This section explains where to find the scenario’s results and how to get the results from outside of the optimization scenario.

Understanding Results inside Model

While the visualizations allow us to have a first glance at the criteria satisfaction, we can dig a little deeper to have a better understanding of where and why some constraints stay unsatisfied.

We can check the content of the tables directly from the model via the Tables button in the top right corner:

Open

Understanding the results will mainly consist of analyzing the different model tables and cross-referencing their content to explain the occurrence of certain behavior.

By looking at the “Results_” tables from the OE run, we can select the table related to a specific criterion and sort by “dissatisfaction”. Any value greater than the corresponding precision value from the problem description means that the criterium is not fully satisfied. The column “state” details the value by distinguishing between Satisfied, Acceptable, and Unacceptable states.

The important distinction between Acceptable and Unacceptable is that, while both have a dissatisfaction greater than 0, the acceptable one is still in the specified margin of error that we (in agreement with the end-user) consider valid.

Therefore, the main problematic lines are the Unacceptable ones. When such a line is found, the column critical gives information on which agent is involved and what it means (e.g., price on a specific product or margin rate). Chances are that the element found here is also present in other unacceptable lines for other criteria, which often means that two objectives pull this value in two different directions. Sometimes, the link is not a direct one as the unacceptable criterion may involve two different computed notions, but each of them applied on the same element (e.g., a common product).

Glassbox Dashboard

A more friendly way to understand the results is to use a glassbox dashboard. It is implemented in the optimization accelerators based on an Optimization Engine run, and documented in Glassbox Dashboards.

If there is no glassbox dashboard for your project, it is possible to implement it easily using the Optimization Common Libraries (accessible to Pricefx employees only).

Dealing with Poor Results

For cases where the system cannot reach its objectives, Optimization Engine Troubleshooting details the most common explanations and solutions to apply.

Querying Results outside Optimization Engine

See https://pricefx.atlassian.net/wiki/spaces/KB/pages/3851026646.