Machine learning (ML) and artificial intelligence (AI), once mere buzzwords akin to flying cars and moving sidewalks, have become commonplace in nearly every industry. From identifying webspam to maximizing revenue through targeted marketing, this technology is integral to today’s business practices across most industries.
In the life sciences industry, machine learning has become an active feature of drug discovery and clinical trial success. In drug discovery, machine learning predicts the effectiveness of different compounds for a specific disease, accelerating prior time-consuming trial-and-error methods. Machine learning has also been used to speed clinical trials and increase success rates by helping to determine patients most likely to benefit from a drug candidate. In fact, ML-supported biomarker selection has been shown to triple the overall likelihood of approval of a drug from phase I to approval.1
Despite this impressive value delivered in R&D, why are life sciences companies only just starting to adopt machine learning to drive commercial excellence? A handful of myths exist surrounding machine learning. Let’s unravel them:
1. Commercial processes are too complex for machine learning
Life sciences companies have a lot of complex data. It’s incomplete, distributed across channels and sources, owned by various departments, and growing exponentially. The industry is unique in that the “customer” is multi-faceted – healthcare professionals (HCPs), healthcare organizations (HCOs) and integrated delivery networks (IDNs), payers, pharmacy benefit managers (PBMs), patients, and so on all play a role in the buying process. The complex nature of commercial processes in the life sciences industry has led to the belief that sales and marketing can’t benefit from machine learning technology. However, the need to accurately find predictive patterns from a large number of data inputs with complex relationships is exactly when domain-specific machine learning should be applied. If the relationships among the data were easily understood, a simple spreadsheet would suffice. Machine learning can synthesize vast amounts of distributed data to help sales and marketing optimize customer engagement across channels.
2. Machine learning is impersonal
While technical aspects of machine learning can be intimidating, it results in improved personalization of both strategy and messaging, creating finer segmentation and communication at the individual level. (Think about your personalized shopping experience on Amazon.com.) Many wince at the idea of a fully-automated approach, in which there is no opportunity for intervention. Machine learning doesn’t limit you to one end of the spectrum or another. Most of our customers prefer to balance human oversight with automation. It allows them to apply constraints or override specific outputs. For example, if a model predicts the optimal cadence between a set of communications to a particular HCP target, the brand manager may take that into consideration and still choose a difference cadence based on an outside factor or “gut feel.”
3. Machine learning will further overwhelm reps with more information
Sales reps are inundated with information – formularies, prescription data, analyses from headquarters, and more. It’s difficult for them to sort out what’s most relevant and timely to move the needle. Machine learning can synthesize and identify patterns in vast amounts of data much more efficiently and accurately than a person can. Consider a rep during pre-call planning: she’ll look at a few different data points to help plan the coming week or perhaps make an adjustment to messaging for a particular HCP. With the help of machine learning, the rep will receive distilled insights from all of the information available to understand the optimal message sequence and mix of channels to maximize chances of an effective HCP engagement.
4. Machine learning and sales reps cannot coexist
That machine learning and sales reps cannot coexist is a recurring theme in most industries, including life sciences, where technology is viewed as a threat to the sales force. There’s little doubt that machine learning makes reps more efficient, largely reducing time for administrative and pre-call planning tasks. While some companies may take this opportunity to drop the lowest performers or expand the role of an existing sales force instead of hiring a new one, experts assert that the rep role is “evolving, not dissolving.” In fact, sales reps are a core part of the machine learning process. Rep input is vital to learning models, providing information for machine learning algorithms to synthesize data sets and identify patterns from which suggested actions are made. Personal engagement between sales reps and HCPs continues to be a significant communication channel (life sciences companies spend up to $15 billion on sales reps), and machine learning is rising to partner with the sales force — not replace it.
5. Personalization isn’t possible without HCP-level prescription data
While rich HCP-level prescription data available in the U.S. can be a very useful input for a machine learning model, a range of alternate data types can be used to construct a clear picture of HCP-specific thoughts and decision patterns. Additional data such as digital activity, meeting participation, and salesforce-recorded feedback is also trackable and can be applied to machine learning algorithms. These algorithms can recognize HCP-level patterns through the complexity of account-level or brick-level data to reduce time spent in analyzing the current needs of specific groups.
While these machine learning myths linger, individuals at many organizations are beginning to realize their untruths. Machine learning is becoming a certainty moving forward as a way to simplify and support ongoing efforts in sales and marketing. By empowering reps and marketers with more pertinent information for each and every HCP, machine learning stands to help the industry adapt to marketplace changes faster and open communication between stakeholders. As more and more companies invest in technology to improve CRM, data input, and analytics, many are already reaping the benefits from machine learning, providing novel and actionable insights that make their organization that much more efficient and valuable to customers.
If you’re interested in exploring this topic further, check out the recording of a webinar we did recently about using AI to achieve commercial excellence in life sciences.
- “Clinical Development Success Rates 2006-2015” David W. Thomas, Justin Burns, John Audette, Adam Carroll, Corey Dow-Hygelund, Michael Hay. Biomedtracker, Sagient Research Systems, Informa, San Diego, California, USA. Biotechnology Innovation Organization (BIO), Washington, District of Columbia, USA. Amplion, Bend, Oregon, USA.