By Rosie Jane teardown: "I said candy"

You: I like things that smell... Sweet like candy. Scent Finder: enter your email to get your results + 20% off your first order You: Can I just see the answer? Scent Finder: skip. (Small, under Submit.) Scent Finder: ROSIE PERFUME You: I said candy.

That's a paraphrase of the shopper's side, but every line Scent Finder says is on the page as of Nov 9, 2026, and the marketing box under the email field is already checked. The bottle that never comes up is Dulce, which By Rosie Jane describes, in its own words, as vanilla bean, brown sugar, and dulce de leche. Candy, basically.

Six questions, then that email screen. The result is Rosie Perfume, white musk and sweet rose from the Rosie page. Dulce is the candy bottle. It never shows.

A quiz that already wrote the notes still asks for an address before it will use them.

What By Rosie Jane does well

Dulce's Smells Like block is written the way a shopper talks: vanilla bean and brown sugar, then vanilla flower and dulce de leche. The discovery set lets you try all seven scents for $29, and it has 1,440 reviews. Build Your Own Bundle already does the bundle math.

That's a brand that knows its product. The quiz and search just need to read those notes.

What By Rosie Jane does badly

The quiz holds the bottle for an email, then asks again. Six questions in, Scent Finder wants an email and 20% off before it will name a bottle, with the marketing box already checked. Skip is under Submit. On the same page, an Attentive overlay can cover Take the Quiz with another 20% off email field. Close the overlay. Don't type an address. The shopper already answered.

Search treats vanilla like a shrug. Type vanilla and you get 140 results. The store's public product feed lists 158 products that day, and only 18 mention vanilla. Almost the whole store.

The FAQ still says $25, and the free box never lands. "You can try every fragrance for just $25," says the FAQ, linking to the discovery set that is $29 and final sale. The bar and the shipping page both say free holiday box at $65, but an $86 Dulce Perfume in the bag still leaves the Holiday Gift Box at $0 with Add to bag, never a cart line. If they write in, it's a contact form with a 36 to 48 hour wait, hello@ on the FAQ, or orders@ on returns.

3 quick wins

1. Answer first, then ask. Show the result card right after the last question and put the 20% offer under it, with the box unchecked. The shopper sees their bottle, then gets asked if they want a discount on it. A yes means a lot more at that point. The quiz owner and the email owner sign off together. Effort is small, since the result already exists behind skip.

2. Teach sweet to mean vanilla. Build a short list that maps quiz answers and search words to the notes already on each product page. Sweet like candy lands on Dulce. Vanilla puts those 18 products first, starting with Dulce and Lake. A merchandiser owns the list. Effort is medium, because it's search and theme work plus someone keeping the list current.

3. One price, one inbox. Fix the $25 line today. Then have support replies drafted from the live product page, the returns page, and the shipping bar, so the $29 set, the $65 threshold, and the inbox all match. A person sends every reply. Effort is small for the FAQ and medium for the drafts.

Managed AI agents

Two agents I'd start with. Both read only what By Rosie Jane already published, and both stop before anything goes out.

Support drafter. Reads the FAQ, the returns page, the shipping page, and incoming Gorgias tickets, then writes the reply for a person to send. It can be built on the Claude Agent SDK with a permission check on every send, or OpenAI's Agents SDK, which documents human in the loop.

Merchandising watch. Checks the quiz skip result, the FAQ sample price, and the shipping number on a schedule. When candy still returns Rosie, the FAQ still says $25, or the bag earns a holiday box that never lands, it flags it and drafts the fix. A merchandiser publishes it.

Technology detected

From response headers, loaded scripts, and public Shopify data on November 9, 2026.

My take

The stack is fine. A custom Shopify theme with Quiz Kit, Okendo, Klaviyo, and Gorgias can do everything above, so don't replatform. Give each product one record with its notes and availability, and have the quiz, search, and product page all read from it. Then the candy shopper lands on Dulce no matter which door they use. Watch Scent Finder: it already loads a dozen third-party tools, and an overlay can cover the first question.

Did your shopper already tell you they wanted candy? That's what my AI Opportunity Teardown is for. For $2,000 I spend a week on your site and data, and you get a written one-pager and a 15-minute Loom on where AI would actually help and where it wouldn't.