Sutra
A forgiving mobile checkout, on a design system two stores share
The problem
A checkout is the one screen where friction converts directly into abandonment — and the usual failure modes are all self-inflicted: surprise fees shown after commitment, errors yelled mid-keystroke, a dead end for every mistake. Sutra’s bet: if every step is forgiving, reversible and priced up front, the flow survives contact with real people.
Sutra is a three-step mobile checkout — contact, delivery, payment — built for the Taar store, on top of a small design system extracted from two real products (Taar and Ghar). It is also a design-system study: the components on the checkout screen come from a token-driven library documented in a living System page, so the checkout and the system ship and age together.
Flow
- One page, three steps — contact, delivery, payment.
- Completed steps collapse into a summary with an Edit button.
- The PIN code fills city and state; the fields stay editable.
- Fees sit on the choice that causes them — delivery on the option, the COD fee on its option.
- UPI first, validated inline; card is visibly disabled and dated until v0.2.
- The live total is one tap away in the order summary at every step.
Every form rule in the flow is built to be undone, not punished.
Wireframes
Try it
Fill the form with a mistake, then open the System tab and switch the accent.
Enter PIN 411001 — city and state fill in, and stay editable.
Design decisions
- Validation on blur, not on every keystroke
- You’re never corrected mid-word.
- Errors say what to fix
- “A UPI ID looks like name@bank” — not what went wrong.
- A count on Continue, not a wall of red
- Press Continue with problems and you get a count, not a wall of red.
- Completed steps collapse into a summary with Edit
- Going back never loses what you entered.
- The PIN lookup assists, never overrides
- City and state fill in, and the fields stay editable — autofill assists, it never overrides.
- Fees before commitment
- The delivery fee sits on the delivery option, the cash-on-delivery fee sits on the COD option, and the live total is one tap away in the order summary at every step. UPI is the default path with the ID validated inline — card is visibly disabled and dated, rather than hidden.
- Trade-off — Card stays disabled until v0.2 — an honest gap on the screen instead of a hidden one.
The design system
The design system underneath has four rules, and the prototype enforces all of them:
- 01Tokens first
- No raw hex or px in product code, only var(--…) references.
- 02Only --accent is themed
- One attribute swap re-skins the whole app from Taar navy to Ghar terracotta, because the neutrals had drifted by a shade or two between the two products and are now shared on purpose.
- 03Components are small and boring
- Each does one job and exposes every state — default, focus, error, disabled.
- 04A living System page
- It documents the tokens and every component state, with a changelog discipline for additions.
- Paper #F6F3EESurface
- Ink #1B1D22Text
- Mute #6B685FSecondary text
- Line #DDD7CCDividers
- Taar navy #1F2A44--accent · product A
- Ghar terracotta #A63D1F--accent · product B
- Danger #B42318Errors
Accessibility lives in the component, not the screen — which is what makes it hold up under reuse: every field carries a real label, aria-invalid on error and aria-describedby to its message; errors are announced with role="alert"; 44px minimum tap targets everywhere, including the radio choices and the chips; visible focus rings on inputs, choices and buttons.
Build & next
Honest limits, stated up front: the PIN table is a mock — five cities, not a postal API. No real payment; the card option is disabled until v0.2. Not yet tested with real users, so no completion or abandonment numbers are claimed.
- A Figma library with variables for the tokens and components with variants.
- The card form — number, expiry, CVV — as the third payment method.
- A usability round with five people, focused on the error-recovery paths.