masthead-services-03-1600x650.jpg

Blog, Comment & Insight

For Pharmaceutical Marketing Analytics

Are Alexa and Siri the Future of Data Analytics in Pharma Marketing?

Posted by Dr. Andrée Bates | Mar 12, 2018 5:40:21 PM

Pharma marketers have access to more data than ever before. Big data can bring together patient registry data, social media data, claims data, sensor data, CRM data and a wide array of other big data sets. And technology has advanced to be able to sort through this vast information.

The exploration of big data is a large job. Each data set alone can take weeks to months of data scientist work in data wrangling and algorithm creation and training to identify the valuable nuggets of what is hiding in the data. To make smart, data-driven decisions, you also need to know where you should be looking or at least what you should be looking for. Big data engineers and data scientists cannot do this job without expert pharma subject matter experts to guide them. We have seen that in several projects we have inherited from big data companies who have failed in pharma projects due to not recognizing this fact. But what if you could program a pharma marketing Siri in your data analytics?

To get more insights from data, most companies add more staff internally or in FTEs externally. That was the case with a recent RFP we participated in, which included a request for ad hoc analytics. The requesting team was asking for FTE staff augmentation on their analytics team, but we approached it slightly differently. Our bid for this was to build essentially a pharma-focused Alexa or Siri on the back of all their combined data. So, instead of just adding people, we proposed that we take their data, combine it, tag it, then write custom algorithms to address the typical types of questions their analytics teams usually got, such as:

  • Why did the business in a sales geography decline?
  • What factors drove the decline and how can I address it?
  • Why did the effectiveness of a certain promotional campaign change this quarter vs. prior year?
  • Which customer segments (could be HCPs or patients) are being effectively targeted, and where are incremental opportunities?

When new questions get asked that are not in the system, we create algorithms for those with our subject matter experts and data scientists, and build them in. That way, all the brand teams have access to self-service answers at the time they need them, with the latest, up-to-date answers. It’s like having Siri or Alexa hardwired into your marketing analytics.

You may find generic self-service platforms which promise drag and drop data in and get answers out. Anyone using these, however, quickly realizes they does not work as expected. I have now done trials of 3 of these platforms which are touted as cutting edge, and yet have found them sadly lacking. The answers were either irrelevant or erroneous. But the reasons for this are clear to me.

Why other attempts have failed

The main problem in all of the platforms mentioned above is – as it always is – the data. In the past, companies assumed if they build a data warehouse, gaining meaningful insights would be easy to do. Alas it is not. The data is not clean, there are disparate types of data that need to be connected, and the algorithms need to be created with a knowledge of what data is relevant and what is not.

Data wrangling is a long and time consuming job. Each data source must be sorted, cleaned, categorised and connected. This can take days to weeks to months per data source, depending on the state of the data. There is no magic quick fix yet despite many promises out there in the market. It takes time and effort from highly qualified data scientists and big data engineers.

The other challenge most have failed to meet is understanding the problem. Many companies with data visualization get confused with drag and drop and insight. Data visualization is certainly part of the process — but only one small part. It does what it says on the tin – visualizes data. I regularly see confusion between data visualization and data analytics, which are related but very different entities.

Data visualization is the front end of a data solution. Analytics is the back end. The visualizations are only as good as the data put into them. 

How to make the dream a reality

So, how do you achieve this utopia of data-driven Siri style Question & Answer analytics?

The first step is properly cleaning, sorting, structuring, imputing and embedding all the data. If you put garbage in, you’ll get garbage out. You need to put it into a big data platform and create relevant AI algorithms to identify and analyze all that data. We use machine learning to refine and improve results.

Now you have the answers, but you still need to input the questions. To translate the voice or text questions into code, you’ll need natural language processing algorithms to speak with the machine learning, and you will need computational linguists as we all know, language can mean different things depending on the syntactic structure. A keyword classification is not going to be sufficient for what I am suggesting. Finally, you’ll need a visualization and text dashboard to show the answer to the question in an easy-to-understand manner.

Over time, create a knowledge database of every question, so that the system knows what the most interest is focused on and create more algorithms to probe into more areas around that specific interest. 

The sky is the limit

The solution I’ve just described is just scratching the surface of what these technologies can do for pharma marketing. I have been simply considering what we can do internally within pharma. However, by collecting data from the digital world from physicians and patients, we can collect vast amounts of data on each customer. And we can combine these insights in the algorithms for a much deeper understanding of these customers.

For example, where do they spend their time? What content do they consume? What channels does each individual favor? What sites do they prefer? Who influences them the most? Then we can apply AI modelling within our marketing systems to serve up the right content, to the right person, in the right sequence, in the right channel.

This makes our marketing hyper-personalized – at scale. We can personalize each and every customer experience and interaction in real time. With data gathering sensors now being embedded in needles and pills and on devices for asthma and COPD as well as devices for diabetes, we are entering a very powerful hyper-personalized data era.

The next step is actually predicting what people want and where they will go, before they know it themselves. Algorithms are already doing this in banking and pharma marketers can apply the same principles.

These game-changers are imminent. Will you be leading the charge or will you be left on the sidelines?

For more information about how you can use these technologies in your marketing programs, please contact the author at Eularis.

Topics: Marketing Insights, Big Data, Artificial Intelligence, Advanced Analytics

Written by Dr. Andrée Bates