Many apply traditional media tactics, metrics and segmentation models to a digital world. However, data and algorithms are far better than any human - no matter how experienced or insightful - in identifying and targeting the right consumer with a message or experience that leads to a conversion. The reason for this is simple: the platforms have the power to analyse data at unimaginable scale and identify patterns that would escape the human eye. Here are four practices that can help marketers to unlock the full potential of machine learning in customer engagement: The major programmatic and social media platforms - including Facebook and Google - have grown into data empires that understand customers’ behaviour at granular levels of detail. Their algorithms can help you identify someone looking for a car loan or a holiday booking, then target them with a timely message tailored to their needs and where they are in the customer journey.Yet this powerful technology is available to just about anyone who can afford to pay for a paid search ad or a social media ad. Leading marketers are therefore also looking at how they can leverage second-party data (for instance, data shared by partners) and first-party data to get an edge. Examples of first-party data that can help a brand to craft a more personalised, compelling engagement including customer relationship management (CRM) data and data signals collected from offline channels such as the point of sale or call centre. This data can make the platform algorithms even more precise, leading to better customer engagements, lower customer acquisition costs and more conversions. Companies that want to delve deeper into using first-party data face a range of challenges. They need robust data governance processes to ensure they comply with regulations such as the Protection of Personal Information Act (POPI) and the Global Data Protection Regulation (GDPR). They also need to think about how they will safeguard their proprietary data within today’s connected data ecosystems.It’s tempting for marketers to think that they know best, when their idea about whom the target customer is comprised of a blend of months-old market research, some historical assumptions about the customer base, and educated guesswork. Research and segmentation may, however, embed some presumptions that limit the brand’s ability to reach some of the most promising prospects. At a time when marketers need to spend every rand efficiently, it’s best to trust the algorithm to identify customers according to a range of behavioural signals and other markers, then target them with appropriate messaging using dynamic ads. Learn from the likes of Netflix and Amazon, which use deep learning to target content to people based on their behavioural preferences rather than age, gender or other demographic factors. Every marketer is aware of the traditional marketing funnel, which moves from awareness and engagement to consideration, conversion, and finally, loyalty. They often think of video or social campaigns as mechanisms to create awareness and spark consideration, tap into search and remarketing to tip customers from consideration to conversion, and use CRM-driven direct marketing to build loyalty.Most understand that customer behaviour in the real world isn’t as linear and tidy as the funnel model would suggest. Consumers may vacillate between consideration and conversion for months before making a buy, for instance, or be lured away by a competitor at the loyalty stage. Yet delivering customised messages and engagements for the many permutations of possible customer behaviours and needs at different parts of the customer journey was impossible before machine learning. With machine learning and dynamic ads, however, it’s no longer as necessary to follow a rigid funnel model to engage effectively with customers and prospects. Instead, marketers can look beyond the funnel and deliver the right message for the customer’s context on the fly. They can use triggers such as scarcity or authority to encourage customers to convert, based on their behaviour. Forward-thinking marketers are starting to look further than traditional digital platforms when it comes to fuelling machine learning with customer data. Voice and visual search are starting to play a key role, though the platforms have some distance to go to offer an integrated approach to managing voice, visual and traditional search to drive better outcomes. This trend is accelerating not only because of the use of augmented reality and voice search on phones, but also because of the explosion of Internet of Things devices like smart cars and smart home technologies. Marketers should be thinking ahead to how embedded cameras and speakers in nearly every home device could change customer engagement in the years to come. The bleeding lines between offline and online are also likely to lead to an explosion in the data available to marketers. The likes of Amazon Go and Alibaba today offer experiences where people can check into a store with an app, take the products they want, walk out and be charged without needing to pay at a point of sale. Scanners and cameras watch shoppers as they move through the store and the AI keeps tabs on the items they have taken from the shelves. South Africa lags this emerging trend so far, which is understandable, given that the technology remains complex and expensive. However, now is the time for local organisations to think about how they can start to bring together offline and online channels and data. The leaders that get it right have the opportunity to offer an omnichannel experience that is consistent and personalised, whether people are shopping in-store or at home. With the technology and data digital marketers have access to today, they no longer need to limit their potential target audience to a set of personas or segments derived through customer research. Masses of data and powerful machine learning tools can understand and predict people’s behaviour and needs with more accuracy than any tools marketers relied on in the past. Yet unleashing the potential of this technology is as much about embracing a new mindset as it is about learning new technical skills.