We’ve improved analysis stability and in-app notifications in FactDecode

---

Before starting an analysis, FactDecode checks the current state of the datasets used by the project.

Previously, when a dataset required an update and an analysis condition also required confirmation, the review dialog could sometimes appear repeatedly, preventing the analysis from starting.

This flow has now been revised.

If a dataset is still behind the latest available date after a refresh, FactDecode continues to show that state as-is. However, when a valid saved dataset version is available for analysis, the analysis can now proceed after the necessary review.

In our verification, analyses containing datasets that still showed a delay were able to complete successfully and reach Results without becoming stuck in a repeated review cycle.

If no usable dataset version is available, or if the analysis cannot be executed safely, FactDecode will continue to block the run as before.

Rather than treating delayed data as if it were fully up to date, FactDecode now makes a clearer distinction between the current freshness state of the data and whether there is a safe dataset version available for analysis.

We also reorganized the notification system used for data refreshes, factor creation, and other operations across FactDecode.

Previously, the way messages were displayed could vary depending on the page or operation. These operation results are now handled through a more consistent notification system.

For example, notifications may appear when:

• a data refresh has completed
• a factor has been created
• there is an informational notice related to an operation

Standard success and informational messages remain visible long enough to be read and then disappear automatically.

On the other hand, issues that require the user to correct something immediately, such as form validation errors, continue to appear directly within the relevant form or dialog.

Rather than moving every message into one place, we now separate messages that should be handled where the user is working from messages that simply report the result of an operation.

We also strengthened the submission flow when creating factors.

Once the final factor creation request has started, FactDecode prevents the same request from being submitted again accidentally.

During the validation and review stages, users can still make changes as before. The submission lock is applied only when the actual factor creation request begins.

This helps prevent unintended duplicate processing caused by repeated clicks while keeping the normal workflow unchanged.

This update focuses primarily on stability before an analysis starts and clearer feedback during user operations.

The underlying analysis logic used for features such as Factor Analysis, SHAP, correlation analysis, and signal generation has not been changed as part of this update.

The goal is to make the existing analysis capabilities more reliable and easier to use.

FactDecode is designed to help users move from data and hypotheses to analysis results with a clear and reliable workflow.

In this update, we made improvements to:

• analysis stability when working with delayed datasets
• repeated pre-analysis review behavior
• notification consistency for data refreshes and factor creation
• notification timing and visibility
• duplicate submission prevention during factor creation

We will continue improving not only the analysis capabilities themselves, but also the overall experience of moving from setup to execution and Results with confidence.