Cards are not the default
Most orders arrive as cash on delivery or through mobile wallets, which foreign platforms treat as an edge case to be bolted on.
An online store with 0 platform fees a month.
A custom commerce platform built for the Bangladeshi market: taka pricing, cash on delivery, local wallet payments, courier integrations and Meta tracking that reports on profit. The shop owns the code and the data, and platform hosting costs $0 a month.
The build
The platform I built.
What I built
The work, up close.
15 pieces of the build, every one a real page from the finished site.
Storefront
Pages load fast and the shop can restyle them
Server-rendered for speed and search visibility, with a section builder that lets the shop assemble campaign homepages, banners, countdowns and collections without touching code or installing an app.
Rendered at the edge, no origin server
Collections
The shop controls its own merchandising
Manual collections for hand-picked ranges and smart collections driven by title, type, tag, price or stock. Sorting and filters are configurable, so a campaign page can be built in minutes.
Manual and rule-based collections
Discovery
Filters built around how buyers search here
Price, type and product-option filters, live type-ahead search, quick add straight from the grid, and recently viewed items, so the catalog stays navigable as it grows.
Type-ahead search with quick add
Catalog
New product ranges add without a rebuild
Each category gets a proper landing page with its own copy and structure, so a new range slots into the store and into search results without a redesign.
4 collections at launch, built to extend
Product page
Product pages carry every buying detail
Ordered galleries with zoom, variant-specific images, compare-at pricing, live stock messaging, and both Add to Cart and Order Now paths. Selecting a variant updates price, image and availability instantly.
Up to 3 option dimensions per product
Variants
Stock is tracked per variant
Variant-level price, SKU, cost, weight and stock, with low-stock urgency drawn from actual inventory rather than a fake counter, plus back-in-stock signup when something sells out.
Variant-level stock and pricing
Presentation
Handmade products get full-size photography
Large imagery, structured specifications, and reviews with verified-buyer matching, so a handmade product is as easy to trust online as it is in person.
Reviews matched to real orders
Brand
The about page carries the maker story
Editable rich-text pages let the brand explain how things are made and who makes them, which is what separates a handmade shop from a marketplace listing.
Merchant-editable content pages
Objections
The FAQ answers the questions that stall a sale
Delivery times, cash on delivery, personalisation, care and returns answered in public. In this market the cash-on-delivery question is the one that decides whether a first-time buyer orders at all.
Objections handled before checkout
Enquiries
Custom orders are tracked as their own flow
A contact form that stores enquiries in the admin with new, read, replied and archived states, alongside the WhatsApp and Facebook routes the shop already uses.
Enquiries tracked inside the admin
Content
Care guides reduce returns
A full blog with cover images, tags, scheduling and related posts. Care guides answer questions after purchase, which cuts support messages and return requests.
Scheduled publishing built in
Delivery
Delivery timings are enforced by the system
Inside Dhaka, outside Dhaka and personalised-item timelines are published, and the same zone logic drives shipping rates and delivery dates at checkout. The promise and the system agree.
Zone-based rates, dates and slots
Merchandising
Every category has its own indexable page
Home and decor gets the same structure as every other range, so the catalog can grow into new categories without a redesign or a developer.
Collections the merchant creates
Product depth
Made-to-order items use a separate lead time
Personalisation, lead times and made-to-order expectations are set on the product page, which is where returns are prevented rather than processed.
Expectations set before checkout
Policy
Published policies reduce payment disputes
Refund, shipping and care policies published plainly. In a cash-on-delivery market, clear terms are part of the operations system, not legal filler.
Written to be read, not to hide
The client.
Mystic Shop BD sells handmade and imported goods across Bangladesh, where demand arrives from Facebook, most orders are cash on delivery, and margin is won or lost on courier performance.
At a glance
The engagement, on paper.
- Client
- Mystic Shop BD, a direct-to-consumer retailer selling across Bangladesh.
- Industry
- E-commerce, with demand driven largely by Facebook advertising.
- Starting point
- Needing a serious store without a growing stack of subscriptions and apps to make local workflows possible.
- Scope
- A full commerce platform: storefront, cart, checkout, payments, orders, inventory, returns, delivery, admin, email, and analytics.
