ExampleFeature Flags · Remote Config

A/B Testing

Split users between two variants of a feature using percentage rollouts and remote config to deliver different experiences.

Setup in the dashboard

Create two feature flags and two remote config keys:

Flagcheckout_v250% rolloutGroup A sees new checkout

Configcheckout_cta_text"Buy Now"CTA for control group

Configcheckout_v2_cta_text"Get Instant Access"CTA for variant group

Flutter implementation

lib/screens/checkout_screen.dartDart

class CheckoutScreen extends StatelessWidget {
  const CheckoutScreen({super.key});

  @override
  Widget build(BuildContext context) {
    // Check which variant this user is in
    final isVariantB = Koolbase.flags['checkout_v2'] == true;

    // Read the appropriate CTA text from remote config
    final ctaText = isVariantB
      ? Koolbase.config.getString('checkout_v2_cta_text', fallback: 'Get Instant Access')
      : Koolbase.config.getString('checkout_cta_text', fallback: 'Buy Now');

    return Scaffold(
      body: Column(
        children: [
          // Show different UI per variant
          if (isVariantB)
            const NewCheckoutFlow()
          else
            const LegacyCheckoutFlow(),

          // Dynamic CTA button
          ElevatedButton(
            onPressed: () => _handleCheckout(isVariantB),
            child: Text(ctaText),
          ),
        ],
      ),
    );
  }

  void _handleCheckout(bool isVariantB) {
    // Track which variant converted
    // e.g. log to your analytics: { variant: isVariantB ? 'B' : 'A', action: 'checkout' }
  }
}

Deterministic assignment

Because bucketing is based on a hash of the device ID, a user is always in the same variant. They won't flip between A and B across sessions. This makes conversion tracking reliable.

Concluding the test

Once you have enough data to determine a winner:

Winner is variant B

Set checkout_v2 rollout to 100% in the dashboard — all users now see variant B

Winner is control

Set checkout_v2 rollout to 0% — all users revert to the control

Clean up

Remove the checkout_v2 flag check from code in the next release and delete the flag