From RankBrain to Google Attribution, Kafka and Spark…it feels like machine learning is fast becoming the most prominent technology concept in digital marketing.
Machine learning is the principle of using self-learning algorithms that can analyse and find patterns in big data sets. These algorithms evolve behaviours based on empirical data and can adapt to new circumstances.
Machine learning is often mentioned alongside digital marketing’s other 2017 catchphrase, Artificial Intelligence (AI). It’s worth stating for those with only a passing familiarity with these concepts – they are not the same thing! AI is the much wider concept of evolving technology that behaves like humans. Machine learning can at best be described as a subset of the general science of AI.
If 2017 was machine learning’s ‘shake-down’ year, it will be interesting to see whether digital marketing’s clutch of machine learning advocates can turn sexy tech into legitimate, every-day use cases for marketers.
This is particularly the case in performance marketing, where machine learning technology is yet to turn column inches into common uses. Delivering on the ground technology from a visionary concept is rarely seamless. Yet in an industry yearning for innovation, machine learning holds the key because it has the power to bring much-needed automation and efficiency to the day-to-day lives of performance marketers.
Forecasting is a great example of this. Predicting uplift for a given promotion and to a given spend target is one of the most common-place tasks for affiliate marketers. It’s also an inherently manual task, reliant on an individual’s knowledge and instinct, and often performed in complicated spreadsheets that are the sole domain of their creators.
Day-to-day forecasting for affiliate marketers across the industry needs to be more empirical. It should be based on machine learning algorithms, which can predict to a publisher and campaign level increased performance based on the type of promotion, changes in spend and seasonality. This sounds straightforward, but to ensure high levels of accuracy complex algorithms are needed that can learn and adapt to new data sets.
Another example is Spend Management. Channels like Search and Display have allowed advertisers to manage their marketing spend more effectively than affiliate has. This is often blamed on affiliate’s diverse publisher base. However, using machine learning algorithms to dynamically process historical performance data, it’s possible for advertisers and publishers to link spend to performance accurately, allowing affiliate marketers to understand the implications of spend changes before they are made. Imagine a world where a CPA change is made based on forward-looking, empirical analysis collated on your behalf, rather today’s status-quo – make the change and then use retrospective analysis to understand the impact of the decision. The impact of the decisions should be known before the decision is made! This is the power of machine learning in the affiliate space.
Machine-learned technology will become increasingly common across the digital marketing industry in 2018. The performance (or affiliate) industry needs to embrace it. Machine learning has the potential to change the way all performance marketers work, with genuine use cases for the every-day challenges of our roles.