States and edge cases

I was the sole designer for two internal MTS catalogs, preparing decisions for three feature teams. Designs reached engineering without some empty lists, errors and intermediate loading states.

Questions arose during implementation. With work distributed across three teams, missing details generated additional clarification requests for me.

I introduced a standard: component states are described and edge cases are explicitly recorded in the file. The team received transition rules alongside the designs.

The handoff included:

• States: default, hover, unavailable, error and empty; loading where the scenario requires it.
• Transitions: which action or system response changes the state.
• Edge cases: what to show when there is no result, an error occurs or an action is unavailable.

This is the structure of the standard described in the project history. The set of states depends on the component: forms, search and lists need different descriptions.

At a working meeting in September 2025, I proposed merging separate designs into the main file, updating components and documenting their intended behavior. I specifically mentioned search and filters: how they work, what should be shown and what should not.

The analyst and I agreed to track this as a separate Story Enabler task. The recording confirms the subject and the agreement; completion of this particular task has not yet been recovered from the archive.

The standard can be explained through a simple question: can an engineer determine the behavior without another verbal explanation?

• What does the person see before an action and while waiting?
• How does an empty result differ from an error?
• When does an action become unavailable?
• What happens on a retry?

This explains the handoff principle; it is not a record of a separate historical test. An interactive reconstruction of five states is available in the combined MTS story.

State and edge-case rules became part of the design handoff. Together with rapid prototyping, this standard reduced engineering rework by 15%.

The contribution of each change was not measured separately. I am still recovering the counting method and comparison period; there is no independent percentage for states or the search-and-filter task.