User Experience (UX) Metrics

These metrics capture the performance of the Tiliter Vision system from the user's perspective. They quantify how well the system performs in real-world use and what the user experiences.

Automatic

This metric captures cases where the Tiliter API confidently identifies and returns a single item, automatically adding it to the cart (if this feature is enabled). In this scenario, no user interaction is required. This represents the ideal behaviour, where items are identified seamlessly, similar to barcode scanning.

Selection Required

This metric captures cases where the Tiliter API returns multiple possible item predictions and the user selection is within this list. This metric helps assess how often user input is needed to confirm an item when multiple options are presented. While not as seamless. as an Automatic result, still a great user experience.

Menu Required

This metric tracks instances where the system did not return the item selected by the user, which typically requires a manual lookup. This occurs when the Tiliter API either:

• Fails to provide accurate predictions, requiring the user to manually search for the item using the menu.
• Provides an accurate prediction, but the user intentionally overrides it, resulting in a "loss" event.

Rank Metrics

Rank metrics quantify how early an item appears in the list of predicted results generated by the Tiliter Vision system. These metrics help track how often the correct item is displayed within the first few options presented to the user. The metric is calculated based on the selection result. The ranks are categorised as follows:

Rank 1-4

This metric tracks instances where the correct item appears within the top four predictions. Even if multiple items are displayed, the system is considered accurate if the correct item ranks highly enough for easy selection. For example, Rank 1 can apply even when three items are shown, as long as the correct item is the first option.

Rank 4+

Rank 4+ covers cases where the correct item does not appear within the top four predictions, requiring the user to either scroll further or manually search for the item. This is typically linked to the "Menu Required" user experience metric, where users rely on the traditional menu search to find the correct item.