← All posts

4 min read

From a parked domain to a shop taking orders: Besamim Perfumes

A perfume retailer had 110 bottles, stock in Shopify nobody could buy, and a domain pointing at nothing. What we built for them, the shop and the back office, and why each piece works the way it does.

Itai Rotem, Mexant

Besamim Perfumes is a perfume retailer. When we started, they had 110 bottles, a Shopify account full of stock that nobody could buy, and a domain pointing at nothing.

That combination is more common than it sounds. The inventory was real and already in a proper system. What was missing was the part a customer touches. So the first job was a shop, and the second was the screen the owner looks at every morning.

Both are running today. The shop is at besamimperfumes.com, and photographs of both are on our work page.

The shop

The storefront is built on Next.js and talks to Shopify through Shopify's own Storefront API. The design decision that matters most is the one you cannot see: stock and prices are read from Shopify at the moment someone loads the page.

The cheaper way to build a shop is to copy the product list into the website and update it now and then. That works until a bottle sells out, the copy still shows it, and a customer pays for something that is not on the shelf. Reading live means the shop shows what Shopify says, and Shopify is where the owner already keeps stock and prices. Nobody has to remember to update the website as well.

On top of that are the parts every shop needs: a product page per bottle with live stock, price and the notes it is built from, a basket, and a real checkout.

The trade portal

Besamim sells to other businesses as well as to the public, and wholesale buyers do not shop the way a retail customer does. They are not picking one bottle for themselves. They are building a case list.

So the shop has a separate trade portal where a wholesale buyer builds that list. A retail customer and a trade buyer want different screens, so they get different screens, on the same shop.

The picker

Most people do not shop for perfume by name. They shop for a moment. So the shop has a picker that reads what a shopper types in plain words, something like "something for a date night, under $80", and suggests three bottles.

The important part is who makes the final choice. The three bottles are chosen by our own code, from the bottles Besamim actually has, at the prices they actually set. So the picker cannot offer a fragrance that is not on the shelf, or a price nobody set. A shop assistant that invents a product is a support ticket at best and a refund at worst. This one can only point at the shelf.

The back office

The second build is for the owner, not the customer: an ops console that reads their Shopify and their QuickBooks at the same time.

The daily screen answers three questions: what is out of stock, what needs a decision, and what the shop is worth at cost. The console reads both systems to answer them, so nobody has to open two tabs and compare by hand.

The money screen puts the takings from the shop next to income and shrinkage from the accounts. It sits side by side on purpose. A shop's sales and its books should agree, and where they do not, that gap is where the money questions live. Seeing both on one screen is how you find it, instead of finding it at tax time.

Two rules the console is built around:

  1. Nothing changes without a person approving that exact change. Before the console changes anything in the shop or the books, a person sees exactly what will change and approves it, or it does not happen.
  2. Every change it makes is written to an append-only log. Entries are added, never edited or deleted. If a number looks wrong six months from now, there is a record of what changed it and who approved it.

That is also why you see the console blurred on our work page. It holds one business's accounts. Every figure on those screenshots is blurred before the picture is taken: their takings, their margin, the value of their own stock. We blurred them rather than swapping in made-up numbers, because a blurred figure is visibly one we chose not to publish, and an invented one would be a lie on a page whose whole point is that the screens are real.

What a small shop can take from this

You may not need a custom storefront or a console. But the rules behind them apply to any business selling online:

  • One source of truth for stock and prices. If your website has its own copy of your products, it will drift. Find out where your site reads its stock from.
  • Keep AI on a short leash. If you add an AI assistant to your shop, let it read what customers say, but do not let it make up the answer. Anything it suggests should come from your real catalog, at your real prices.
  • Put your sales and your books side by side. If the shop says one figure and the accounts say another, you want to know this month, not next April.
  • Approve changes, and log them. Any tool that can change your shop or your books should show you the change first and keep a record after.

The shop is built on Next.js, TypeScript, Shopify's Storefront API and Supabase. The console uses TypeScript, the Shopify Admin API, the QuickBooks API and Supabase. Both run on Railway. The full set of screens is on our work page.

If your stock lives in one system, your books in another, and your website in neither, tell us about it.