Making better decisions regarding labor, inventory and production cannot rely solely on better forecasts. It is also critical to understand the costs of decisions being too high or too low versus actual demand in order to calculate a series of ‘best bets’ that will collectively yield significant profit gains over time.’s technology combines a state-of-the-art forecasting module with a sophisticated economic analysis engine in order to weigh the most important trade-offs and produce superior decisions.


While we support multiple application areas, our underlying methodology is consistent across our Labor, Inventory and Production tools, which is summarized below:


Modern AI techniques give the potential to improve forecasting. This is particularly relevant for businesses with hundreds or thousands of sites, where new techniques allow us to create forecasts that learn from other locations or skus. This can boost performance by identifying patterns that are shared — for example — by all suburban locations or all locations with a particular competitor present. These types of refinements improve performance, as does the incorporation of novel data sources such as local events.


Making better decisions is not just a function of improving forecasts — after all the objective is to improve profits. also applies sophisticated AI techniques to identify the actual cost and revenue impact of having too much or too little capacity vs the actual demand. This is typically a complicated relationship, where the cost of being above and below is neither symmetrical, nor simply linear—- with small errors often having little impact, but large errors becoming very expensive.



An optimal decision on staffing, inventory or production may not be the one associated with the most likely forecast demand level. We also must consider the sizes of the error costs and the likelihood of different demand levels.


For example, the ‘best bet’ decision may be to hedge to avoid an expensive error. automates this and replicates it over thousands and thousands of individual decisions, to improve the overall outcome.


Predion can also automatically explore the implications of different decision levels and learn from the results of those explorations to improve its knowledge of key tradeoffs.

Our Product Philosophy

We’re a team of experienced executives, not just talented engineers and data scientists. We’ve built our product to reflect hard-learned lessons from making system investments pay off in real-world implementations, not just look good in the lab.



AI solutions often gloss over the amount of effort needed to prepare data. We minimize internal IT effort by managing all data transformation and supporting flexible, simple ingestion and transfer methods.


Change management is often a cause of AI deployment failure. We are laser focused on minimally disturbing existing processes, keeping change management efforts highly focused.


In practice no machine is perfect, and people will know important things the machine cannot. We build our tools to make it simple to incorporate this specialized knowledge, while automating the rest.


Nothing goes exactly as you expect, so we structure our engagement model to pilot quickly, measuring financial impact and then designing the system to learn more and adapt further in quick cycles.

1920 L Street NW, Suite 800
Washington, D.C. 20036


A company