Platform Overview

ABetterChoice (ABC) is an all-in-one experimentation and configuration platform built for game studios and internet products. It integrates "which users see what" and "whether this change actually works" into a single product — you no longer need to switch between A/B testing tools, parameter management backends, and data analysis tools.

Three questions ABC helps you answer

Question	Platform capability
"Which version is better?"	Run controlled comparisons with multiple experiment types; the stats engine gives you a trustworthy conclusion
"Who benefits most?"	Causal inference (HTE) identifies how different user segments respond differently to the same change
"How do I release safely?"	Remote Config supports gradual rollout, canary traffic, and one-click rollback — no redeploy needed

Unlike standalone A/B testing tools, ABC puts the full decision loop — from hypothesis → experiment → analysis → deployment — on a single platform.

Four platform modules

Module	Problem it solves
Experiments	Validate product changes with controlled experiments — supports layer experiments, config experiments, and MAB (multi-armed bandit) experiments
Remote Config	Change parameters, copy, and reward values without a redeploy — the server pushes values in real time, effective within seconds
Audiences	Define segments like "high-spending users" or "new users in Japan" as reusable objects that experiments and configs can reference by name
Statistics	Translate "the metric went up" into "is this real" — includes CUPED, FDR, SRM detection, and HTE using industry-standard methods

Capabilities at a glance

Experimentation

Scientifically validate the effect of a change through controlled experiments; make decisions based on statistical significance, not intuition.

Experiment type	Description	Best for
Layer Experiment	Isolates traffic using a layer-and-domain architecture; multiple experiments run in parallel in independent traffic spaces	Most A/B scenarios, especially when multiple experiments need to run simultaneously without contaminating each other
Config Experiment	Binds Remote Config parameters to an experiment; runs independently at 100% traffic for different audiences simultaneously; results are automatically written back to the config when the experiment ends	Granular optimization across multiple regions or user segments
MAB Experiment	Multi-armed bandit algorithm that automatically shifts more traffic toward the better-performing group during the run	Quickly selecting a winner among multiple variants, or time-sensitive campaigns

Remote Config

Manage parameters from the server in real time — no redeploy required. Use a rule engine to serve different values to different user segments and roll out progressively by traffic percentage.

Capability	Description
Parameter management	Supports String / Number / Boolean / JSON; set a default value
Rules and targeting	Multi-level IF / ELSE IF / ELSE rule chains matched by audience and traffic percentage
Traffic rollout	Gradual rollout from 1% → 100% to reduce full-launch risk
Change audit	Every change is automatically logged; supports version comparison and one-click rollback

Statistics and analysis

Capability	Description
Variance reduction (CUPED)	Uses pre-experiment covariates to reduce metric variance — converges faster with fewer samples
FDR multiple comparison correction	Controls false positives in multi-metric scenarios to avoid "significance inflation"
Bayesian inference	Provides posterior probability and credible intervals — suited for small samples and continuous decision-making
Drill-down analysis (Explore)	Break down results by dimension to locate sub-groups that drive positive or negative effects
HTE heterogeneous treatment effect	Identifies differential responses across user segments — answers "who does this work for?"

HTE is a key differentiator for ABC compared to most competing tools — mainstream A/B testing tools generally do not include native causal inference.

Traffic quality assurance

Capability	Description
Dual-hash assignment	Double-layer hashing guarantees uniform random assignment, eliminating selection bias at the source
SRM detection	Monitors sample ratio mismatch in real time; alerts immediately on anomalies
AA / AB Backtrack	Pre- and post-experiment data quality diagnostics to identify root causes of anomalies

Where to start

Onboarding by role:

You are...	Recommended path
Product manager / Operations	Quickstart → Remote Config overview → Experiments overview
Data analyst	Experiments overview → Reading results → Statistics methods
Engineer / SDK integration	Quickstart Step 2 → SDK integration docs → Layer Experiment overview

What to read next

• Quickstart — run your first experiment in 6 steps
• Feedback and Support — how to reach us when you have a problem