Using predictive personalization to improve outcomes

 

 

By Keri Hettel, VP, group director of analytics, Razorfish Health

 

 

In today’s world, consumers have come to expect personalized experiences – it is not a nice-to-have, but rather, a point of entry if you want to remain a competitive brand. At the crux of personalization is, you guessed it…data. Personalization can look like many things – an advertisement for a famous seafood restaurant in Miami, even though you live in Boston (but you did recently book a vacation to Miami?); or it might look like a kiosk in a drug store with a platform to stand on to find out your perfect in-sole thickness and density.

It’s often easier to imagine truly personalized experiences for consumers, but as healthcare marketers, we also know the importance of physician marketing. We have to remember that, in many ways, physicians are no different from the everyday consumer – they also expect personalized experiences tailored to their needs and interests. When it comes to marketing to physicians, you have the data to enable this level of personalization. The question is: Are you using your data to its full potential? Most likely, no.

Evolving the impact of your data
The marketplace is getting more sophisticated in how we collect, analyze, and use client data. However, to advance your marketing efforts, data analysis must move beyond descriptive data – meaning, marketing optimizations should not stop at taking an action based on past performance data. To stay ahead of the curve, data and optimization programs should include predictive personalization.

Predictive personalization at work
Predictive personalization is not just about what your client’s customers have done; rather, it’s what your customers might do based on what they have done, who they are, what their colleagues are doing, and so on. While most are not already doing this, the good news is you likely have a lot of the data to get started!

Enhancing efforts across channels
Have you ever worked with a client on utilizing predictive personalization to enhance their targeting efforts across online and offline channels?

Situation:
A client had launched a brand in a highly competitive and genericized category where there was also a significant amount of competitive promotional spend. During the launch year the client noticed that physicians who were trying the product, were adopting it; however, trial was not meeting initial expectations. After six months, product adoption was significantly underperforming forecast driven by limited trial.

A predictive personalized solution:
Most likely the client is already optimizing their marketing channels, shifting spend to the highest performing programs. They might be utilizing A/B testing to ensure the most effective email subject lines were used, or analyzing the types of content physicians were interested in to develop key messaging.

However, to really drive up the impact of optimizations, an agency should enable predictive personalization using a two-step approach:

First, there is a need to deeply understand current product trialists – who they are, what their prescribing behavior is across the category, what content they are interested in (e.g. email messaging, website content), and what channels they use and respond well to (e.g. field force, paid search, email). By building a unified database to bring together all of these pieces of data, an agency can enable a truly holistic approach to their analysis.

Second, to predict additional physicians who are not currently prescribing a product, an agency will want to utilize advanced statistical modeling techniques and machine learning. These two tools allow the target audience to be segmented into low and high priority groups based on the intersection of their current behavior and predicted future behavior.

Identifying high priority physicians:
The models should be developed to deliver strategic groupings of physicians as the output. Each group should then be addressed in very different ways strategically – hence a tailored, personalized solution. For example:

• Physicians who don’t currently write, and are not predicted to write in the future become low priority. All spend should be shifted away from these physicians, enabling the client to save money and/or shift spend to high impact groups.
• Physicians who don’t currently write, but are predicted to have a high likelihood to write in the future become high priority. Shift a lot of effort towards these groups – focus on determining the right channels to reach these physicians, the right message to shift their behavior, and so on.

The level of predictive personalization that comes out of this analysis is in large part determined by the data and knowledge put into the analysis. In the case above, an agency can use demographic data, geographic data, writing behavior across a category, and content/message preferences. This robust knowledge of a select group of physicians allows the agency to really get to know their audience. Not only does this help with the accuracy of the model, it also allows for very actionable outcomes. In this case, the agency was able to eliminate spend that was having no impact on their targets and to shift their focus to creating content, messaging, and communication plans for specific physicians who they were more likely to move the needle with. Without this level of personalization some optimizations could still have occurred, but they would have been less impactful on the bottom line.

Not all agencies or brands have all of this data at their fingertips, but it’s likely that at least some of it is available. You probably know more about your consumer than you think you do. Get started with a simplified model to start dipping your toe into predictive personalization. For example, you may be able to look at the overlap between field force calls and emails received, both of which require you to know exactly who the recipient is. Over time, you can develop and implement a measurement program that allows you to collect a more robust data set and evolve the level of predictive personalization you apply. Regardless of how much data you currently have available it is key that you make the most of the data that you do have.

At the end of the day make sure your data is always working (hard) for you!