Roughly twenty thousand Iranian-owned businesses operate in Canada with no trusted Persian-language directory between them — and building one meant deciding, in code, what money is allowed to buy.
Interface
What the user touches
Application
Where the rules run
Data
What persists
Platform
What it runs on
There are roughly twenty thousand Iranian-owned businesses in Canada — lawyers, doctors, restaurants, advisers, shops, trades — and no single Persian-language reference that covers them. Discovery happens through Instagram posts, Telegram channels, and word of mouth, which works until you need someone outside your own network.
The project started as čārana and was rebranded to GOPLAZA in August 2026. The rebrand was a naming decision; the product question never moved. Listing businesses is the easy half of a directory. The hard half is keeping the data true and keeping the trust signals honest once there is a paid tier.
That second problem shaped the architecture more than anything else. If a business can pay for visibility, the platform has to be able to state — structurally, not in a policy page — that visibility is all it bought.
In a directory, the moment a paid placement can look like a trust signal, every unpaid signal on the page stops meaning anything.
A Turborepo monorepo running a Next.js 16 web app and an Expo mobile app against Supabase, with a shared core package that holds every rule both clients must answer identically.
Directory across four axes
Businesses browsable by category, by city, by city-and-category, and through community channels — the ways people actually look for a local service.
Claim and correction flows
Owners claim their listing through a reviewed process, and anyone can submit corrections or complaints, with an admin queue behind each.
Contribution standing
A «اعتبار مشارکت» ladder that recognises people who keep listings accurate, with levels computed at read time rather than stored.
Owner loyalty, deliberately separated
Tenure earns discounts and room, never a credibility badge. "Three years on Plaza" is publishable; "trusted business" is not.
Mobile app on the same core
An Expo client sharing the entitlement, standing, and schema logic with the web app, so both answer identical questions identically.
Persian as the primary locale
Persian digits, Iranian calendar handling, and Vazirmatn typography built into the core rather than patched at the view layer.
Operator tooling
Nineteen admin surfaces covering claims, reports, reviews, corrections, listings, jobs, health checks, and data cleanup.
A directory is only as valuable as its data is current, and business data decays constantly.
Once listings can be paid for, every visible signal becomes suspect unless the platform separates money from credibility by construction.
Two clients reading the same database will drift apart the moment either one implements a rule locally.
Stored derived state goes stale silently — an expired plan still reads as paid until something remembers to downgrade it.
Persian-first is not an English product with translated strings: digits, dates, and typography are all different systems.
Two principles ended up driving most of the technical decisions, and both came from failures we could see coming rather than ones we waited for.
Paid features buy placement and capacity, never credibility. The loyalty module is written so that nothing it exports can be used as a trust signal, because the moment one paid signal looks earned, no unpaid signal is believable.
Standing level, plan entitlements, and owner tenure are all recomputed on read. A lapsed contributor loses their level the moment the window passes — no cron, no decay job, no stored verdict to go stale.
A parity audit found the web app ordering listings randomly with a paid boost while mobile ordered by creation date: the same directory answering the same question two ways. The shared core package exists to make that class of bug impossible.
Expiry is judged against the paid-through date rather than a status column, so a delayed Stripe callback neither over-charges the customer nor keeps a lapsed listing featured.
Entitlements, standing, schema, and formatting live in packages/core so web and mobile cannot answer the same question differently.
Anything derived from time — levels, expiry, tenure — is calculated at read time, which removes an entire category of stale-data bugs.
The paid tier is bounded to placement and capacity. Trust comes from claims, corrections, and contribution standing, none of which are for sale.
Phase 1
Model businesses, categories, cities, and channels on Supabase, and ship the browsing surfaces across each axis.
Phase 2
Build the claim, correction, complaint, review, and report flows with the admin queues that adjudicate them.
Phase 3
Extract the shared rules into packages/core, ship the Expo client against it, and add entitlements, standing, and loyalty with the money-trust separation enforced in code.
GOPLAZA is one of the ventures we design, build, and run in-house at AshaVid.
