Building Analysis That Survives an Interruption

A long analysis should belong to the account and the job, not to one open screen. If a device sleeps, the app closes, or connectivity changes, completed work should remain durable and recoverable.

Store progress at meaningful boundaries

Checkpoints are most useful after verifiable stages such as upload completion, media normalization, round segmentation, athlete-track approval, metric extraction, and report assembly. Each checkpoint should record the inputs and analysis settings that produced it.

Adjust without starting over

When a user changes analysis criteria during a long job, compatible stages can be retained while affected measurements are recomputed. Clear dependency records make that selective work possible and keep progress estimates honest.

• Resume from the last verified checkpoint.
• Notify the user when processing finishes or needs review.
• Keep partial results private until their publication state is explicit.
• Never present an interrupted stage as complete.