A redesign shaped by award-winning benchmarks and behavioral data, rebuilt inside a new design system on a launch deadline that would not move.
Brastemp's new premium refrigerators sat on a generic product page. Value went unperceived, sales leaned on discounts, and leadership had named the redesign one of the company's top 10 initiatives.
Lift conversion on a high-ticket product and give the page a bespoke, aspirational, premium feel worthy of the new positioning. All on a fixed launch timeline, mid-migration to a new design system, and with hundreds of possible directions to choose from.
I led it end to end: the discovery (competitor benchmark, behavioral data, CES survey), the redesign from generic catalog to a real buying journey, and a high-fidelity prototype built on the new design system's tokens and components. I coordinated my product squad and partnered with Design Ops, Content, and Data. My hardest call was defending, with the Insurance & Services team, that the page should secure the product's conversion before selling the add-on.
Add-to-cart nearly doubled when the content stopped hiding behind a click.
Also new: the structure was shipped as a reusable blueprint for every premium product launch that follows it.
Mad Max was Brastemp's move to reposition the brand into aspirational territory, alongside the launch of its premium appliance line. Leadership defined it as one of the company's top 10 initiatives, in their own words:
We want to recreate our Product and Category pages with more immersive and interactive content and communication, simplifying the way we present our products.
| AS-IS | TO-BE |
|---|---|
| Technical barrier: complex features, value not perceived | Assertive communication: spotlight on differentiators and features |
| Reliance on incentives: selling on discount, coupon, shipping | Immersive content: inspirational, interactive product pages |
| Engagement gap: static content, technical language | Brand and consumer connection: solving everyday pains |
My first move here was to reduce risk: to find the real opportunity inside the space that had just opened up.
I grounded discovery in three axes:
From the benchmark, six patterns stood out:
The page as a cinematic narrative, where each scroll reveals a chapter.
Overview on the surface, detail on demand.
Cinematic visual treatment, aspirational styled photography.
Modules you navigate by angle or feature, where the user investigates rather than scrolls past.
The product’s visual language dominates the whole digital experience.
Value shown with visual tools and short video, not just text.
During this stage I also mapped the dependency that worried me most: our graphic materials were simple next to the competitors'. Without premium photography and per-feature video, scroll storytelling had nothing to fill the screen. That became a work stream of its own.
A UX critique of the current product page then confirmed, from the UX side, exactly what the kickoff had flagged from the business side: static content, value not perceived, over-reliance on incentives.
What's missing to reach the next level: Bespoke Aspirational
Two of those critical gaps were not mine to fix on a screen. The line launched with just 2 SKUs, small enough to produce the missing photography and video by hand, but more launches were coming, so I did not solve for two. I gave the content team a production guideline built to serve every future launch, and defended the budget for a new image set by arguing its reuse across inbound and outbound.
The benchmark opened more directions than one page could hold. Crossing them against how the current page actually behaved is what turned a list of possibilities into where to act, what to change, and why.
Three sources, none of them answering the same question:
I ran the behavioural cuts with the Data team, and built the personas with our UX Researcher from the two together.
I’m just having a look. If the page doesn’t show me what I want right away, I close the tab and I’m gone.
Hook, don’t close.
Rarely buys on the first visit, so the job is not the sale: it is an impression strong enough to anchor value for the next one, and enough pull to keep the scroll going.
I want to buy, but I’m not impulsive. I need to know it fits my space, that the math on shipping and installments works, and that other people trusted it.
Answer the basics, then sell the features.
Colour, dimensions and the cost of ownership settle first. Only then do the killer features land, and they land better shown than described.
Two themes carried 83.6% of the complaints, at 41.8% each. Together they describe one problem: the old page asked too much effort, first to find the truth, then to get past the noise.
Information like the height should be at the top. Hunting for it is exhausting, I’ll give up.
Make technical information easy to find.
It produced two modules: a dimensions view for the objection, and the Product Tour for the features nobody was reading.
Too many pop-ups in your face in front of what you want to see, and you just give up.
Strip the clutter before adding anything.
The dense add-on block moves out of the first fold, which turned out to be the hardest call in the project.
Lift is not conversion rate. It is how much more buyers use a module than non-buyers.
Each one settled a different question:
That left me with hypotheses rather than convictions, which is the point: make the basics findable and the Validator stops leaving to search for them; expose the differentiators instead of hiding them behind a click and the Window Shopper goes deeper; move the add-on behind the decision and the product's own conversion rises without the attach collapsing. Wave 1 tested the cheapest of the three before anyone spent a sprint on the others.
