It’s been five years since Gartner called Prescriptive Analytics “the final frontier for Big Data, where companies can finally turn the unprecedented levels of data into powerful action.” Will prescriptive analytics be the final frontier for health marketers?
Pharma Marketers’ Analytics Adoption Curve
Let’s start by understanding where pharma marketing is on the analytics adoption curve.
Do you retroactively analyze what happened with your media campaigns? If YES, you’re using Descriptive Analytics.
Do you evaluate, in a deliberate and controlled fashion, why it happened? If YES, you’re using Diagnostic Analytics.
Do you have enough data and intelligence to predict what is most likely to happen? If YES, you’re using Predictive Analytics.
Do you use logic and math to decide how to make it happen? If YES, you’re using Prescriptive Analytics.
While almost all pharma marketers are using Descriptive and Diagnostic Analytics to a certain extent to measure and evaluate their media campaigns, no one to my knowledge is using Predictive and Prescriptive Analytics approaches with measurable success in health. The approach has been used successfully in mass brands for marketing, and in pharma for supply chain management, customer service modeling, and even clinical trial management, but not for marketing and media.
What Prescriptive Analytics Could Look Like
And yet we talk about prescriptive analytics because the promise is extraordinary. Let’s imagine the possibility.
For example, every month nearly 100 million people visit Healthline. With big data models, we would be able to predict the condition profiles of our readers, what content they’re going to read, what ads they’re going to click on, and what brands they will ultimately buy. With that information, we could use AI in real time to individualize and personalize the experience for each reader and determine which ad unit(s) to show them, in which platform (site, social network, newsletter), and in what order, to influence specific behaviors such as visiting a particular website page, and delivering a specific outcome like a new brand start.
Sponsoring brands could manage the reach of their ad campaigns and forecast the increased demand generated by their ad spend far more accurately. The prospect is enticing.
Limitations of Prescriptive Analytics in Health Marketing
Before we fully embrace the allure, let’s recognize some of the limitations of predictive analytics in health marketing.
The scale of the data just doesn’t exist. Sure, if we’re talking about a condition like flu, which affects large swaths of the world’s population, we may be able to build models that work. But what about advanced metastatic breast cancer? Or ankylosing spondylitis? The numbers of cases of these conditions are small, so the predictive power is far from confident and the models fall apart.
Media campaigns don’t really relate to each other. To get the historical learning at scale, we’d need to combine the learning across all media campaigns, across all brands, even across all conditions. But brands are different. Insights are different. Messages and levers are different. The experience from one campaign doesn’t apply to another, even though the holistic experience can yield powerful insights.
The patient journey is truly individual. Perhaps, with really big data sets, we can find demographic or behavioral markers that correlate with condition profiles, advertising sensitivity, or likelihood to get on brand. But the correlation won’t be high, and won’t necessarily portend future behavior. It’s because health is multifactorial and every person is truly unique. Not everyone with advanced metastatic breast cancer is the same. They live under different economic conditions, social determinants, value systems, and risk curves. There may exist a handful of people who mimic similar behaviors, but never segments large enough to truly validate the effort.
Last but not least, patient privacy protection limits widespread use of patient data. Industries leveraging prescriptive analytics with machine learning and artificial intelligence have significantly fewer barriers than health and pharma marketers. HIPAA makes data matching and outcomes-based analysis extremely challenging. Add to that the data loss now with GDPR and CCPA, further diminishing the sample sizes available for testing and optimizing, and making universal recommendations impossible.
How Pharma Can Be Data-Driven
Instead of chasing big data models, marketers can be predictive and prescriptive with their media campaigns by using descriptive and diagnostic data combined with human computing: human discernment, sensitivity, and gumption.
Here are our four principles:
Live by the data. Continue to invest in machines to run the decision rules behind your dashboard, alerting you when a campaign is in the “red” or “green” zone. But make sure you have empathetic humans available to receive those alerts and act on them.
Surround yourself with a team of competitive hard-core analysts. They’re hard-wired to go beyond current standards. They’re your human engine, to fine-tune future results and predict future results.
Be smarter about how you use historical data. Ask “What worked?” Also, scenario plan by asking “What could work better?” to understand and perfect campaign effectiveness. Be strategic. Start with the end goal in mind, test and experiment, and deftly piece together the answers to “how” and “why.”
Optimize actively, not passively. When your agency or partner presents their optimization approach, don’t gloss over that section. Approach optimization as a critical strategic lever. Build rapid review cycles, and learn and codify that learning everyday.
Prescriptive analytics gives us an ideal to aspire to, but not something we can realistically participate in today. However, we can take the key principles from that ideal and incorporate them in the way we run our campaigns. We can adhere to scientific rigor, transparency, and connecting humans to the numbers to rethink and refine campaigns. For now, it may not be a golden goose, but it’s not an empty promise either.