- Stack
- Cloudflare Workers and D1, Hono, Drizzle ORM, TypeScript, and a React admin console.
- Ownership
- Full source code, database schema, and deployment configuration held by the business.
- Platform cost
- $0 a month while usage stays within Cloudflare's free allowances.
The challenge
6 problems with global platforms here.
Hosted commerce platforms are built around a Western card-payment model. Selling in Bangladesh means solving a different set of problems.
Subscriptions compound
A monthly platform fee is only the start once reviews, wishlists, returns, page building, email, and analytics each need their own paid app.
Couriers need real integration
Booking parcels by hand and copying tracking numbers does not scale, and it hides the delivery data that protects margin.
Cash on delivery is a margin risk
Fake orders and failed deliveries cost real money on every parcel that ships and comes back.
Facebook drives the demand
That only works if Pixel and server-side events agree with the product catalog, and most setups quietly disagree.
Rented platforms have ceilings
A new checkout rule or report can mean a higher plan, another app, or simply being told it is not possible.
Project goals
7 goals, in order.
Each goal made the next one possible.
01
Own the whole stack
Source code, data model, and deployment held by the business rather than rented from a vendor.
02
Make cash on delivery first-class
Design the checkout around how customers actually pay, not around a card form with COD attached.
03
Connect local payments
bKash, SSLCommerz, and manual wallet methods handled natively, with an incentive to pay in advance.
04
Automate the courier loop
Book parcels from the order, then bring delivery status back into fulfilment automatically.
05
Measure profit, not revenue
Join cost snapshots to orders so reporting shows margin instead of a flattering top line.
06
Defend the margin
Risk scoring, phone and IP blocking, and order verification before an expensive parcel ships.
07
Run it for nothing
Architect it so the platform bill stays at zero while the business is small, and scales deliberately.
My strategy
The decisions I made.
6 calls that shaped the build, and the reasoning I apply to custom commerce work.
Start from the local reality, not a template
Most platforms begin with card payments and add Bangladesh through apps and manual work. I began with taka, cash on delivery, wallets, and couriers, so the awkward parts are native instead of patched.
Own the checkout so incentives can change
Checkout is where margin is decided. Owning it means an advance-payment discount, a delivery slot, or a verification rule can ship the same week it is needed.
Make ad IDs and feed IDs the same IDs
Pixel, server-side events, and the product catalog all use the same variant identifiers. Without that, retargeting quietly misfires and reporting never reconciles.
Send courier results back to the order
Booking a parcel is the easy half. Returning delivery status to fulfilment is what turns courier choice into a measurable decision rather than a habit.
Report profit, not revenue
Cost is captured on the order line at the time of sale, so historical margin stays accurate even after prices and products change.
One codebase instead of a plugin chain
When reviews, returns, email, couriers and analytics share a source tree and a schema, an update cannot silently break the checkout.
Built for Bangladesh
What global platforms treat as edge cases.
The conversion loop
The conversion loop.
Every stage writes back to the order, so the shop can see which advert produced a delivered, profitable sale.
01
Advert
Campaign, source, and medium captured from the first click and carried through the session.
02
Product view
Browser and server events fire together, deduplicated, using the same variant IDs as the catalog feed.
03
Cart
A server-backed cart with early email capture, so an abandoned basket can still be recovered.
04
Checkout & payment
Local methods, an advance-payment incentive, and risk checks before anything is dispatched.
05
Courier
The parcel is booked from the order and tracked without anyone retyping a consignment number.
06
Delivery
Courier status returns to fulfilment, turning delivery performance into data the shop can act on.
07
Review & repeat
A review request goes out after delivery, and the customer enters segmentation for retention.
The economics
This platform against a rented store.
Shopify is an excellent product for merchants who want a standard hosted platform. For this shop, owning the code won on both cost and control.
