If you read my article on the barriers to growth that youmay be imposing on yourself, you should have a basic understanding of the difference between AI and what you are using now. In addition, you should also understand that anything is possible (and I mean anything). You probably have a truckload of data, and there are a bunch of AI approaches to be used with specific types of datasets, so where do you begin? What is the process when you really don’t have a roadmap?
1. Figure out what your core challenges are and define them
The point of all this is not to have AI for the sake of having it but to solve your challenges and get your company making the best decisions possible for profitable growth. Therefore, the most important step is to consider your biggest problems and define them. You may not even know the cause, so it could be simply, “Why am I not getting trial with my new drug?” or “Where is the pocket of opportunity within my market that I am missing?” It could be more specific: “Which combination of messaging and channels will help me achieve specific revenue results?” or “How can I find more rare disease patients?” You need to begin by defining your top challenges.
2. Find all the data you can get your hands on
You know you can get it. You are drowning in it. You cannot produce results without data, so get as much as you can. You don’t need all of it to figure out what you can do – just samples of as many datasets as you can access is enough at this stage.
3. Understand what different data can uncover
Now, you have a lot of data sitting there or that you can access. What can this data do? You no doubt have heard a lot about ‘Big Data’ and how valuable it is to your company. Many people do not understand the difference between ‘Big Data’ and ‘Small Data’.
Essentially, ‘Big Data’ does not fit in a spreadsheet, nor does it fit on a single server. So if your data is in spreadsheets or on a server, it may be a lot but it is not ‘Big Data’. However, quite frankly, whether your data is big or small, it doesn’t really matter. What is important is knowing what the difference is and what can be done with each type. So, let’s start small.
‘Small Data’ is manageable size-wise. It is also often structured (like Excel, with rows and columns) and can come from Veeva, your distributors or other systems. No doubt you have analyzed this data for many years and it may not be that sexy right now. However, it is essentially for understanding your customers and your performance.
‘Big Data’ is wild and sometimes a little disobedient, unstructured, too big to fit on a single server, and constantly changing. You can see why people often prefer ‘Small Data’. Nevertheless, within its wild and disobedient ways are sparks of genius that are often unnoticed. It is likely to be what customers are saying to your call centre staff, on social media, or in other places. It is a great mass of confusion but also a great opportunity if you can figure it out. To do that, you need the right people…and we have found that the right people tend to be theoretical physicists and data scientists.
4. Understand different types of analytics and implications of each
I have described the evolution of analytics for the past decade since I began doing guest lectures in business schools. It started with Descriptive – a little like Google web analytics – which describes what happened e.g. ‘X’ people came to your website, ‘Y’ people clicked on a page, and so on. It was used to compare reps – ‘X’ mins of a detail by rep ‘X’ and ‘Y’ by rep ‘Y’, and so on.
Then we moved up to Diagnostics to understand why things happened as best we could. For this we included market research analysis. Thereafter we moved to Predictive analytics, which is predicting the future. What’s missing?
Prescriptive analytics, which is prescribing the future, and allows you to change your future results by prescribing a set of actions which enable you to change your future results. It includes the other components but adds far more accuracy to the diagnostic and predictive, and far more power as it can run through billions of combinations of data and see exactly where the problems lie and how to solve them. For this we use Artificial Intelligence as it is the only way known so far to get the most accurate results possible today.
5. Work with people who know what they are doing
Finding the right people to do this is quite possibly the biggest challenge you will be faced with. It certainly was for us, and we are an analytics company that’s been around for well over a decade, and is constantly changing and growing. AI has been the biggest change for us in our history and the new and exciting applications we are using AI for amazes me on a daily basis.
However, the biggest challenge we had to solve was finding leading Professors in theoretical physics and data science, as well as Linguistic professors for some of the phases of our unstructured data analysis, who have the level of ability to do what we needed. This has not been an easy road but we persisted, with years of frustration at times, and it was well worth it.
We now continue to push the boundaries of what we are doing every week. In our weekly math meetings with our Professors and math team we spend half the time looking at data and half the time brainstorming new ways to solve client problems using AI. Once you begin using AI, the light is glorious and the insights awe-inspiring, even transformative, for the client brands involved.
