Written by Earlene Worrall & Kevin Troyanos
Machine Learning Based Techniques Have the Power to Significantly Augment Traditional Research Techniques – Not Replace Them
In our respective roles as analytics and research practitioners, we interact with a wide array of practicing market research professionals in the health and wellness space. In recent months, we have noticed an underlying hesitancy to adopt algorithmic-based research techniques. Many traditional life sciences market research professionals have directly confessed a fear that big data and machine learning will take over their jobs in the coming years.
As we see it, those fears are unfounded — and perhaps even a bit myopic.
We believe that advanced analytics and machine learning will play an important role in the future of life sciences market research by complementing — not replacing — the capabilities of traditional research. In our work together at Publicis Health, we have seen the true power of a combined, iterative approach to customer research.
Machine Learning as a Tool to Enhance Human Understanding
So how exactly can researchers harness the power of data and algorithms to optimize the caliber of insight achieved? The key is to strike an artful balance between the inherent strengths on each side (Human and Machine) in order to achieve a depth of understanding not possible with either approach alone.
Algorithms are great at analyzing data and predicting specified human behavior based on stimuli, and can efficiently process data over thousands of variables in order to detect human responses. By using data to identify existing behavioral patterns, they can readily predict future behavioral outcomes based on features captured within the data. Such predictions offer brands an unprecedented opportunity to not only anticipate and meet consumer needs, but to produce strategic communications aimed at positive behavior change where it matters most.
The rapidity and accuracy with which algorithms analyze data and predict future behavior simply cannot be rivaled by any human. Machine learning dramatically accelerates the time to insight for researchers, by taking a vector of inputs based on real-world data and spinning out probabilities of future behaviors – at massive scale. If market researchers want to understand consumer sentiment in real-time, predict consumption patterns and even outbreaks, or understand which factors impact the adherence behavior of entire patient populations, algorithms are the tool of choice.
Social listening platforms are a good example of how some of these algorithms are being put to use by many market researchers today (whether or not they realize it). For instance, researchers and marketers have the ability to tap into and understand customer sentiment, buzz, and trends in real time, at impressive scale. They can gather data on the sentiments expressed about certain drug classes or brands, and use it to target hyper-specific patient groups, fill urgent information gaps, and inform future outreach campaigns. However, as any social listening expert will tell you, the human lens is essential in order to distill truly actionable and disruptive insights from that data, regardless of the platform’s technical abilities.
Explaining the “Why” Behind the “What”
No matter how big the data gets, or how advanced the algorithms become, the fact remains: to be human is to be complex. We cannot stop at what the data suggests. Rather, we want to harness the power of that data in conjunction with deep human understanding from qualitative research. Qualitative techniques permit us to explore all the potential influencers of behavior, in order to interpret the data optimally.
Explaining the “why” behind the “what” people are doing requires us to step into our customers’ shoes to understand their experience, how they make sense of the world, and what shapes their behaviors and perspectives.
So once the data identifies existing behavioral patterns (or predicts future behaviors), it’s time to unleash qualitative methodologies to get beneath the surface to underlying barriers and motivators that truly explain those findings. Qualitative insights enhance understanding of the data and, in our experience, can materially change interpretation of observed or predicted behavior.
For example, a classification algorithm (such as a Neural Network, or Support Vector Machine) trained on historical prescribing data may predict that a group of physicians have a high likelihood of adoption of a certain medication. Deeper qualitative dimensionalization of this group may confirm that this high degree of predicted behavior indicates true brand preference for that group of prescribers – or, for a subset of this group, it may show that this historical and predicted behavior is in fact circumstantial and actually driven by other outside influences (such as the brand preference of patients, or the influence of payers). This qualitative understanding dramatically shifts the target of effort and strategies for influencing customers toward the desired behavior.
Machine Learning Has Limited Functionality without Human Interpretation
While this is all very exciting, we haven’t yet answered the question on researchers’ minds: what does this mean for my job?
It means you shouldn’t panic. In fact, you should be excited by the research synergies that are now possible.
Machine learning makes it easier to find patterns, or make predictions, but it will not tell us the underlying reason why something is the way it is beyond the data alone. For all its utility, machine learning algorithms are inherently limited on the universe of data on which they are trained. It’s up to researchers to uncover and dimensionalize the driving human forces behind predicted and observed behaviors.
And while there is immense utility in an algorithm’s ability to analyze whether a new tactic is or isn’t likely to drive behavior change, they can’t tell marketing and research teams how to enact that change. Overlaying qualitative research will identify the strategic levers available to meaningfully connect with the selected target and influence behavior. Strategy, communication, and branding are still very much human capabilities.
Unprecedented actionability results from leveraging the art and science of research through a combination of machine learning-based analytics and deep human-centric insight. We are able to hone in more precisely on most promising predicted targets, and determine the right way to communicate with them to appeal to their in-going mindset and positively influence behavior.
Embracing New Tech to Lead in Human Understanding
The ultimate goal is to influence positive behavior change as it relates to health care. Any actionable strategy derived from research requires: (1) a description or prediction of behavior; (2) understanding the ‘why’ behind the ‘what’; and (3) recognizing how to most effectively influence behavior. Algorithms can be an invaluable tool in identifying behavioral patterns and developing predictions, but the other two pieces of the puzzle remain squarely under the purview of researchers and marketers.
By harnessing the strengths of both approaches in an iterative way, research professionals can take their depth of insight to new levels. Only by operating synergistically can we stay abreast of the pace of change, and inspire the behavior shifts we want to see.