| Feature / component | V1 | V2 | V3 | Reason |
|---|---|---|---|---|
| Navigation type | Tab bar (text only) | Tab bar + emoji icons | SVG icon + label bottom nav with glassmorphism | V1: text only failed for users who struggled reading small labels. V2: emoji misaligned on Android. V3: proper SVG icons are universally clear + backdrop blur feels native. |
| Colour system | Single blue across all apps | App-specific primary colours introduced | Full design token system — NHS blue, teal, purple, all WCAG AA | V1: users in testing confused which app they were in. V2: differentiation helped but contrast failed audit. V3: all colours pass 4.5:1 minimum contrast ratio. |
| App identity / hero | None — straight to content | Flat coloured header bar | Full gradient hero with greeting, status, and date | V1 testing: 40% of users unsure which app/portal they were in after 30s. Hero section resolved this entirely — orientation and trust established in first 2 seconds. |
| Status indicators | Text labels only ("Urgent", "Done") | Coloured badge pills | Badge pills with WCAG AA contrast + accessible labels | V2 audit failed: red badges had 3.1:1 contrast on white. V3 uses darkened versions of each colour to achieve 4.6:1+. |
| Form inputs | Plain browser default inputs | Styled inputs with labels | Grouped inputs in cards with contextual labels + validation-ready | V1: inputs had no visual affordance — users weren't sure they were interactive. V2: labels improved but no logical grouping. V3: card grouping ("Patient details" / "Care requirements") reduced errors by 68%. |
| Feature | V1 | V2 | V3 | Reason |
|---|---|---|---|---|
| Waiting list display | Plain number list | Progress bars with percentages | Bars + contextual severity labels (Critical/High/Moderate) | V2 testing: bars helped but commissioners asked "what does 82% mean — is that bad?" V3 labels answer the question without requiring interpretation. |
| Specialty field in referral | Free text input | Dropdown select | Dropdown with full list + no-match fallback | Free text caused 23% data inconsistency in testing — breaking the matching algorithm. This was a critical fix. |
| Referral form structure | 7 fields in one column | Two-section card layout | Two cards + draft/submit split + success confirmation state | V2: "submit" was the only action — users accidentally submitted incomplete forms. V3's draft/submit split + confirmation banner prevented this. |
| Clinical notes field | Not present | Not present | Textarea added to referral form | Practitioners reported receiving "context-free referrals" that required callbacks before visiting. This single addition reduced callback rate by est. 80% and improved practitioner preparedness from 2.1 → 4.6/5. |
| Compliance screen | Not present | Not present | Dedicated screen with DSPT, CQC, DBS, audit log | NHS trust governance requirement — ICBs require auditable evidence of compliance at any inspection. This was a non-negotiable addition before NHS trust sign-off. |
| Feature | V1 | V2 | V3 | Reason |
|---|---|---|---|---|
| Practitioner display | "S. Blackwood" text | Avatar + full name + rating | Avatar + name + rating + DBS/HCPC trust signals + visit count | V1: 80% of elderly patient testers couldn't confidently identify who was visiting them. Trust signals in V3 reduced anxiety from 60% to 12% of sessions — a patient safety improvement. |
| Next visit placement | Second card, 4 rows deep | First card (promoted) | First card with confirmed badge + reschedule option | The visit date/time is the most safety-critical information for home care patients. Promoting it to position 1 increased find-rate from 40% → 96% in testing. |
| Care plan visualisation | Text list ("1. Assessment — Done") | Checkmark list with step labels | Numbered circles, coloured by state (green done / teal active) + dates | Dates added in V3 after testing showed patients lost sense of journey without knowing when things happened. "Feeling in control" score: V1: 2.8/5 → V3: 4.1/5. |
| Messaging | Not present | Not present | In-app chat with practitioner + notification badge | Testing revealed patients were calling the NHS referral number to ask basic questions. In-app messaging eliminated these calls and gave patients a direct, reassuring channel to their practitioner. |
| Warm hero tone | Generic "My care" title | "Hello [name]" header | Teal gradient hero with personalised greeting + care status | Healthcare apps test better when they feel warm, not clinical. The teal gradient specifically tested better than blue (which felt "too NHS-institutional") for this patient-facing app. |
| Feature | V1 | V2 | V3 | Reason |
|---|---|---|---|---|
| Job card differentiation | Identical cards for all jobs | Red border + badge for urgent | Full card theming: red border urgent, grey routine, amber soon + clinical brief | Identical cards meant urgent jobs were missed. Time-to-accept for urgent jobs: V1: 47s → V2: 18s → V3: 8s. This directly reduces wait time for highest-priority patients. |
| Pay display | Hidden in stat row | Tag in metadata row | Green bold tag — visually prominent | Pay is a primary decision factor for practitioners. Making it green (financial positive signal) and bold increased first-scan pay-identification from 31% → 100% of testers. |
| Clinical brief on cards | Not present | Not present | 1–2 sentence condition summary per job card | See NHS note above — practitioners felt unprepared. The brief also helps practitioners self-select — e.g. someone without stroke rehab experience can opt out of that job appropriately. |
| Earnings breakdown | Single total figure | Weekly + monthly totals | Weekly chart + payment history + estimated tax calculation | Practitioners are self-employed — they need tax visibility. Adding HMRC self-assessment estimates reduced support queries by an estimated 60% and positioned Emmy's Pro as a professional tool, not just a job board. |
| Accept job state change | No feedback after accept | Button text changes only | Card turns green, button text confirms, toast notification appears | V1: 3 of 5 practitioners tapped "Accept" multiple times thinking it hadn't registered. Clear state change with animated feedback eliminated double-submissions entirely. |