At a glance
- Role
- Product manager and builder. I set the problem, the decision logic and the scope, and built it with AI coding tools
- Product
- RoomReady, a shopping prototype that turns a budget and a room’s purpose into a buy-now plan
- Built on
- Shopify’s Global Catalog API and the Universal Commerce Protocol (UCP): real products, prices, ratings and per-merchant checkout links
- Period
- November 2026, in a two-to-three-hour window
- Stack
- React and Vite, deterministic logic with no AI at run time, 12 automated tests
- Try it
- The prototype is live at oladejoqazeem.com/roomready
Context
People increasingly find products through AI assistants, conversations and social content rather than a store. Finding a product is not the same as feeling confident enough to buy it. Shopify is exploring shopping apps that close that gap while helping merchants reach the right buyers.
One example given is a first-time renter furnishing a home. They have not done it before, the choices are endless, and a mistake costs money they do not have. I chose that problem and built RoomReady on Shopify’s Catalog API and UCP in the two-to-three-hour window the exercise allowed.
The constraint shaped everything. With a few hours, the question was not what a complete shopping app would do but which single decision, done well, would make a first-time buyer confident.
The problem
A first-time renter is not buying a sofa. They are trying to make a room work without a costly mistake, and every tool they have is built for buying one product at a time.
I wrote down what goes wrong before I wrote any code.
The budget is for a room, not an item
Search and recommendation show products one by one. Nobody tells the buyer that the sofa they like leaves nothing for a lamp, or that the lamp can wait.
Everything looks essential
A rug, a side table and an accent chair all look like things a living room should have. The buyer cannot tell which ones the room needs first for the way they will use it.
Trade-offs are invisible until checkout
The buyer discovers they are over budget at the end, with no idea what to give up, so they either abandon the cart or overspend.
Shipping and tax arrive last
A plan that fits to the dollar fails once shipping is added. First-time buyers do not expect it.
Finding the cause
Plan before building
I set the decision logic first: which categories a living room needs for each way of using it, in what priority, and what the plan does when the budget is short. The code comments carry that hierarchy, dated, so a reviewer can see the thinking was not generated with the prototype.
Pull real products, then review them
I queried the Catalog for seven categories at three price bands and got 161 products. I reviewed them by hand and kept 29 that were the category asked for, with a usable image and a checkout link. The prototype says so on every page: it is a snapshot, not a live search.
Decide what the first version would not do
No room dimensions, no live search, no combined checkout, no AI at run time. Each of these is either a separate product problem or something the Catalog data could not support reliably in the time available.
Product decisions
Ask for how the room is used, not what the buyer wants
The inputs are a budget, what the room is mostly for (relaxing, working from home, hosting), what the buyer already owns and one thing they will not compromise on.
Priorities follow from use. A desk is essential for working from home and not needed for hosting. Asking about use lets the plan explain every item in one sentence the buyer can check against their own life.
Essentials always go in, even over budget
If the essentials for the chosen use cost more than the budget, the plan shows the shortfall and keeps them. It never drops an essential quietly.
A plan that fits by removing the sofa is not a plan. Showing the gap and letting the buyer decide is more honest than a number that looks right.
Optional items are suggested, never bought
Items marked Can wait sit under Buy Later and are not counted in the total. Leftover budget is shown, not spent.
The buyer’s question is what they need first. Spending their remaining money on a rug they did not ask for would reproduce the overspend problem the product exists to prevent.
Every trade-off is proposed, then accepted
When the buyer picks a more expensive product or moves something to Buy Now, a dialog shows the shortfall and the options, ordered by how little they disturb the room: cheaper alternatives first, postponing an optional item next, raising the budget last. The plan does not change until they choose.
Confidence comes from seeing the consequence before committing. The engine only ever proposes; the buyer decides.
Hold 10% back for shipping and tax
The plan spends 90% of the budget and labels the rest as set aside, because the Catalog does not confirm shipping or tax before checkout.
Better a buyer finishes under budget than discovers the overrun at a merchant’s checkout page.
Hand off to each merchant’s own checkout
The Ready to buy panel groups items by merchant and links each to its checkout URL from the Catalog. RoomReady does not take payment.
UCP gives a checkout link per product, not a cart across stores. A prototype that pretended otherwise would misrepresent what Shopify’s platform does today.
What shipped
Inputs
Budget, room use, items already owned, one non-negotiable item. Changing any of them rebuilds the plan.
Buy now, Buy later, Left out
Each item carries its priority, a one-line reason tied to the chosen use, the real product, price, rating and dimensions where the merchant provides them.
Budget panel
Spent, set aside for shipping and tax, and what is left to keep or put towards an upgrade.
Trade-off dialog
For any change that breaks the budget: the shortfall, the options ranked by functional impact, and a plain warning when an option touches an essential item.
Undo
Every accepted change shows what changed and can be reversed in one click.
Checkout handoff
Items grouped by merchant, each with its Catalog checkout link.
Tests
Twelve automated tests cover the plan hierarchy, the budget maths, the trade-off ordering and the rule that essentials and protected items are never removed.
Delivery
I planned the logic and the scope before the first prompt, then built with AI coding tools in short cycles: one behaviour at a time, each checked in the browser and in tests before the next.
The Catalog query and the hand review of results came first, so the product decisions were made on real data, not placeholders.
The prototype states its limits on every page: the snapshot date, the number of products kept, that shipping and tax are estimated, and that each merchant checks out separately.
The version on this site has two small changes made after the exercise: it works on phone screens and no item is protected by default.
Measuring success
The job of the product is confidence, so the measure is whether buyers complete a room, not whether they click a product.
Plan completion: the share of sessions that reach a plan with every essential inside budget.
Trade-off acceptance: how often a proposed option is accepted rather than cancelled, and which option types win.
Checkout handoff: the share of Buy now items that reach a merchant checkout, and how many merchants a plan spans.
Returns and regret: whether buyers come back to change a plan, which would show the first plan did not hold.
Outcomes
A working prototype on real Shopify Catalog data, built and submitted inside the window.
A plan builder and trade-off engine that never change the plan without the buyer’s say, with the reasoning for each item visible on screen.
A live version on this site, so the decisions can be tried rather than read.
Lessons learned
The scope is the product
With a few hours, every feature I did not build was a decision. Cutting room dimensions, live search and a combined cart left one problem done properly: allocating a budget across a room.
The weakest part is silent quality
The plan upgrades an essential item one price step when the budget allows, and the card does not say so. A buyer could think the cheaper option does not exist. That is the first thing I would change: show the upgrade and the option to undo it.
Real data changes the decisions
Dimensions came back for some products and not others, which is why a fit check was cut. A prototype on mocked data would have shipped a feature that fails in practice.
What I would build next
Live Catalog search with the same hand-review rules applied automatically, a fit check for products with dimensions, bedroom and home office as room types, and a saved plan the buyer can return to as prices change.
Explore the product
Three ways to use a room, three different plans
The same seven categories, prioritised differently by what the room is for. Pick a use.
Relaxing and everyday living+
Essential: sofa and lighting. Important: coffee table. Can wait: rug and side table. Not needed: desk and extra seating, unless the buyer adds them.
Working from home+
Essential: desk and lighting. Important: sofa and side table. Can wait: coffee table and rug. Not needed: extra seating.
Hosting friends and family+
Essential: sofa and lighting. Important: coffee table and extra seating. Can wait: rug and side table. Not needed: desk.
Try it
The prototype is live on this site with the real product data. Set a budget, choose how the room is used, and change something to see the trade-off dialog.