Modest fashion is not a generic ecommerce category. Customers are buying garments where fabric weight, drape, and finish matter as much as fit — and where authenticity isn't a marketing claim, it's the entire point. The category in Bangladesh runs largely on Instagram DMs and templated checkout pages; very few brands manage to look online the way their pieces actually look in the hand.

Our brief was to build a storefront that did three things at once: present each piece (abaya, hijab, thobe, kufi, attar, prayer accessories) with the care a careful shopper expects, speak natively in both English and Bengali, and accept Cash on Delivery as a real first-class payment — not as an afterthought tacked onto a Stripe-first flow. Get those three right and the rest of the funnel falls in line.

The storefront is a headless Next.js front end against a Magento backend. Magento carries the ecommerce primitives we didn't need to reinvent — catalog, carts, orders, customer accounts, inventory, taxes, returns — while Next.js gives Al-Bayt a fast, custom, fully-controlled experience. Product pages render statically with on-demand revalidation; cart and checkout are dynamic against Magento's GraphQL API. The seam holds because the boundaries are clean.

Tailwind carries the visual system. We built a small set of primitives — product card, price strip, variant selector, COD-aware checkout step — and composed every page from them. Photography does most of the work; the layout's job is to get out of its way. Type scale, whitespace, and image ratios are tuned for garments, not gadgets.

Cash on Delivery sits inside the checkout flow as a peer of card and wallet payments, not behind a 'more options' fold. The address step validates against deliverable postcodes; the order confirmation page is bilingual the moment the customer picks a language. Cloudflare in front of Vercel keeps cold-cache load times under a second across the Bangladesh-to-Gulf corridor that carries most of the international traffic.

The first checkout iteration asked for full delivery address before payment selection. COD-first shoppers — the majority of the catalog — bounced when they saw the form before the trust signal. We re-sequenced the steps: payment method first, then address, then confirmation. COD completion rose materially in the first two weeks after the change.

Search was the second adjustment. The initial implementation searched only product titles. Bengali shoppers searched in transliterated form ('abaya' vs 'আবায়া'), and category words ('prayer', 'attar') as often as product names. We extended the index to include transliterations, category, and tag fields, and added a small popular-search row beneath the search input (Hijab · Abaya · Thobe · Prayer mat · Kufi · Attar). Bounce rate on the search page dropped sharply.

Inventory is the next room to unlock. The catalog has grown past 500 products with a fast-moving subset that goes in and out of stock daily; the team is moving from manual updates in the Magento console to a small in-store-app the warehouse staff use to mark stock on receipt. The headless boundary makes that work invisible to the customer.