The challenge: ship faster without breaking trust

Lingokids runs a high-scale mobile app with two distinct audiences and success criteria: kids need a delightful learning experience, while parents need clarity and confidence to subscribe. That makes experimentation high leverage, but also high risk: a “small” paywall change can swing revenue, retention, and long-term engagement.

The monetization team wanted to move faster and test more ideas in parallel, without letting overlapping experiments invalidate results or let a bad change quietly roll out to 100% of traffic. At the same time, they needed guardrails to ensure short-term monetization wins didn’t damage retention.

‍

The approach: experimentation as a team sport

GrowthBook is used across Lingokids by engineering, product, and data, with roughly 40 active users inside a 150+ person company. GrowthBook has been broadly adopted across the organization for feature flags and experimentation, with multiple read-only stakeholders following along. The platform supports Lingokids’ React Native mobile app and Ruby backend, with results analyzed alongside their data stack.

“We use GrowthBook extensively for both experimentation and feature flags. We often have five or six experiments in flight, and things have been quite smooth.
— Paulo Alves Pereira, VP of Engineering

On the monetization side, experimentation is deeply embedded into how the team operates:

• Primary metrics: conversion + revenue
• Guardrails: retention and longer-term engagement signals
• Cross-functional workflow: product management shapes hypotheses, engineering implements safely, and the data team validates results and monitors health checks

“For the monetization metrics we normally use conversion and revenue. From a guardrail point of view, we keep an eye on retention.”
— Filipa Batista, Product Manager, Billing, Pricing & Monetization

‍

Doubling velocity: How the monetization team runs more experiments

Over time, the monetization team dramatically increased experiment throughput. Today, they run around 15 experiments per month, roughly doubling the number of tests they can execute in parallel.

Two GrowthBook capabilities unlocked this scale:

1. Namespaces for parallel testing

As experimentation increased, overlap became a real problem. Multiple experiments touching the same purchase flow could clash and contaminate results.

Namespaces solved this by letting the team bucket users into isolated traffic segments, for example, in 20% slices, so multiple experiments could run simultaneously without interference.

“Namespaces really shifted the velocity difference. We can bucket experiments, avoid clashes, and run more in parallel.”
— Alejandro Martí, Engineering Manager, Monetization & Billing

2. Configuration-driven experiments without app releases

For many experiments, Lingokids uses GrowthBook configuration parameters (via JSON) to control behavior dynamically.

Instead of shipping new code or waiting on mobile release cycles, the team can immediately launch new variants to adjust:

• Frequency rules
• Image URLs
• Paywall layouts

“We don’t need any code changes, we don’t need an app release. We just configure the new tests and launch right away.”
— Filipa Batista, Product Manager, Billing, Pricing & Monetization

This approach significantly reduced time-to-data and removed mobile release bottlenecks from the experimentation process.

‍

Risk management: Catching losers before they ship

High velocity only works if risk is controlled. Lingokids uses a disciplined rollout process for every experiment:

• 3–4 day health-check phase on very low traffic
• Data team validates metrics, distributions, and bugs
• Riskier experiments ramp gradually, often to 50% traffic
• Experiments are stopped early if conversion or retention turns negative

“Whenever we see it impacting conversion or retention by a certain amount, we stop it right away.”
— Filipa Batista, Product Manager, Billing, Pricing & Monetization

This process has helped Lingokids avoid shipping costly losers at scale.

‍

The results: Repeatable wins and a paywall breakthrough

15 experiments per month, with meaningful wins

The monetization team now runs around 15 experiments/month, roughly a 2× increase in parallel experimentation volume since adopting GrowthBook more heavily. Their typical “solid win” is ~5–6% uplift, with occasional standouts nearing 15%. Roughly one-third of experiments are winners.

A flagship win: 7% conversion and revenue lift from paywall redesign

One of Lingokids’ proudest examples is a paywall redesign in the gaming experience. The team iterated 4–5 times, evolving from a bare-bones layout (title, two products, button) to a clearer flow emphasizing one product, reduced cognitive load, and better communication of subscription benefits. The outcome: ~7% increase in conversion and revenue, without harming the play experience.

‍

Looking ahead: From faster tests to smarter learning

As experimentation scales, the team is candid about attribution complexity. With overlapping experiments and variability from user acquisition channels, it’s not realistic to sum per-test uplifts into a single “conversion went from X to Y” number. Instead, they focus on repeatable lift ranges, strong win rates, and post-rollout validation with their data team to confirm experiment effects show up in real-world business performance.

“It’s one thing to run one experiment. It’s another to connect it to all the previous iterations to understand if it’s worth continuing in that direction.”
— Alejandro Martí, Engineering Manager, Monetization & Billing

The team is actively exploring GrowthBook’s APIs, data access, and upcoming capabilities to help automate insight generation and guide future experimentation.