Decisions

Firing a real-time bank transfer on every scan inherited bank uptime, latency, fees, and redirects. Validation proved that bet couldn't hold mid-build.

Decoupled funding from checkout. Banks are touched only when a user tops up their wallet; payments clear instantly against wallet balance.

Dynamic QR infrastructure is expensive to build and operate, and an MVP doesn't need it to prove the loop.

Ship a static merchant QR where the payer enters the amount. Print it once, accept money forever.

A QR-only model breaks peer-to-peer payments, where there is nothing physical to scan.

Every user gets a name@shappay virtual address. QR codes and typed addresses resolve through the same engine.

Every export tool optimizes for completeness, producing noisy output that burns AI tokens and confuses the model.

Built the opposite bet: a semantic representation defined by what it confidently throws away.

A clean export only proves itself against messy, real files made by people who never expected a machine to read them.

Shipped to the Figma Community so the output had to survive contact with files it was never designed around.

Any download or sign-in step kills conversion at the table, where a hungry customer has zero patience.

The QR opens a web app deep-linked to the table. Scan, browse, order, and pay by transfer, with nothing to install.

People arrive in groups, read the menu at the same time, and hand the waiter one combined order. A separate cart per phone breaks that ritual.

A shared table session: everyone at the table adds to one synced order and it is placed together, mirroring how groups actually order.

Staff cannot juggle several different tools per shift while the floor is busy.

A single role-gated admin app spans kitchen, floor, and management, showing each person only what their role needs.

In Nigeria's big cities, operating cost (rent, power, overhead) runs higher than India's while shopper spending power runs lower. Mass-market baskets are too small to support 15-minute delivery economics.

Target the tier-1, convenience-paying customer who already values twenty minutes over a few hundred naira. The choice of who to serve fell straight out of the economics, and quietly decided everything downstream.

The moment you teach this customer to hunt for discounts, you start a fight the margins can't win.

Lead with speed and proximity everywhere; the home screen's top line shows the delivery time and store distance, not a sale banner. The customer is paying for speed, so the product talks about speed.

For the youngest, most frequent shoppers most orders are unplanned. They open the app because they remembered they're out of something, not with a list, so an empty search box is the wrong tool.

The home screen shows aisles, not products: each aisle a 2x2 peek of what's inside. You read the whole store in a glance and tap straight to it, while the catalog gently tempts the impulse buyer.

The easy way to lift average order value is to discourage small orders, but small frequent orders are exactly the habit that keeps a customer coming back. Push too hard and you win the order while losing the customer.

Design toward what a customer is worth over a month, order size and frequency together. The cart nudges baskets upward gently instead of bullying them.

Asking who someone is should never stand between them and the thing they came to buy. A sign-up wall at the door kills speed.

Guests shop straight past the sign-up screen. Identity is collected only at the one moment it matters, placing the order, and only the minimum a delivery needs (phone, name, address), as a sheet over checkout so no one loses their place.

The fifteen minutes between paying and opening the door is where the promise either feels kept or doesn't. It is not dead time to hide behind a spinner.

Show the order moving through real stages (accepted, packed, on the way) with a live countdown on screen, making the brand's whole claim visible and accountable.

Period-care branding defaults to either clinical or overly cute, and neither earns trust.

A single leaf-droplet mark built into the wordmark around Cheerful Violet, carrying the system without leaning either way.

Claiming to be a product designer and engineer is cheap; anyone can write it on a page.

Hand-coded this site end to end (Next.js, Tailwind, GSAP) and labeled the stack, so the container is itself the evidence.