The shopping-journey redesign behind R$296M in new revenue

By Q3, the dashboards were telling a story nobody wanted to read: revenue was sliding, with no single reason to point to. That ambiguity was the assignment. Rather than guess, I set out to find where the money was actually leaking.

Working with the analytics team, I walked the funnel step by step. The drop-off wasn't isolated to one page; it compounded across a handful of core touchpoints, product category pages, the product page itself, the add-to-cart action, and the cart-to-checkout flow, each one shedding a little more intent than the last.

A drop-off chart tells you where people leave, never why. Knowing the funnel bled hardest at the product page was a starting point, not an answer, so I went looking for the behaviour behind the numbers. My next move was to bring in Hotjar and, alongside the CRO team, analyze the conversion data module by module across the product page, measuring how much each section actually pulled its weight toward a purchase.

To separate signal from noise, I triangulated three sources.

1. Attention was pooling in the wrong places. Heatmaps plus a Customer Effort survey showed interaction clustering tightly on the product title, price, and Buy button, confirming those as the real decision-making zones, while the gallery, specs, and reviews got far less attention despite mattering to the purchase. On mobile, price and CTA weren't even visible on landing, which raised effort before a decision could start.

2. The content that drove purchase wasn't the content people could reach. A behavioral analysis of which sections most influenced purchase surfaced a clear mismatch. The Buy CTA (93.9%) and product characteristics (87.4%) were top drivers and sat above the fold, but User Reviews (57.5%) sat well below it, and specs were buried deeper still.

3. The reframe lived in one relationship. Plotting each module by scroll-engagement against conversion-impact made the core opportunity impossible to miss: User Reviews had high conversion impact but low engagement, because they were buried too deep. People who reached the reviews were far more likely to convert, so the reviews needed to come up to meet attention, not wait for it.

My main insight from the data was that the elements with the highest influence on purchase,reviews, trust signals, the CTA, and clear pricing, were exactly the ones users had to work hardest to reach.

To make that concrete, I mapped the whole product page into four engagement zones, from where attention peaked at the top to where it flatlined at the bottom. People scrolled roughly 60% deeper on mobile, but the shape held: the revenue-driving modules had to move up.

To turn the diagnosis into design decisions, I plotted every product-page module by how much users engaged with it against how much it influenced conversion. Two areas of the chart told me exactly where to act.

The amber zone was the headline opportunity. User reviews (buyer_opinions) drew one of the highest interaction-versus-conversion ratios on the entire page: shoppers who buy actively seek them out. Yet the module sat buried at the very bottom, far below where that intent lived. Moving it up and redesigning it was the single highest-leverage move the data pointed to.

The green zone was a cluster worth compounding. These modules already share the first and second fold, exactly where attention concentrates, so they earned high engagement but were still converting below their potential. Because they sit grouped in the same stretch of page, they could be redesigned together to pull more weight, rather than competing for attention that wasn't there.

The constraint I designed around. Given the technical limits of the VTEX platform, I deliberately chose UI solutions that could deliver high impact while staying lightweight, which also served an internal KPI: Core Web Vitals and loading speed. I passed on heavier interaction patterns that would have traded performance for polish.

I presented the findings and recommendations to the PM and PO, reviewed feasibility with the tech lead, and ran an Impact × Effort prioritization to turn the diagnosis into a roadmap: four product-page pain points plus one high-value checkout pain point.

The prioritization turned the diagnosis into a short list: four product-page pain points and one in checkout. Five problems, five moves, each tied to the value it had to deliver:

The voltage step was quietly bleeding users. Two problems compounded:

I pre-selected the default voltage to remove the redundant confirmation, surfaced the selector earlier, right on the product card, and made the Buy CTA persistent on mobile so the purchase action never scrolls out of reach.

The original first view worked against the buyer: a cropped main image, weak hierarchy, a pricing box crowded with competing values, and a Buy CTA pushed below secondary content. I reorganized it around what actually drives the decision:

Feature content was buried inside modals, forcing people to choose between reading a description and seeing the product. I moved it into a side-by-side, horizontally browsable layout so copy and imagery are absorbed together.

Being honest about this one, two things are worth calling out:

The reviews experience was the clearest high-impact opportunity the data pointed to, so I redesigned it fully: voting, filtering, media upload, a responsive layout, and a way to actually write a review, surfaced earlier where the data said it would do the most work. It was designed and validated, then blocked at the finish line by an organizational dependency, decommissioning the legacy review tool and onboarding its replacement. It never deployed. This was the sharpest lesson of the project, and it's the reason my first reflection below is about alignment, not craft.

With roughly 85% of users on mobile and Pix strategically important to cash flow (it's paid in full, upfront), the payment step was leaking intent: people chose Pix but didn't finish it. I redesigned the step around four moves: