Understand how Optimize predicts whether a variation will outperform no change in AI Optimize.
Good to know
Strength is used only in AI Optimize optimizations — it doesn’t appear in test or personalize optimizations.
In AI Optimize experiences, strength is a performance metric that represents the probability a variation will perform better than No Change, based on recorded data. The higher the strength value, the more likely the variation is driving improved results. The lower the value, the more likely No Change is performing equally or better.
How strength works
When you launch a variation, its strength starts at 50%. As data is collected over time, the system calculates strength based on how the variation performs compared to No Change.
For example, if a variation consistently outperforms No Change, its strength might increase to 80% or 95%. If No Change performs just as well or better, the strength will decrease accordingly.
How strength is calculated
Optimize compares conversion rates and traffic for each variation against No Change over time. Based on this historical data, the system assigns a strength percentage to each variation. That percentage represents the variation’s probability of beating No Change moving forward.
This metric is designed to help you interpret performance more intuitively and make faster decisions about what to keep running — or what to replace.
What to do based on strength
- High strength (e.g., 70–100%) — the variation is likely outperforming No Change. Consider expanding on this idea or using it to inspire other variations.
- Low strength (e.g., 0–30%) — the variation likely isn’t performing well. You might want to pause it and try a new approach.
Deeper dive into strength
Strength is specifically designed for AI Optimize — where traffic is dynamic and variation performance evolves over time. Unlike traditional A/B testing metrics like statistical significance, strength accounts for real-world conditions like uneven traffic allocation and shifting visitor behavior. It’s more stable, more intuitive, and better suited to adaptive optimization.
If you want to take a deeper dive into the statistics and logic — learn why strength is used instead of stat. sig.