How machine learning has evolved digital advertising

Machine learning is the practice of programming computers to “learn” without manual input and has been fundamental to the growth of real-time bidding (RTB), which lies at the very heart of programmatic advertising. Relying on a series of algorithms that allow computers to adapt and learn from observing certain data sets and behaviours, machine learning can predict a consumer’s likelihood to purchase when exposed to an ad.

Back to the future

Arthur Samuel wrote the first machine learning program way back in 1952, in which a computer learned to play checkers by building up patterns of winning behaviour and strategies that were most likely to secure a victory. Machine learning steadily progressed from there, with events such as the introduction of Explanation-Based Learning (EBL), where a computer analyses training data and created a general rule it can follow by discarding unimportant data, as well as the creation of NETalk, where a computer learned to pronounce words the same way a human baby does.

The 1990s is where machine learning moved away from machines relying on a knowledge-driven approach, shifting instead to a focus on learning with data. At this point, data scientists began creating programs that allowed computers to analyse large amounts of data and draw conclusions from the results.

For example, a machine learning program can learn to recognise pictures of cats when shown a sufficiently large number of examples of pictures of “cat” and “not cat”. I’m not sure why, but that’s apparently what machine learning scientists are into. No, honestly. Or an autonomous driving system learns to navigate roads after being trained by a human on a variety of types of roads. As the program gains “practice” with the task, it gets better over time, much like how we humans learn to get better at tasks with experience. The machines aren’t quite as good as humans yet, but they have their own global race and they are learning at a geometric rate.

Machines in marketing

The same core value applies to machine learning in marketing, with the intent of a machine predicting the strategies that will lead to sales. Historical data is drawn into a computer from a variety of sources, evaluated by a set of algorithms, and used to determine which ad to serve to a browser or potential customer.

As programmatic adoption in the region continues to rise, machine learning will continue to be at the forefront of accurately predicting human behaviour – specifically their ever-changing wants and needs. As algorithms become more refined, machines will be able to predict things more accurately, like who is most likely to purchase, when they are likely to be in the mood to buy, and what they are most likely looking for. Static data rules will no longer be a viable way for marketers to use data, as machine learning provides much deeper insights in real-time.

With massive amounts of data now feeding into marketing systems, refinements to a campaign can be made as required, reacting almost instantly to a rise or fall in demand, shifts in browsing behaviour, geo-positioning data, and more. Ultimately, machine learning empowers marketers to optimise their campaigns, therefore absolutely maximising the investment that their company or client has made.

Moving forward: Intelligent marketing automation

With big data now becoming data obesity, the challenge facing marketers is which signal to pay attention to. Discerning which data points are most useful for a campaign will become an important role for machines to take on, with algorithms driving the outcome. Machines will continue to take more of the guesswork out of data-driven marketing, applying rules for certain data sets, and elevating one above another depending on changing needs and circumstances.

Personalising customer experiences is also at the core of modern marketing and ML can power this, even across multiple platforms. Rather than relying on a small focus group to draw conclusions on customer behaviour, machine learning can take on board a whole body of data across different touchpoints and optimise it to draw a very wide range of conclusions. This, in turn, allows for a much more precise focus on the individual – their likes, dislikes, browsing behaviour, and more – and that knowledge can be used to offer them better, more tailored online marketing.

Ultimately, machine learning will allow marketers to do what they do best. By removing a lot of the heavy lifting from a campaign and providing deeper insights that allow marketers to tailor their approach machine learning enables marketers to drive better outcomes and make better use of data, making sense of this wealth of information that we now own and ultimately reach the right customer at the very best time and with the most meaningful messaging. So just like builders will soon leverage robots to help them build houses, for marketers, machine learning is a tool for a more efficient, effective marketing process.

Zachary King is vice president of commercial in Asia.