Redesigning the trial-to-paid journey

How a freemium plateau became structural conversion growth: trial-to-paid CVR within 14 days nearly tripled (+170%) over 18 months.

Nuvemshop’s freemium model brought new merchants in fast, but upgrades to paid plans didn’t keep pace. The cost of serving the free base kept growing without the expected return. Instead of one big fix, I designed six sequential interventions over 18 months: I owned the definitions, designed the flows and supported the development team through each release. Each layer was launched first in Brazil and then expanded to Argentina and Mexico.

The problem

The free plan accelerated acquisition across a high-volume merchant base, but conversion to paid plans had stagnated. More stores meant more operating cost, not more revenue.

The diagnosis

Premium features were hidden on the free plan. Store owners saw the tool as limited and, never seeing what they were missing, had no reason to upgrade. Interviews surfaced a second pattern: store owners postponed paying as long as they could and tied the investment to making their first sales.

The primary metric throughout this case is CVR within 14 days: the share of new trials that complete payment within 14 days of upgrading.

From diagnosis to validation, with data and people

  • Interviews with store owners. In-depth interviews with active and churned store owners to map trial behavior, upgrade motivations and drop-off points.

  • Flow audit. Heuristic mapping of usability barriers in the billing flow, finding where the interface created invisible friction and hid value.

  • Engagement analysis. Day-by-day cohort curves showed when store owners went passive, revealing that the conversion window was shorter than the model assumed.

  • Hypothesis validation. Test sessions with real users confirmed the drop-off points and validated the readability of the new modals before rollout.

  • Variable isolation. Whenever feasible, each intervention was validated with A/B tests and cohort tracking before full rollout, isolating the effect of each layer.

  • Progressive rollout. Brazil worked as the validation market: most layers launched there first and only then expanded to Argentina and Mexico, reducing operational risk across LATAM, especially on sensitive changes like the hard-block.

A new contextual upsell

Before, the upgrade path was a generic, scattered banner with no clear context (“Go to plans”) that sent store owners to an external plans page, away from what they were doing. The new model replaced it with in-context modals, built on three UX pillars:

  • Value exposure. Visual triggers inside the locked feature itself, at the moment of use (for example, bulk actions on the Sales screen), so the upgrade becomes a natural consequence of use.

  • Friction reduction. A single-CTA in-app modal overlays the task and keeps the screen visible in the background, so store owners decide without losing context.

  • Continuity. The feature unlocks on the same screen right after upgrading, with no redirect or reload: an immediate reward that reinforces the decision to pay.

Show the limit at the moment of action

Instead of silently blocking, the modal appears exactly when the store owner tries to use a paid feature: at the point of highest intent, not before it. It launched on the Sales screen in March 2024 (Brazil) and became the base model for every layer that followed.

The accepted risk

Interrupting a task can feel intrusive. We accepted the risk of irritation and session abandonment, betting that the context (a blocked action, not free browsing) made the interruption feel justified to the store owner. It held: support tickets and abandonment showed no negative impact after launch.

Six interventions, three markets, 18 months of data

Whenever feasible, each layer was validated with A/B tests and cohort tracking before full rollout. Most layers, especially the initial and structural ones, were tested in Brazil before expanding to Argentina and Mexico.

  • Mar/2024 · Brazil · New upsell model (Upsell v1). The first structural redesign of the conversion flow. It set the new baseline for everything that followed.

  • May/2024 · Brazil · 7-day trial. Competitors offered 14 to 30 days, so a shorter trial put acquisition at risk. Cohort data backed the call: daily engagement dropped sharply after day 7, no matter how much trial time was left. Because a shorter trial also pulls payments earlier, total CVR is the fairer read of this layer, not only the 14-day window.

  • Jun/2024 · Brazil · Freemium limit. Theme publishing was blocked on the free plan, a trigger at the store owner’s moment of highest effort. It was the largest single step in the series, and a new, higher floor consolidated.

  • Aug/2024 · All markets · Soft-block. A redesigned modal with a single CTA and an expiration soft-block made the financial commitment explicit before payment, turning an impulsive click into a deliberate decision. The gap between total CVR and CVR within 14 days started to close.

  • Nov/2024 · Brazil, Argentina, Mexico · New feature limits. Plan-based limits create conversion pressure independent of the trial: every time a limit is hit, it becomes a contextual upsell point. The higher floor held through 2025.

  • Jul/2025 · All markets · Hard-block. Unlike earlier layers, it acts after the upgrade: the admin locks from day 2 past due, with direct language about the commitment, removing the soft-block gray zone for store owners who upgraded but hadn’t paid.

A permanent baseline shift

Compared with the pre-intervention average (Jan–Mar 2024 = index 100), CVR within 14 days averaged index ~275 in the last quarter of the series (Jun–Aug 2025): nearly 3x the baseline. Total CVR rose about 60% over the same periods and settled on a new sustained floor. No single layer explains the result: each one raised the floor, and the gain held.

What I learned: high CVR without real commitment doesn’t become LTV

The simplified modal sped up immediate conversion, but CVR within 14 days showed that part of the upgrades were impulsive. Our read was that impulsive upgrades tend to churn earlier and distort cohort metrics, so the quality of a conversion mattered as much as its volume.

  • Immediate CVR isn’t real LTV. CVR within 14 days measures who actually pays, not who clicks. Raising it became our proxy for healthier revenue, and total conversions kept growing at the same time.

  • Positive friction as a requalification lever. The soft-block and hard-block didn’t reduce conversion; they requalified it. A store owner who goes through a clear block and still subscribes is making a deliberate decision. Used transparently, friction became a product lever, not just a cost.

  • What I’d do from the start. The risk was in the speed and order of iterations: a layer launched out of turn masked the effect of the previous one. Launching the soft-block alongside the first modal would have shortened the cycle and improved conversion quality from the first layer.

Metrics are shown as relative growth and indexed values, averaged by period (Brazil). Screenshots use dummy data from staging environments, with names and values blurred for confidentiality.