Shopify BasicMystic
Platform subscription$29 a month billed annually, or $39 month to month$0 a month within Cloudflare's free allowances
Platform cost over 3 years$1,044 to $1,404 before apps$0 in platform hosting
Platform transaction feeThird-party fees can apply with external gatewaysNone added by the platform, gateway fees still apply
Bangladesh operationsOften needs apps or custom integrationTaka, cash on delivery, bKash, SSLCommerz, Pathao and Steadfast built in
Checkout controlBounded by platform rules and plan levelSource-level control over fields, incentives, and risk logic
Core featuresAdvanced workflows may need paid appsReviews, returns, email, analytics and couriers in 1 codebase
OwnershipRuns on a proprietary platformSource code and database schema owned outright
MaintenanceHandled by the platform, which is genuinely easierThe shop's own responsibility, traded for control
Shopify list pricing checked in August 2026 and subject to change. Payment-gateway fees apply to both. The $0 figure covers platform hosting within Cloudflare's free tier and excludes domain, advertising, gateway, SMS, and email costs.
The outcome
The platform, by the numbers.
What the shop runs on today, and what it is no longer paying for.
$0Monthly platform cost
$1,044+Avoided over 3 years
5Payment methods
2Courier integrations
Business impact
What this gets the shop.
Budget moves to the business
Money that would go to a platform subscription and a stack of apps goes into products, delivery, and advertising instead.
Margin is defended, not hoped for
Risk scoring and verification stop a share of the failed cash-on-delivery parcels before they ever ship.
Advertising can be trusted
Deduplicated events and matching catalog IDs mean the numbers in Meta reconcile with the orders in the database.
Demand becomes visible
Failed searches, wishlists, and views on sold-out items show what customers wanted and could not buy.
Courier choice becomes a decision
Delivery performance by courier, payment type, and risk level turns a habit into something measurable.
No ceiling on what comes next
A new rule, report, or integration is a code change, not a plan upgrade or a search for the right app.
Questions
The short version.
Is the platform really free to run?
The platform hosting is, with a clear boundary worth stating:
- What is $0: the Cloudflare Workers and D1 usage that serve the storefront, admin, and database, while usage stays inside the free allowances.
- What is not: domain renewal, advertising spend, payment-gateway fees, SMS, email above a provider's free tier, and courier charges.
- What happens at scale: passing the free tier means a deliberate paid upgrade, which is a growth event rather than a failure.
Why build custom instead of using Shopify?
Shopify is the right answer for many merchants. It was not the right answer here:
- Local workflow: cash on delivery, bKash, SSLCommerz, Pathao, and Steadfast are native rather than assembled from apps.
- Checkout control: advance-payment incentives, delivery slots, and verification rules need source-level access.
- Cost shape: no monthly platform fee and no app stack growing alongside the catalog.
- Honest trade: Shopify handles maintenance and brings a huge ecosystem. Owning the code means owning the upkeep too.
How does the Meta tracking differ from a standard setup?
Most setups fire a Pixel and stop. This one closes the loop:
- Two signals, one event: browser Pixel and server-side Conversions API with shared event IDs, so nothing is double counted.
- Matching identity: the variant IDs in the product feed are the same IDs sent with commerce events, which is what makes dynamic retargeting work.
- Attribution stored: campaign, source, and medium are frozen onto the order, so revenue can be traced to a campaign later.
- Consent aware: tags and server events respect the visitor's consent state.
What stops fake cash-on-delivery orders?
Several layers, because no single check is reliable on its own:
- Risk scoring: orders are flagged for elevated risk before dispatch rather than after a parcel is lost.
- Blocklist: phone numbers and IP addresses linked to abuse can be blocked directly from a suspicious order.
- Verification step: flagged orders wait for confirmation before a courier is booked.
- Feedback loop: delivery outcomes by risk level and payment type show whether the rules are actually working.
Can you build something like this for my business?
Custom is the right call when your workflow is the thing that makes you money:
- Good fit: local payment or delivery models that hosted platforms treat as edge cases, or app bills growing faster than the business.
- Poor fit: a standard catalogue with standard payments and no appetite for maintenance. Hosted platforms win there, and I will say so.
- Next step: book a free call and I will tell you honestly which one you need.
Keep exploring
View all
More case studies.
Your turn
Outgrowing your platform? Own the next one.
If your workflow is the thing that makes you money, it should not be limited by what an app marketplace happens to offer.