The Core Problem with Traditional A/B Testing
In performance marketing (Facebook Ads, Google Ads, TikTok Ads), testing multiple landing pages (A/B testing) is critical to boosting campaign ROI. However, traditional A/B testing methods suffer from major flaws:
1. Wasted Budget: Marketers split traffic evenly (e.g., 50/50) until they collect enough data. During this learning phase, 50% of your ad spend is continuously funneled into the lower-performing landing page.
2. Delayed Analysis: Manually calculating conversion rates (CVR), comparing Earnings Per Click (EPC), and adjusting traffic weights is slow, tedious, and prone to human error.
3. Stale Optimization: A landing page that won yesterday might lose today due to ad fatigue, audience shifts, or network changes. Static configurations can't adapt in real-time.
To address these pain points, PostbackFlow features a built-in Smart AI Auto-Optimization Rotator powered by the Epsilon-Greedy algorithm and Poisson-variance Z-Test verification.
---
What is the Epsilon-Greedy Algorithm?
Epsilon-Greedy is a mathematically proven reinforcement learning model used by large-scale dynamic routing engines. The rotator automatically splits traffic into two concurrent phases:
- Exploiting Phase (85% of Traffic): The system dynamically routes the vast majority of incoming clicks to the landing page identified as the statistical winner (the Winner). This ensures you capitalize on the best-performing asset instantly, maximizing conversion revenue.
- Exploring Phase (15% of Traffic): The system evenly splits the remaining traffic among the other landing pages. This continuous exploration checks if alternative landing pages improve their performance over time, preventing your campaign from getting stuck on a declining winner.
---
The 5 Campaign Rotator States
PostbackFlow automatically manages your campaign through a logical state machine:
1. Cold Start: Triggered when new landing pages are added and have not yet reached 100 clicks each in the last 24 hours. Traffic is split equally to establish baseline statistics.
2. Exploring: Occurs when landing pages have crossed the 100-click threshold, but the conversion difference is not yet statistically significant. Traffic remains split equally.
3. Exploiting: The Z-test identifies a winner with at least 95% statistical confidence. The winner receives the 85% traffic ceiling, while other pages share the remaining 15% floor.
4. Recovering (Outage Protection): If the winning landing page suddenly breaks or fails to convert (e.g., hosting crashes), the system enters safety mode, redistributing traffic to other active pages to prevent budget drain.
---
How to Enable AI Auto-Optimization
Activating the AI rotator in PostbackFlow takes less than a minute:
1. Add Multiple Landers: In your Campaign configuration page, add two or more destination URLs by clicking Add Lander URL.
2. Activate the Campaign: Ensure your campaign status is set to Active.
3. Turn on AI Controller: Toggle the AI Auto-Optimization Enabled switch to active.
PostbackFlow's background cron task runs every 5 minutes to evaluate click distributions, perform Z-test calculations, and update traffic weights automatically. You can sit back, relax, and let the algorithm optimize your campaigns!