How research uncovered a faster path to

How research uncovered a faster path to purchase

Saúde iD

company

UX Researcher

role

1-month validation

timeline

WhatsApp, Google Optimize

tools

A/B Test

method

THE PROBLEM

Sales were stuck, and users were leaving early

Saúde iD was underperforming in exam sales, reaching only 27% of its quarterly target. Recent interviews and data revealed that customers were unfamiliar with medical terminology, making it hard to complete purchases through the existing search experience.

WHERE USERS DROPPED OFF 3-months

64%

of visitors left the site without interacting with anything

10%

of the rest engaged with the homepage search

of those searchers reached an exam page and completed a purchase

THE INSIGHT

The same exam appeared under different clinical names depending on the medical reference used, increasing cognitive load and abandonment during search. I surfaced this by pairing funnel analysis with an in-page popup survey and quick follow-up interviews with users who struggled to find their exam.

THE EXPERIMENT

What if we removed search entirely?

>> THE SETUP >>

To test whether searching by clinical taxonomy was the real barrier, I ran an A/B test with roughly 20,000 visitors.

Using Google Optimize, I split traffic 50/50 — 10,000 per arm (A = original search, B = experiment flow)

In the prescription arm, our support team read the prescription, built a pre-filled cart, and sent a checkout link, letting us validate real demand without building anything.

>> WHAT I MEASURED >>

I compared completed purchases between the two arms, alongside CTA click-through and prescription sends, the same funnel where users had been dropping off.

43% of visitors who entered the B flow engaged with the feature, and the prescription arm drove 30% more completed sales than the control.

THE RESULT

The barrier was findability, not search quality, and fixing it lifted sales 30%

Against the control, the prescription arm converted 30% more visitors into completed purchases, all fulfilled through WhatsApp. Because both arms saw the same catalog and prices, the lift pointed to discovery, not search refinement, as the primary driver of conversion. Even with a fixed price shown up front, engaged users completed payment, which suggests price wasn't the immediate blocker for people who reached the cart.

Flow B results (the winning arm):

10,000 visitors entered flow B
43% CTA click-through
2,142 WhatsApp conversations initiated
641 prescriptions sent
54% of those completed the purchase

→ Flow B converted 30% more visitors into completed sales than flow A (control)

THE IMPACT

Building a new feature

The experiment gave the product team enough evidence to prioritize a native prescription-upload feature. The prescription reader is now part of the platform, a lasting product change informed by research-led experimentation, with WhatsApp kept on as a supporting channel while the in-platform solution was built.

After the feature shipped, I tracked homepage visits, interaction with the new feature, prescription sends, and completed purchases to confirm the effect held.

reflections and takeaways

This case reinforced that search optimization isn't always the right problem to solve. Here, replacing an interaction proved more effective than refining it. Controlled experiments like this are powerful research tools when development is constrained, letting a small team (in this case, me and one PM) validate demand quickly and make more confident investment decisions. The main limitation: the winning arm bundled several changes at once, so while it clearly beat the original flow, isolating exactly which element drove the lift was left for later iteration on the native feature.