From install to a first ride you can trust
- Short on time? Read the two-minute summary.
- A self-initiated concept for a new rider's first session on PickMe in Sri Lanka, covering install to booking (Part A) and booking to receipt and help (Part B).
- Desk research, a funnel map with no invented rates, hypotheses, a coded trilingual prototype and an A/B test plan.
- Unofficial. Not affiliated with, endorsed by or sponsored by PickMe (Digital Mobility Solutions Lanka).
The primary bet (H1, Part B) is that a rider who can hold the fare will ride again. The proposal: the booking screen shows a fixed fare with its breakdown and the only three things that could change it, and the receipt explains any difference before payment. Reviewer reports (not verified in the app) point to a trust gap between the fare shown and the fare charged, and this is where the work is focused. A rider-side trip end and a one-tap dispute route are separate follow-on tests.
The supporting bet (H5, Part A) is that a new rider who sees their own fare before creating an account will reach a confirmed booking more often. The proposal moves the sign-up step to the moment of booking. Permission timing is split off as its own later test. This bet rests on benchmark evidence, not on PickMe complaints, so it is ranked lower and will be tested separately from H1.
Open the live prototype to tap through every screen. Each image below opens it.
One screen, one promise. The ride screen in the generic Control pattern and in the fixed-fare Treatment.
From install to a receipt that matches. Pick a route, see a fixed fare, read the breakdown, get matched, and get a receipt that explains any change.
Built to pass AA. Colour tokens with their contrast ratios, the Manrope type scale, and the Sinhala and Tamil screens.
https://susithd.github.io/pickme-first-session/
The prototype is a phone-sized frame with a demo panel beside it (a bottom sheet on mobile, opened from a "Demo" button).
- Variant switches between Control (a generic ride-hailing pattern, not PickMe's interface) and Treatment. The chip in the frame header always shows which one is active.
- Scenario switches the ride between normal, no driver, driver cancels, route change and driver asks for cash. Peak pricing, language (EN, සිං, தமிழ்) and a reduced-motion override are in the same panel.
Note
This is desk research only. About 30 public reviews and posts were collected (Trustpilot, the App Store, press and social posts), several seen only as search snippets, and six ride-hailing apps were benchmarked: Uber, Grab, Gojek, Careem, Bolt and Rapido. Google Play reviews were not read at all, and Trustpilot skews negative. There was no teardown of PickMe on a clean device, no interviews with riders, and no PickMe data. The funnel contains no invented rates, and sample sizes are shown as tables over assumed baselines. Claims about PickMe's behaviour are reviewer reports, not verified in the app, and the Control variant is a generic pattern. The Sinhala and Tamil copy is a machine draft that needs native-speaker review. Primary research is the first step in docs/next-steps.md.
pickme-first-session/
├── README.md
├── LICENSE MIT (code)
├── PLAN.md decisions, hypotheses, build plan
├── research/
│ ├── 00-teardown-guide.md how to capture a clean-device teardown (not yet done)
│ ├── 01-context.md company and market fact sheet
│ ├── 02-review-mining.md public review themes and sources
│ ├── 03-benchmark.md six-app benchmark with evidence keys
│ └── 04-interview-guide.md five first-use sessions (not yet run)
├── docs/
│ ├── summary.md two-minute summary
│ ├── funnel-audit.md funnel map and evidence by stage
│ ├── hypotheses.md H1 to H5, evidence and what would change my mind
│ ├── decisions.md D1 to D17 (D12 removed), dated, with overrides
│ ├── prototype-spec.md screens, routes, states, copy, tokens, variants
│ ├── experiment-plan.md H1 test and H5 spec, sample sizes, guardrails, events
│ ├── next-steps.md primary research first, then follow-on tests, open local questions
│ └── images/ README banners (1600 px)
├── presentation/ Behance boards and LinkedIn banners, rendered from the live prototype
├── prototype/ React 18, Vite, TypeScript, vanilla CSS, tests
│ ├── src/
│ ├── tests/
│ └── package.json
└── .github/workflows/
└── pages.yml lint, unit tests, e2e, build, deploy to Pages
Docs are CC BY 4.0. Code is MIT.
The prototype lives in prototype/. Node 22 is used in CI.
cd prototype
npm ci
npm run dev # local dev server
npm test # unit tests (Vitest)
npm run e2e # keyboard walk and accessibility checks (Playwright, axe)- Research: desk research and benchmark done. Teardown, five interviews, a usability test of the prototype, a Google Play review read and native-speaker review of Sinhala and Tamil are still to do.
- Prototype: built as 13 screens (ten core screens plus three Part A screens), in English, Sinhala and Tamil, with a Control and Treatment variant. The build is tracked in
PLAN.md, section 8. - Experiment: H1 test and H5 spec are written in
docs/experiment-plan.md. Nothing has been run. - Next: primary research, then the H2 (rider-side trip end), H4 (dispute route) and H3 (payment agreed before pickup) tests. See docs/next-steps.md.
- Code: MIT. See LICENSE.
- Documentation (README,
docs/,research/): Creative Commons Attribution 4.0 (CC BY 4.0). Third-party quotes remain their authors' and platforms'.
Susith Deshan Alwis. https://susith.theqexle.site