An outline of machine learning capabilities within the field of digital advertising. And what it means for your patient recruitment campaigns – both in the future and right now.
First of all, let’s have a look at what machine learning is and how it’s used by Facebook – the king of utilizing machine learning for digital advertising purposes.
What is Machine Learning?
Facebook and other digital platforms have access to lots of information about how users behave while on the internet. In Facebook’s case, this comes from their enormous user base and how people behave while engaged with Facebook itself. Plus the huge amount of data they’ve gathered about how people behave on other sites that have the Facebook Pixel installed.
Using complex algorithms, Facebook analyses this data in order to improve the results it delivers, based on learning from experience. It’s literally a ‘machine’ – the hardware, software and algorithms that come together to create Facebook – that is ‘learning’ from the data it has access to.
How Does Facebook Use this Data?
Facebook looks at what you want to achieve with your campaign – for example, generating potential patient leads for clinical trials – then shows your ads to people who are most likely to perform the action that you want them to. Such that, where Facebook can see that people who match up to a particular set of data points are the ones who usually perform this action, they will then try to target other people who, from their previous behaviours, match those same data points.
Having seen the results over many campaigns I’ve been involved with, I’m convinced that, over time Facebook would be able to use its machine learning capabilities to target the right people, even without any manual optimization at all during the period the campaign was running. (As long as you understand and set your goals correctly within the Facebook system). That is, the ongoing interpretation of the new data by Facebook itself would be continually improving the results of your campaign, without requiring any input from you.
Of course, as a Facebook ads specialist, that conjures up the possibility of me being done out of a job by the very system I’m an expert in. However, what it really means is that the tools I’m working with have become so sophisticated, that the combination of expert human-devised strategy, with enhanced machine learning functionality, helps to deliver the best possible results.
Where previously I might have spent a lot more time interpreting the data provided to try and match my targeting to the right sort of people; now I can put more effort into experimenting with and optimising elements such as audiences, creatives, and different parts of the funnel along the patient journey. Plus I can give the system a ‘head start’ through using my experience and knowledge of the best methods for getting the most out of machine learning. Which leads to better results overall and maximises the return from your ad spend.
Working with Machine Learning for Optimal Results
Marketers such as myself, who have a strong background in the way things were done in ‘the olden days’ (ie more than five years ago), are usually well-versed in the methodologies of such things as ‘split testing’ and ‘beat the control’. (Things that originally date all the way back to the earliest days of advertising as a proper profession). Indeed, I’d suggest that, were Claude C Hopkins (one of the founding fathers of the methods of modern marketing) around today, he’d understand the basic principles behind digital advertising as being essentially the same as those he wrote about in his classic text ‘Scientific Advertising’.
The difference nowadays is that Claude C Hopkins and people like myself were experienced in analysing and managing campaigns manually – that is, interpreting the data for ourselves and making adjustments accordingly. Facebook and the other main digital advertising platforms have taken that basic idea and automated it to a level that would have been inconceivable even just ten years ago.
Fundamentally, the amount of data that Facebook has to work with now is so enormous, that a single individual – even one as skilled as Mr Hopkins – would never be able to get anywhere near to being as accurate with their predictions of how people will behave. Which is why Facebook’s machine learning capabilities help to keep it head and shoulders above any other advertising platform out there.
Of course, the system is only as good as the information you provide to it. One of the easily-overlooked elements for success when working with machine learning is that you have to give the machine the right parameters to work with in the first place. Thus ensuring that the ongoing optimisation through the algorithm is actually targeting the right sort of people – those who are most likely to register their interest in participating in a clinical trial. (Or whatever it is you’re trying to achieve with your ads).
Conclusion
Machine learning is not something from the future as imagined in science fiction. It’s here today, and has been for many years in one form or another. In order to deliver the best possible results from your patient recruitment digital advertising campaigns (and indeed, any form of digital advertising campaign), you need to embrace this fact and incorporate the possibilities of machine learning into your strategy. Which will ultimately provide for much better outcomes than could be achieved without it.
