The Benefits of an Analytic Product - Democratizing AI Insights

As a company that specializes in creating artificial intelligence-driven insights to support the marketing and sales operations of life sciences companies, we appreciate the complexity involved in designing, building, and executing models. These algorithms are high in value, and used to help make important decisions that impact both company bottom lines and people’s lives. But how do you ensure the insights reach those who can benefit the most? The answer is to create an interactive wrapper over the AI or ML models that empowers users to run their own experiments.

Let’s be clear: not every data science effort needs a customized front-end interface. It may be appropriate to allow your data science team to work solely within a Jupyter notebook or maintain a project within an AI platform (like Dataiku or DataRobot). As far as AI development goes, these are perfectly acceptable technologies for writing, training, and publishing models. This is also fine if you have siloed data scientists who are focused on solving discrete business problems.

But what do you do when you have a specific workflow that you want your user community to run, and you want it to be easy, understandable, and responsive? You also do not feel that it is necessary to provide a commercial software license to each and every member of this analytic user community. An analytic product designed around user behavior and extracting business value pushes control of running an analytic out to a wider community. At the same time, it facilitates the generation of experiments and insights because the best parts of the model can be accessed by users in a controlled, managed fashion. 

For example, in our Next Best Action (NBA) analytic product, which anticipates promotion ordering and timing, we want users to specify the model version, date range, and tactics that they want to evaluate. Once chosen, the users are free to generate results from a pre-trained model and evaluate the output. With a business-driven analytic product like NBA, there is room to do something further with the results. Maybe the output is used to populate a dashboard, or perhaps the application integrates with the company CRM system for personalized suggestions. Whatever the ultimate purpose, the application has built-in guardrails for preventing (or at least minimizing) incorrect decisions.  

A well-designed analytic product also empowers non-data scientists to have a say in how the model is run without them having to contemplate all of the decisions that the data scientists and operations people have to make when the model is developed and trained. Underneath the covers, there must be a supporting Data Ops effort. There must also be a set of machine learning operations (ML Ops) to manage the models, and there should be APIs for programmatic ingestion and delivery. And of course there is the self service administration and utilization component. See below:

• DataOps: automates the data life cycle including the quality assessment process 

• Analytic Data Layer: hosts the different stages of the data life cycle from source files to fit for purpose tables ready for ML consumption 

• MLOps: supports the development, deployment and maintenance of ML models

• APIs: facilitate the interoperability of the model with other applications to ingest inputs and provide results 

• Self-Service: Provide stakeholders with access to all refined and unrefined elements of the product…this makes the models actionable.  

The alternative to having an analytic product (like NBA) is that the models and corresponding insights are accessible to only one or a small handful of data scientists. Many organizations will yield to a standby refrain that “we would never let business analysts adjust the model inputs” or “the people who would consume the models do not know what they are doing.” So…what if there was an easy-to-use analytic product that offered accurate insights and reliable predictions that more people could use? 

Data scientists are not cheap. They also sometimes take vacations. Relying on them to run commodity jobs that can otherwise be handled and managed by a wider population is avoidable. An analytic product has the benefit of being always available and always ready to generate the insights required to successfully steer the ship.   

----

Trusted by life science companies, Sentier provides a data-driven understanding of their customers. For the past 5 years, we have been delivering disruptive insights through innovative AI-powered solutions. Always at the forefront of the most effective data integration and machine learning approaches, we have created unrivaled models. As a result, our clients are better able to optimize their marketing and sales resources, and build stronger customer relationships.

Sentier is committed to the concept of High-Velocity Decision Making and we believe it can only become a reality with a strong and innovative analytics component.

We strive to be the leaders in the actionable application of new and emerging data and analytics approaches and to remove the barriers to our clients of benefitting from these solutions.

We believe that the ethical application of our services will benefit patients and doctors as well as the pharma and biotech companies we serve.