Perform Causal Discovery / Causal AI methods to identify causal relationships in your complex data with the Xplain CausalDiscoverer (XCD).

What it can do for you

The Xplain CausalDiscoverer (XCD) uncovers causalities in your complex data by excluding all available confounding factors. It, therefore, reveals the genuine relationship between two events (cause and effect), where the second event is a consequence of the first.

By understanding causality, companies can gain deeper insights into their operations, customers, and market trends to plan effective actions and make targeted decisions that drive growth and success.

The CausalDiscoverer leverages the capabilities of the back end to implement a usability concept for the analysis of objects “as a whole”, i. e. across all available data streams which are linked to the root object. Classical DWH approaches replicate parts of data into “star schemas” or “OLAP cubes” – by comparison to ObjectAnalytics, they are like the view through a keyhole.

You will no longer require experts to coerce data into these constrained analytical schemas, and with that, you bring analytics from the ivory tower into your daily business. You will be able to interactively follow your train of thought, from questions to follow-up questions, and – supported by predictive models – discover potential “cause and effect” relationships.

How the integration into your development environment looks like

The CausalDiscoverer supports various interfaces. You can generate artifacts in the XCD (e. g., a relative time axis), view the definition of that artifact, e. g. in terms of a piece of Python code, and simply copy that code into your Python script. Indeed, you will be able to convert an entire session into a Python script (or into a Jupiter Notebook).

The Xplain CausalDiscoverer and your Python environment seamlessly interoperate – making the XCD your extended development workbench.

Analysis of the production process

Example from manufacturing – cylinder heads production: analysis of the production process and root cause analysis in the case of faulty parts.

Why causality is important

Predictive modeling tells you, for example, which machine is about to fail.

But the more important questions are: Why is it failing? And what can I do about it?

This is where Predictive Modeling ends – and Causal Discovery begins. Only when you know why something happened (its cause) can you plan targeted interventions to eliminate causes of failure or achieve a desired effect. This will enable you to prevent or change future outcomes.

Knowing cause and effect is the foundation for developing smart strategies to achieve a goal. Whatever your business objective, you better know what is critical to your success – or failure!

The challenge – and our solution

In Real World Data there are literally “billions” of correlations to your target variable (your business goal). However, 99.9 % of them are meaningless. “Correlation does not equal Causation” – a well-known fact. But why is it so?  And how can we penetrate that fog of correlations to see the real causes driving your business?

Confounders (factors that affect both the supposed cause and the supposed effect) mask the “real” impact of a factor on the outcome. Understanding cause and effect based on observational data primarily means finding these confounding factors. Thus, a wealth of information is required to avoid overlooking important confounding factors – and technology that can search such necessarily complex data for confounding factors.

Based on our ObjectAnalytics Database, our Causal Discovery approach is currently the only approach that can handle the complexity of “Real World Data” – and at the same time use the wealth of information in such data for a deep search for confounders. Without an experiment, this is still not proof of causality. However, Causal Discovery can sort out myriads of meaningless correlations, helping you quickly arrive at relevant hypotheses about causal effects.

Listen to Paula to understand our Causal Discovery approach on an intuitive level.

What Causal Discovery can be deployed for

Typical applications are:

  • Manufacturing: Understand complex production processes and why manufactured parts are not meeting quality targets. Implement continuous process monitoring: a DiscoveryBot that constantly searches for newly occurring  causes of defects.
  • Maintain an installed base of machines: Understand why machines fail and develop an intelligent maintenance strategy. Let artificial intelligence look for new problems and their causes and predict failures to intervene early.
  • Healthcare: Understanding the causes of noncompliance and treatment changes. Recognize early signals of side effects. Understand who is at risk and why – and establish targeted care management.
  • CRM: Comprehend the causes of churn. Constantly look for new causes and how competitors are attacking your customer base. Measure the causal effectiveness of your campaigns.

… and many more.

What our Causal Discovery algorithms look like in practice

The Causal Discovery / Causal AI algorithms are embedded into the CausalDiscoverer and its interactive usability concept. Once configured, just click on a target, and you will see what drives or potentially causes this target. The domain expert can reject certain factors, and ask for alternatives. In this way, you can arrive at results, which combine knowledge from data with domain expertise.

Any of these algorithms can also be used from within a Python environment.

DiscoveryBot: It constantly monitors your processes for emerging issues and reasons that jeopardize your business. Your insurance against unpleasant surprises.

Cancer Chart

Example: Factors potentially causing breast cancer, including a graphical representation that visualizes direct and indirect effects on the target of analysis.