The launch dates would not wait for the Orbit roadmap, so I validated the direction on the old site first. The central move was a new differentiators module: collapsed, click-to-open sections swapped for open ones in an alternating image-and-text sequence, in vibrant marsala. The bet was simple: expose the content the data called high value, instead of hiding it.
Before the A/B test even closed, the job shifted: v2 was the whole page rebuilt inside Orbit, a reusable blueprint rather than a one-off screen. Orbit lived in Storybook as a library of components, so I composed each module inside its logic, reusing agnostic components where their quality held and building from tokens, atoms, and molecules where it did not, negotiating with Design Ops which parts they would ship and which were mine. My rule for spending: new-build budget only where impact was high and nothing existing solved it. Everything else, recompose or reuse.
I documented each decision as a test card: we believe that… / to do this we will… / we will know we were right if…
| Module | Build | We believe that… | …we were right if |
|---|---|---|---|
| First fold: expanded gallery + anchors | Recomposition | the #1 CES pain (validating the basics) blocks the purchase | internal search for dimensions drops and qualified scroll rises |
| Dimensions & manuals toggle + exploded view | Recomposition | the fear of "it won’t fit" is a silent objection | post-purchase questions and returns by dimension drop |
| Relocated cross-sell (warranties/installation) | Recomposition | the attach happens better post-decision | product conversion rises without a proportional drop in attach |
| Rebuilt reviews | Reuse | social proof (+65%) supports the Validator | engagement rises and scroll reaches the module |
| FAQ and nearby alternatives | Reuse | residual objections stall the final step | abandonment from unanswered questions drops |
| Product Tour | New build (big bet) | it is the only module that attacks the inertia (detractor #1) | the Window Shopper interacts and findability of technologies rises |
The old reviews block was legacy VTEX. Instead of designing a new one, I reused the component Orbit had already shipped for Whirlpool Colombia's launch: aggregate rating, score distribution, filters, and verified purchase, all inherited. The lift was positive but moderate (+65%), so it didn't earn a new build. It earned good reuse.
The remaining modules followed the same logic. FAQ and nearby alternatives were anchored in the real objections: voltage, installation, warranty. The cross-sell relocation was specified and never shipped, which is the subject of the last section.
A styled kitchen with the refrigerator at the center, navigation by feature, and a layered feel that puts the user inside the scene. My most deliberate call was an automatic timer, 5 seconds per feature, so each technology presents itself without a click.
These are Wave 1's numbers, measured on the old site during the launch period: the differentiators module de-collapsed, its content exposed instead of hidden behind a click, against the version that still hid it. The rebuilt page in Orbit is not in them.
The result
How it got there
The gains do not come from the layout alone. Parallel initiatives, such as Project Themis (tracking accuracy) and the Complementary Feed (qualified traffic via regional offers), were fundamental levers. The cleanest causal reading comes from a category where price was neutral and the decollapsed layout still beat the average. That is the argument I stand behind without caveats.
What we still need to measure. Wave 1 proved that exposing the content was worth doing, which is what justified building the rest. What it cannot say is what the rebuilt page is worth. The full v2 in Orbit has not closed, and that figure will follow here when it does.
Moving Insurance & Services out of the first fold was my hardest call, and I treated it as a decision about a page. It was a decision about an organisation. Warranties, installation, trade-in and parts belong to other teams with their own targets, and a display strategy that settles where they live has to be built with those teams, not defended against them. I made the case from both sides, the behavioural data and the CES, and designed the move as a controlled test on a single SKU so that nobody had to take it on faith. It never ran. Those teams were absorbed by the platform migration and the capacity was not there for this front, so the module is still in the first fold and the proposal is still a proposal. That gap is the thing standing between this page and the branded journey it was meant to become.
It also helps to be exact about how far this went. On the maturity ladder I use for premium pages, the page shipped at stage three of five.
Reaching stage three on 100% existing assets was a choice, made to fit a launch date that would not move. Stages four and five do not need more design. They need content produced alongside the product.
So I took it to the PM and the managers as a production problem rather than a design one, with two proposals. Every launch ships two cuts of each feature video: one with burned-in text for social, and one clean, usable in a product page, a landing page or the catalogue. And one track from launch kickoff to live, where design, content and marketing agree the asset matrix per feature up front, instead of each channel commissioning its own. The production already happens. It is just fragmented, so the model reorganises it rather than adding to it.