We conducted a heuristic evaluation and identified several key issues blocking a smooth user experience. Here is a breakdown of the critical friction points.
- Data Privacy: Critical breach found on the History page. Users can currently see and modify other users' projects.
- Expectation Mismatch: The workflow is backward; users generate a concept before setting constraints, leading to frustration when the filter screen appears later.
- Dead Ends: Users cannot complete actions (submit/save) on final screens due to missing buttons.
- Hidden Scroll: The initial questionnaire lacks a visible scrollbar, hiding content below the fold.
- Blind AI Generation: The AI process only shows a percentage. Users need descriptive updates (e.g., "Refining shapes...") to trust the system is working.
- Lost Progress: Selecting an alternative design wipes the original instead of moving it to a comparison sidebar.
- Confusing Labels: Terms like "Feedback" and "Recalculate" in the editor don't match how jewelers actually talk about design changes.
- Ambiguous Navigation: The "End Project" button implies finishing production, but it just redirects Home.
- Click Targets: Selecting a design requires clicking a tiny button rather than the image itself.
- Trap Doors:
- "Confirm Order" triggers an immediate purchase without a summary or confirmation dialog.
- No "Undo" exists after confirming a design choice.
- AI generation cannot be stopped once started.
- Popup Traps: Critical price info popups lack a close button ("X"), trapping the user.
- Forced Flow: After AI generation, users are locked into choosing a concept with no option to start over.
- Inconsistent UI: Download buttons change style across screens; the "Creation" button is in the wrong corner (top-right vs. standard bottom).
- Confusing Icons: The survey confirmation button uses a "Back" arrow icon, but flipped horizontally.
- Unpredictable Actions: Clicking "+" in a deleted slot auto-generates content instead of letting the user choose.
- History Corruption: Users can accidentally modify already submitted/closed projects.
- Open Inputs: Text fields lack validation, accepting nonsensical data.
- Visual Clutter: Drafts and submitted projects look identical in the History page.
- Memory Load:
- Generated designs lack descriptive labels (e.g., "Vintage Style"), forcing users to remember which is which.
- The "Similar jewels" label scrolls away instead of sticking to the header.
- Blind Spots: The jewel is only shown from one angle.
- Slow Iteration: Customizing an inspiration design requires a tedious download-reupload loop.
- No Power Tools: The History page lacks search or sort filters. Expert users have no keyboard shortcuts.
- Intrusive Alerts: Price changes trigger a screen-blocking popup instead of a subtle banner.
- Information Overload: The cost breakdown is a wall of uniform text, making it hard to scan.
- Cramped View: The editor lacks a full-screen view for detailed inspection.
- Silent Errors: When materials clash, the system blocks the action without explaining why.
- Hidden Features: Users must guess that designs can be customized; parameters are hidden inside menus rather than visible.
Chosen Approach: Prototype #1 (Tablet)
We chose the tablet prototype over the desktop version because it offers a more natural flow for jewelry design.
- Context: A tablet is a better collaborative tool for a jeweler to hand to a client than a desktop computer.
- Interaction: The touch-optimized interface supports the visual, tactile nature of customizing a physical object better than mouse clicks.
- Flow: The iteration process on the tablet version was simply clearer.
Feature Migration: To ensure we don't lose functionality, we are moving the "Estimated Crafting Time" feature from Prototype #2 to our selected design. It will appear as a label near the price during iteration to manage user expectations.
- Screen 1: Setup Form
- Screen 2: Design iterator
Goal: Transform our paper concepts into a responsive, interactive prototype that solves the major friction points identified above. We are prioritizing User Control and System Feedback.
- Safety Net for Orders (H3): We will implement a proper Confirmation Dialog with a summary (preview, price, time) before any purchase is final.
- Transparent AI (H1): Replace the static percentage loader with active status updates (e.g., "Analyzing preferences...", "Polishing design...").
- Escapability (H3): Add clear "X" close buttons to all popups and an "Edit Design" button on the confirmation screen so users are never trapped.
- Flow Correction (Non-Heuristic): Move the filter/constraints screen before generation to align user expectations.
- Clearer Labels: Rename ambiguous buttons (e.g., "End Project" → "Exit to Dashboard") and add descriptive tags to generated designs.
- Better Feedback: Provide specific error messages when material combinations fail (e.g., "Gold is too soft for this setting").
- Visual Hierarchy: Redesign the cost breakdown to be scannable (larger totals, grouped sections).
- Undo/Cancel Generation: While useful, stopping the AI mid-process is less critical than fixing the ordering flow.
- Breadcrumbs: Adding navigation context (Home > Editor) will be added if time permits.