Experiments (also known as A/B Tests) are a crucial Zyntex feature. They allow you to run experiments on your game to test different features and see which one performs better.

​
Use cases

  1. UI/UX decisions: Test two different shop layouts, button placements, or HUD designs to see which version players engage with more.
  2. Monetization tuning: Compare different product prices, gamepass bundles, or Robux offers to maximize conversions without guessing.
  3. Onboarding & retention: Try alternate tutorials or first-time player experiences and measure which flow gets more players to return.
  4. Game balance: Adjust weapon stats, XP curves, or difficulty scaling between groups to find the “sweet spot” for fun and fairness.
  5. Event & feature launches: Soft-roll new events or systems (like quests or progression tracks) to only half your servers, so you can measure impact and catch issues before a full release.

​
Experiment terminology

​
Group

A group is a single group that is part of an experiment. Typically, there is a Control group and a Test group.

​
Entry

An entry is a log of a player entering an experiment. For example, if a player opens the shop, that would be registered as an entry.

​
Conversion

A conversion is a log of a player converting on an experiment. For example, if a player buys a product, that would be registered as a conversion.

​
Experiment flow

The experiment flow is as follows:
  1. A player enters the game.
  2. The player is assigned to a group via random assignment.
  3. If the player performs an action that is registered as a conversion, that would be registered as a conversion. Otherwise, it is registered as an entry without conversion.
  4. The results are aggregated and analyzed on the dashboard to determine which group performs better.

​
Experiment duration

It is recommended to run experiments for a duration of at least a week to ensure that the results are statistically significant.

​
Experiment best practices

  1. Run an A/A test: Before running an experiment, it is recommended to run a simple A/A test to ensure your setup is working properly. An A/A test is a test where you run the same experiment on two of the same group, so if the results are widely different, that suggests you have an issue with your setup.
  2. Use a large enough sample size: The larger the sample size, the more statistically significant the results will be.
  3. Use a large enough duration: The longer the experiment runs, the more statistically significant the results will be. It is best practice to keep an experiment running for at least a week.