Defining the Problem – Getting Started with Causaility

By Kristi Furrer, COO Causaility

We’ve been talking to a lot of folks about solving their complex business problems, and I’ve learned that understanding and defining the problem accurately is the lynchpin to success for an analytics project. 

Causaility’s AI analytics platform, Causaition, is key to optimizing your decision-making process and preventing costly mistakes. We help solve your most pressing business problems and explore various scenarios using biomimetics which imitate nature to solve complex problems.

I’ve taken a few minutes to jot down my thoughts and help clarify what’s involved…

When trying to solve a problem, first, we need to understand it, its nature, and its characteristics to know where to turn for help. You don’t always need sophisticated analytics tools like Causaition, but if your problems look like those below, then we can help.

1. The problem is complex – a problem that has many possible solutions, is difficult to analyze, and is not easily understood.
2. The problem is multidimensional and requires data from diverse sources to analyze effectively – multidimensional problems have many different aspects or features such as complexity, multiple factors, and multiple dimensions.
3. The data is multi-scalar – data that encompasses multiple levels of measurement or scales.
4. The problem domain is dynamic and evolving – the issue is constantly changing and developing over time, with new aspects or factors emerging, requiring a flexible approach to address it as the situation unfolds.

You may not fully understand the problem you want to solve yet but this may guide your thinking. We believe we add the most value when we work with these kinds of problems.

Next, we need a clear articulation of the problem to be solved or what you are trying to learn or validate.: To help you understand what I mean, let’s look at what medical labs are experiencing as they try to embrace personalized medicine and migrate to a value-based care model that prioritizes outcomes and patient value. Many believe they can improve outcomes for complex urinary tract infections (UTIs) and bladder cancer by demonstrating the clinical utility or efficacy of polymerase chain reaction (PCR) testing over traditional culture testing to accurately detect these conditions.

This is a comparative analysis of two different testing modalities. These comparisons have a lot of patient admissions, treatment, and outcome data associated with them organized chronologically. This type of comparative analysis is applicable across healthcare for applications such as vaccine adverse reactions and drug treatment outcomes.

To understand the value of one modality versus another, we ask if there are factors about these two testing modalities we want to explore.  Are there factors that affect the data over time for example, this could drive the outcome and the results of that testing modality. If the results are intermittent or erratic and difficult to predict it becomes a multidimensional problem, where we are looking to explore the data that’s available related to a given testing modality and discover.

We let you see the factors that may be at work within certain populations and reveal what those linkages are and what the relative relevance is. This leads to a better understanding of the relevance of the various factors and what impact they’re having, developing knowledge around that.

For the analysis, our team builds a multidimensional model based on you and your team’s input, as the subject matter expert(s) (SME). The SME identifies the questions they want to answer, or what they want to explore and discover, based on the available data, and defines what they want to include. This can be any kind of real-world data (RWD) including the dark data created during the normal course of business but rarely explored.  RWD sources are the multiplicity of available databases ranging from siloed enterprises to public and private records. We collect this information and download it to our AI platform with its data lake implementation that eliminates system integration complexities. We use small/wide data methods to connect diverse and dynamic information from disparate data silos.

We overcome siloed data without the need to cleanse, normalize, or integrate delivering groundbreaking and insightful results, removing data challenges while increasing efficiencies and mitigating bias, saving time and money before moving to a “real-world” environment.

Our team works with you to deploy your model and access the data to populate it based on what questions we’re looking to answer, or what things we’re looking to explore and discover.

In general, domains such as research often have the type of complex problems where our platform excels. Other domains that can benefit include:

• Complex Scenario Analysis – Why and What if analytics is an example.
• Simulation of Natural or Complex Systems – Digital Twins representing Human Biology is an example.
• In-silico project execution – Biopharma Clinical Trials is an example – In-silico refers to experiments performed on a computer rather than humans.

We believe our technological approach represents a paradigm shift, moving from obscuring reality by narrow technological constraints to discovering reality by applying real-world reasoning (RWR) to real-world data (RWD).