Modeling the Mechanism

Across science businesses we see the same challenges. Everyone wants winners faster and cheaper than they can test.  If we could just perfect the process of "science," we could avoid getting stuck in the trap of waiting on unnecessary tests and phases.

Others have tried to solve this problem with "black box" AI/Machine Learning/statistical models which rely solely on passively observed data – they may even flaunt that they are "purely" data-driven and "uncontaminated" by the assumptions of theory or structure, but they only predict associations. Black Friday yields sales. Rain yields grain.

With recent advancements in the science of causality, we now know that we can do better. Our team of PhDs, computational scientists, MBAs, and applied mathematicians built a different modeling approach to predict outcomes under intervention. They tell you how to achieve goals you've never met by doing things you've never done. They work because they know why.

Our technology solution pairs a "catalog of causal mechanisms" mined from the academic literature – whether it be consumer behavior or fundamental chemistry – with Bayesian Inverse learning. Bayesian Inverse uses data to quantify the causal effects of those mechanisms even when they are entangled with each other and the environment.

Build Data Richness

You struggle with enterprise systems because they cannot capture the richness of complex business and scientific processes. To get work done, subject matter experts have to preserve data crucial to understanding the product as tribal knowledge and ad-hoc pockets of data–outside of any corporate ERP, CRM, PLM, LIMS or ELN. Our solution enables quick integration from everywhere, yes even for that crucial data and metadata - getting past the roadblocks of ERP, etc.

Reveal Hidden Mechanisms

Your Design of Experiments is often focused on getting a specific outcome: "How do I improve this battery design's capacity?" or "How do I get a juicier tomato?" However, you run a disconnected series of experiments that don't reveal the "why". We disentangle, quantify, and structure unknown material properties (activity of an enzyme, diffusion through a crystal lattice, etc.) using our Bayesian Inverse technology and your existing testing data.

Capture Environmental Factors

Concurrent mechanisms aren't just entangled with each other - they're highly entangled with the environment. Research can't live in a vacuum, but scientific reductionism dissects mechanisms by excluding the richness of the environment. How do you know if your gene increased corn yields or was it just a wet year? Our environmental simulators recreate the experimental conditions of your physical tests and the complex real-world conditions of your products.


Price effect on sales

We capture how likely consumers are to purchase at each retail channel as price varies

Susceptibility to Ads

Total impact of advertising

We capture how likely consumers are to purchase at each retail channel as media exposure varies

Impact of Seasonality

Varying Purchase Likelihood

We model true seasonality of a product, excluding all other input factors.

Actionable Forecast

Causal Model with All Effects

We make this model totally accessible to our users, allowing them to explore the results and explore the results and build plans

Technology Platform Components

Our software encodes the exceptionally sophisticated mathematics and computation required to implement this strategy in any domain

"Super Suit" Architecture

Our platform's data architecture supports new data primitives for expressing the rich (nested, conditional and multimedia) structure of science-business data and a compact toolbox of new action primitives (like tree-aggregate and tree-compare) for accomplishing any complex workflow without app customization.

Universal Model Language

Our platform uses a form of Bayesian probabilistic programming as a universal language to describe any mechanistic simulation model. This allows us to weave mechanisms gathered from disparate academic literature sources into a singular whole.

Universal Parameter Learner

Our platform's Bayesian Inverse engine can learn–and learn the limits of its own knowledge–to parameterize the mechanisms of any mechanistic simulation model from your existing testing data without software or mathematical customizations.

Universal Business Optimizer

Our platform uses compiler technology and advanced mathematical methods to build environmental and business simulators that allow our software applications to translate mechanisms into optimized business decisions.