Are you having a marketing problem you cannot solve? I bet the answer to solving it is somewhere in your data.
What are the biggest barriers to using AI in Pharma?
Does this sound familiar? To many people AI is a buzzword; they don’t really understand what it is, what it can do or how it differs from older linear approaches. Many Analytics Managers have been around for a while, and many have failed to keep up to date with what is going on with regard to this evolution in the field of math.
Most Pharma Analytics Managers only understand AI in terms of what they are aware of, and appear to assume that it is a new and superior way of doing what they are doing. However, in reality, AI is night and day from all the usual suspects such as regression analyses, multivariate statistics and the like. As a result, teams whose leaders only feel comfortable with the old ways keep plugging away studiously with the linear analyses as that is what they know and understand. They then wonder why they cannot find that nugget of truth that is the very thing they need to create stronger growth in their sales and marketing results.
Lack of understanding of what is possible
People who grew up with older linear analyses are finding it hard to get their head around the possibilities Artificial Intelligence presents as they keep thinking of it as a version of what they already know. However, that is a mistake.
Of course, with AI we can deal with all the usual suspects, but with far more accuracy, such as:
• Understanding optimal strategic direction
• Enhanced value proposition identification
• Enhanced competitor differentiation
• Ability to see which segments will offer the most value to brands as well as what levers to pull to maximize the growth from them
• Optimal resource and budget allocation for maximum market share gain, revenue and profit
• Optimal drug pricing based on the current realities of outcome and value
What amazes me most in our weekly math meeting between our Professors and our math team (and me – I don’t like to feel left out) is the breadth of challenges we can now solve using AI.
Not only can these systems have hundreds of simultaneous conversations but they don’t forget anything from any conversation and can answer each question thoroughly and completely. We can use AI for pricing drugs to ensure superior market access. We even discovered we can use it to get far better results out of Pharma mergers and acquisitions than is currently being achieved (in terms of making the best match, getting the best price for the acquisition and knowing what to keep and what to lose).
Can any statistics and linear math do all of these things? I don’t think so.
You don’t always need new data, you need to be able to get more out of the data you have… a lot more.
In the old days we had to collect data and apply linear math to that data, but now that we use AI, in the majority of client cases we find we don’t need any expensive new data as our clients already have all the data they need sitting under their noses. However, they don’t have the right tools to get value from it as the skill set for doing this has changed dramatically over the past 2 years.
You are bound to have loads of data, both structured (Veeva, Marketo, market research quantitative, IMS, Payer) and unstructured (market research qualitative, free text call center, emails from reps to docs, etc). All of these types of data can be used together with AI, and the beauty of AI is that it learns and modifies the algorithms itself as it grows. Again, have you ever seen linear approaches change their algorithms themselves as new data is added? I don’t think so, and if you have, perhaps it was happy hour and you saw more than that.
Barrier 4
Knowing what to do with the data
You can create different AI approaches to your data but the secret I have understood from my math meetings is two-fold:
1. Different data screams different AI approaches, so you need to know which AI approaches are best for your specific datasets and combination of datasets.
2. A combination of AI approaches often works for even stronger results.
There are many different approaches and sub-approaches within each, and unless you know what you are doing, you could be applying the wrong AI approaches to a dataset and not achieving the most out of it that you could be. See Prof, I do listen in these weekly math meetings!
Conclusion
Part of the problem in applying AI to Pharma reminds me of something I read about a flea that was put into a jar. Since fleas are capable of jumping approximately 30 times their own body length, a lid was put onto the bottle to contain the flea. Whenever the flea jumped, it hit the lid of the jar, and eventually it figured out how to jump to a level just below the lid. After a while, the lid was removed and the flea was free to go, but by now, it could no longer jump high enough to escape from the bottle. Physically it could, but the flea now believed that if it jumped any higher, it would get hurt.
The story is a metaphor, of course. However, it illustrates the concept that we perform not according to our abilities but according to the expectations we have that are put there by ourselves, or by someone else. This seems to mimic what I see in Pharma analytics today in quite a few ways.
For more information on the use of Artificial Intelligence in Pharmaceutical marketing, including Machine Learning, please contact the author – Dr Andree Bates at Eularis: www.eularis.com.
Found this article interesting?
To learn more about how Eularis can help you find the best solutions to the challenges faced by healthcare teams, please drop us a note or email the author at abates@eularis.com.