Organic Olfactory

Context

A real shop, audited before it was redesigned.

Organic Olfactory is a niche-fragrance retailer in East Legon, Accra, selling 698 perfumes on a free Shopify theme. I audited the live store on 4 October 2026, on a phone and on desktop, and pulled its public Shopify data, so every finding rests on a number rather than an impression. Then I redesigned it as a working front end. Every bottle, price and size in it is the shop's own, and it sells nothing the shop doesn't. I sent it to the owner as a pitch, with the audit and a plan for taking it live. There's been no reply yet.

Problems

The hurdles, named plainly.

In one line: the store sells 698 high-priced perfumes, at a median of GH₵4,000, to people who can't smell them. Five things make that harder than it needs to be.

• There's no way to find a scent you don't already know by name. The only way in is a flat list of 58 brand names, and the products carry no tags. Yet 437 descriptions already list their notes, so the data exists; it just isn't structured.
• Expensive blind buys, with nothing to lower the risk. No samples, no reviews. The shipping policy is blank, and the refund policy is template text that excludes "flammable liquids", which in practice means perfume.
• Product pages don't sell the scent. 690 of the 698 products have a single photo, the notes are buried in paragraphs, and nothing links onward, so every product page is a dead end.
• Buying on a phone is slow. Add to cart sits below the fold, behind a size picker and a gender picker that each have one option. Delivery is "calculated at checkout".
• It looks like a default theme. Three loud colours compete, the wordmark sits on the green at 1.69:1 contrast, and white on the orange is 2.58:1. Both fail WCAG.

Approach

Design principles

Growing pains

What the old design could no longer carry.

The store runs on Shopify's free Sense theme with its defaults left in place: a generic template, now holding 698 products from 72 brands. Nothing in it is broken. It just never grew past the template.

• The homepage has two sections, and "Featured" shows the first four products alphabetically. The banner still reads "Welcome to our store".
• On a phone, a large brand-logo tile pushes the product grid below the fold, and mixed photo ratios make the grid jagged.
• The homepage HTML weighs about 1.07 MB, mostly a 240-country selector that prices everything in cedis anyway. It loads 60 scripts, and none of the 698 product images has alt text.

The redesign

Quiet luxury, built from tokens.

The direction is the restraint of a fragrance house rather than a marketplace: warm paper, ink, thin lines, generous space and one accent, a forest green deepened from the shop's own. Type is Instrument Serif for display and Inter Tight for everything else, including every price. Every colour, radius and space comes from one token file, with Tailwind's defaults cleared out, so an off-brand grey can't be used by accident. The product photos are the shop's own, set on one linen stage so their mismatched backgrounds read as a single surface.

• The cedi sign decided the typeface. My first choice for the sans, Hanken Grotesk, has no ₵ at all, so every price would have fallen back to a system font. I checked each font's glyph table before choosing.
• And then the file size. The ₵ lives in a 90 KB font file. Split out into a 1 KB font loaded only for that character, prices render in brand type and pay almost nothing for it.

Introducing

A scent finder that says why.

Five questions, about a minute: when you'll wear it, how you want to feel, what you love, how noticeable it should be, and your budget. It cuts the choice to three, and every pick says why it fits, like "Amber · Warm & cosy · Evenings out". The answers live in the address, so the results can be shared, which suits visitors who arrive from TikTok.

• Does this. Treats budget as a hard filter, never a suggestion, so nothing it recommends is out of reach.
• Shows this. The reason for every match, so a recommendation reads as advice instead of a guess.
• Invites this. Sharing the results, with a link that opens the same three scents for a friend's second opinion.

Ways in

Browse by feeling, and search that understands notes.

Families, moods, occasions and presence are stored the way they would be added to the real Shopify store, as product metafields, and the filters live in the address so every view can be shared. Search reads notes as well as names: "vanilla", "oud" and "creme brulee" work, accents optional, and an empty result points to the scent finder instead of a dead end.

The product page

A page that sells the scent, not just the bottle.

The notes are laid out as a pyramid, from the first fifteen minutes to the lasting trail. Under that: how it feels, when to wear it in Accra's climate, and how far it carries, labelled as an editorial guide because the shop's data doesn't say. On a phone, Add to bag stays one tap away in a bar that appears only while the main button is off screen. The bag suggests up to two bottles from the "pairs well with" list of what's already in it, never a random upsell.

Checkout

A checkout written for Accra.

A guest checkout with about five required fields. Mobile Money comes first, with the network chosen up front, and there's never a card-number field on the page. The rider needs a phone number and a landmark, not a postcode, so GhanaPost GPS is optional. Errors are listed at the top with links to each field, and the draft survives a refresh.

Alternatives

The routes not taken.

An earlier draft offered 5ml samples, discovery and gift sets, gift wrap and a travel atomiser. They would have answered the blind-buy problem better than any screen, and I took them all out, because the shop doesn't sell them and a prototype that does would be designing a different business. They became recommendations for the owner instead, and the design keeps room for them: a size choice on the product page, and suggestions in the bag. I also chose a React prototype over a Shopify theme, on the same router Shopify's Hydrogen framework uses, so the work carries over if it's built for real with the shop's checkout and stock staying on Shopify.

Reflection

Project takeaways.

• Some problems aren't design problems. The biggest risk in this store is a GH₵4,000 blind buy, and the real fix is a 5ml sample, which is a change to what the shop sells. Design's job there was to say so plainly and leave room for it.
• The data was already there. The store had no tags, but 437 descriptions already named their notes. Structuring what existed made browsing by feeling possible without inventing a single fact.
• Check the font before you fall for it. A typeface with no cedi sign would have broken every price on the site. Reading the glyph table first saved a redesign of the type.