In Google Ads, Conversion Lift studies will begin to use Bayesian statistical methodology to increase the certainty of lift you can expect at the end of your study. Currently, Google Ads uses Frequentist methodology for most accounts, but in 2025 there will be a gradual transition to Bayesian methodology for all Google Ads accounts. Reach out to your account manager to learn more about transition timing. The Bayesian approach has a few key benefits for reporting and feasibility (also known as “Study power”) tools:
- It primarily relies on data collected during the study, while also leveraging historical campaign data to supplement the analysis and improve study power.
- It allows you to run studies with lower budgets and fewer conversions than traditional methods while increasing the chance of observing lift.
On this page
- How it works
- How do you interpret Bayesian results
- Frequently asked questions about Bayesian methodology
How it works
When you set up a conversion lift study, Google Ads will be able to use Bayesian statistical methodology based on aspects of previous studies you’ve run (the campaign type, performance metrics, as well as product vertical) in order to increase the likelihood of detecting lift. Because Bayesian methodology leverages historical campaign data (also known as priors), you can get more precise, reliable results than with the Frequentist methodology while leveraging fewer conversions and a lower budget. Learn more about Conversion Lift feasibility. Additionally, read more to learn to Set up Conversion Lift based on users. To learn more about Bayesian methodology, you can reach out to your account team.
How do you interpret Bayesian results
Bayesian methodology uses “Credible Intervals” which are similar to “Confidence Intervals,” but have a different interpretation. The true value of your lift has an 80% probability of being within the Credible Interval, which is determined by combining your experiment’s data with the historical campaign data.
How does this change the way we report results?
Google will still report your point estimate (for example, the percentage of relative lift), but the interpretation of confidence intervals may change.
Consider the following example: An experiment concluded with a 90% certainty of lift and upper bound of 15% and a lower bound 5%, and a lift of 10%.
- In Frequentist language, the lift score would have shown that your lift is statistically significantly positive. That means that if your campaign truly had no lift, it would have a less than 10% chance of showing equally positive results (lift of 10% or higher). Also, there is an 80% chance that our reported confidence intervals contain the true lift.
- In Bayesian language, the interpretation of the lift score would be “There is a 90% probability that the campaign had increased your conversions." In addition, because 80% credible intervals were reported, the interval could be interpreted as having an 80% chance the lift is between 5% and 15%.
Frequently asked questions about Bayesian methodology
Why can't I find Conversion Lift study results?
You may not be able find the Conversion Lift study results in the following scenarios:
- Results won’t start showing until the 4th reporting day of your study. Due to the time required to process conversions, this may take up to 10 days. You can expect variations in your results as we collect data for the experiment.
- Results will be reported in the Reporting tab in Lift studies once if your certainty of lift is above 50%.
- You haven’t met the minimum conversion quality threshold (150 conversion coming from the treatment group, and 65 from the control group).
- Your spend is below $5,000 USD.
What are the minimum study requirements?
For advertisers looking for directional lift (lower than 90% confidence), any budgets above $5,000 USD and 1,000 conversions will allow access to these results.
