML will also be able to consider macro factors (weather, inflation, government debt, etc.) along with micro factors, and in effect radically change the way we communicate and deliver value to consumers. Companies that will be able to align marketing and data via cutting edge machine learning technology will always stay ahead and get a bigger slice of the pie.
No comments:
Post a Comment