PM360 asked data experts what hidden insights companies may already have if they just know how to look for it as well as the best way to get buy-in on predictive analytics or next best action recommendations. Specifically, we wanted to know:
- As we undergo a data revolution in terms of the volume of data now at our exposure and the tools to analyze it, what is the biggest unrealized potential hidden in a company’s data assets?
- As companies turn to more predictive analytics and automated next best action recommendations, how can they best ensure these findings are being properly implemented in sales and marketing strategies?
Make the most of the data you already collect to support your key performance indicators (KPIs)—it has wider value. Companies often prioritize quantitative outputs and forget about unstructured data types such as customer feedback, user generated content, and employee insights. These provide helpful insights when it comes to existing customer experience, as well as competitor activities, employee satisfaction, and brand perception. First-hand customer insight in particular provides additional context for numbers and plug gaps—allowing you to turn analytics into actionable insights.
Analyzing this data can be challenging, but some of the newest technologies—AI, advanced analytics, predictive modeling, and natural language processing (NLP)—are now helping companies access these untapped sources. Leveraging this data enhances understanding of customer behavior, market dynamics, and competitive trends, ultimately boosting marketing strategies and overall operations.
While the potential within unstructured data types is huge for businesses looking to maximize the value and impact of their existing data, it’s important to consider how much data you are collecting and why. If you don’t know how to use it effectively to make improvements then it might just slow down your analysis, or even lead you to the wrong conclusions.
We have a data problem—there is too much of it. This makes it difficult to, as my old fraternity brother used to say, “separate the fly poop from the pepper.” When there is too much data and it is not integrated, or “fused,” you are drowning in it and it’s nearly impossible to find insights. Once you execute a data strategy plan and bring order to the chaos, you can do more high-level analytics, but in the meantime it’s possible to find value by doing simple stuff.
For example, at my last company I built a monitor that followed which of our drugs and competitor drugs were active in the market. At one point, we launched a line extension, but unbeknownst to everyone, the data service companies that maintained the master list of National Drug Codes (NDCs) misinterpreted our communication of line extension as a replacement and marked one of our most profitable products as discontinued. Fortunately, I had this check in place, corrected the problem, and saved the company millions of dollars. Until you’re no longer drowning, you cannot take advantage of all the low-hanging fruit that may provide the easiest and highest dollar-value impact on your bottom line.
Pharmaceutical companies have one chance to get their product launch right. Whatever revenue trajectory is set at launch—either meeting expectations or missing them—is the course that drug will follow for the rest of its lifetime. Any data that will help steer a launch in the right direction is pure GOLD. One dataset that has traditionally been unavailable is sales training data.
In prep for a product launch, sales reps are trained, certified, and sent out into the field to sell. In the past, it was hard to determine if training was effective in impacting behavior and, ultimately, business outcomes. But now, comprehensive, real-time data around sales rep behavior and performance is available, which can be used to make decisions that can change the course of a new drug’s success.
Correlating data about what training material reps are going back to, what information they are struggling with, and what messaging content they are using in the field, with individual sales performance data reveals incredible insight. With this data, Marketing, Sales, and Commercial Training leaders can identify performance promoters or diagnose areas of risk and take immediate action to impact rep behavior and improve drug launch performance.
Think about how Google or Amazon serve you an ad for something right when you are talking about it or searching for it. As businesses, we want to embed that type of personalized and predictive capability into our commercial functions. However, the selection of what’s next in Google or Amazon is not really a right or wrong answer—but there is when talking about medical products and potentially billion-dollar decisions. So, we shouldn’t completely trust predictive and next best recommendations in business like we would in our personal lives. Rather, it should be a vetted recommendation.
The most successful programs start with a proper pressure test and learning, and typically, six months minimum is needed to see if a predictive model is working. Along with constant refinement, control groups, A/B testing, etc., you don’t need to make a binary decision on whether to use predictive and next best models—that is a false choice many organizations get stuck in. Everyone should be experimenting in a controlled manner. Pick a brand, a sales team, a segment, and try something. Evaluate the effectiveness and move forward accordingly. The worst thing to do is nothing or to get stuck in paralysis by analysis.
Being strategic is vital when implementing next-best-action technologies. Two crucial steps in a successful implementation process include:
- Setting clear expectations and goals with relevant stakeholders
- Tailoring the campaign around specific objectives to increase effectiveness
We recommend a crawl-walk-run approach when utilizing these technologies. For example, we start manufacturers and brands with message optimization of their existing assets. This pragmatic approach also gives stakeholders credibility with the MLR process, setting the stage for more sophisticated future uses. While the impact from these technologies may not be a silver bullet, they will continually strengthen over time through optimizations: the more services, data, and insights gathered, the more accurate the analytics become.
Additionally, specific checks and balances should be implemented to ensure the desired data is used appropriately—in a privacy-safe and HIPAA-compliant manner. These safeguards can help assuage concerns from compliance and other stakeholders, but marketers must demonstrate that their next-best-action investments improve targeting and overall campaign effectiveness. Ultimately, if the campaign proves these technologies result in positive patient outcomes, that equals success.
Amid growing adoption of predictive analytics in healthcare marketing, brands are excited, curious, and sometimes intimidated by what can feel like the “black-box” nature of predictive modeling. Clients are increasingly seeking guidance on how to effectively harness data science to drive their marketing efforts while remaining aligned with their brand’s core strategy and simultaneously trusting the validity and utility of the results. This challenge underscores the fact that healthcare data science cannot operate in isolation.
Marketing predictive models should incorporate the minds of both internal and external strategy counterparts and even engage key opinion leaders within the healthcare industry to “gut check” the process and ensure strategic alignment. Project briefs for healthcare predictive models should mirror the process already in place for creative and media briefs, which already benefit from full team collaboration across data, analytics, brand strategy, and medical. Having cross-team strategic collaboration is key to addressing potential client concerns.
The collaborative approach ensures that, from initial project brainstorm to model development, the results of any predictive models align with real-world healthcare challenges, comply with regulations, and tackle critical issues in patient care that matter to brand teams. This ensures the resulting work is both statistically robust and practically valuable in enhancing patient and brand outcomes.