Buyers weren't failing to start a purchase. They were failing to finish it.

Context

A marketplace with two ways to buy a box.

Businesses could buy standard boxes like in any online store. The real differentiator was custom packaging. Picture a pizzeria that sells an 80 cm pizza: without an account, the owner builds a packaging project from scratch, choosing board, thickness, coatings and artwork, and the five best matching suppliers can quote it.

Projects often looped back, with a supplier asking for a sharper artwork file before quoting. Only then did the buyer pick a quote, pay and follow production.

The brief

Make a mature product mobile first.

Packfy was built for desktop: long forms, big tables, repeated information. The full mobile first rethink would take well over a year. Two problems couldn't wait: poor search ranking with slow pages, and conversion that had fallen for months to a single completed order.

Were users leaving because the purchase was hard to start, or hard to finish?

Mapping the experience

First, understand the system.

I mapped every journey, the jobs behind them and the friction along the way, with a heuristic evaluation and a WCAG accessibility review of each screen.

The custom journey was a chain of fragile handoffs. Quotes came by email, one per supplier, over days, to addresses that were never validated. Links broke or expired. Returning buyers hit a long registration form, then struggled to find their order, with the spec on one screen and the quotes on another.

Funnel analysis

The data turned friction into evidence.

With the Data team, I rebuilt the conversion funnel in Mixpanel one question at a time. Did people reach registration from the quote email? Start the form? Finish it? Find their quotes, reach checkout, pay?

Registration was one of the biggest drops. Buyers arriving from a quote link had to give fiscal and billing details before they could see a price.

User research

The person asking for a quote was rarely the person buying.

Structured interviews across four profiles: standard buyers, custom buyers, buyers who abandoned registration, and sellers. We reached five people in most profiles but only two who had abandoned registration, so I weighed their input accordingly.

The pattern held. The person asking for a price was often an employee, an owner scouting a supplier for Finance, or a new business checking costs. None had a reason to hand over fiscal details yet. The platform had confused asking for a price with being ready to buy.

Sellers saw the same journey as a queue of open questions, often pricing outside the platform and coming back to enter the quote.

The real problem

It was also breaking at the end.

Years earlier, to simplify development, standard and custom purchases had been given one shared checkout. For users they were very different: a custom order arrived with a full spec and an agreed quote, a standard one with almost nothing.

The team assumed users weren't reaching checkout. Checkout itself was one reason they weren't converting.

Prioritization

We couldn't redesign everything at once.

I ran an impact and effort workshop with Product, Marketing and Engineering. Checkout came first: it sat closest to revenue, it affected both ways of buying, and both the funnel and the interviews backed it. It became the first flow rebuilt mobile first, and the reference for everything after.

Quick performance fixes ran in parallel to help search ranking while the redesign moved forward.

Design

Less information, at a better moment.

I mapped every field the checkout collected against everything the buyer had already given, then went through it field by field with Engineering: what had to be stored, what was duplicated, what could move elsewhere, and what was collected only out of habit. The goal: ask the buyer for less at the moment of deciding.

1. 01Stop asking twice.

   Information from registration or past orders is carried into checkout to confirm, not retype.

2. 02Show a custom order as a custom order.

   Checkout shows the agreed spec, the seller and the accepted quote, instead of squeezing it into a catalog template.

3. 03Let returning buyers skip the card.

   Packfy's customers reorder. With Legal, we cleared stored cards, within PCI DSS and Brazil's data protection law, the LGPD.

4. 04Start from the smallest screen.

   Without extra space to fall back on, hierarchy, grouping and a visible sequence of steps had to carry the complexity. Desktop scaled up from there.

Validation

Proven in the same funnel.

Mixpanel events were built into the new checkout from day one, so it was measured against the discovery baseline rather than a feeling. Conversion improved across the checkout. The exact figures are under the client's confidentiality agreement.

Beyond the checkout

A foundation for the rest of the platform.

I documented guidelines for the next phases, helped plan a roadmap from the most critical flows down, and reviewed redundant data with Architecture so the platform would keep asking for less. WestRock's internal team took the transformation forward from there.

Impact

The biggest change was where the redesign began.

How the work was built

Methods, in order.