Find winning variations that drive more traffic and increase user engagement.
The point of an A/B test is to control external variables. Testing content headlines through social media posts or tweets can make for significant variables, such as a time variable. For example, the majority of your audience might be active at 11am, so posting at that time might drive more engagement and create false winners.
Choose winning tests based on defined metrics like clicks instead of eyeballing based on multiple metrics.
Without a clear definition of a winning test, you’re opening yourself to confirmation bias, where you look for evidence to confirm your existing beliefs. Having a defined set of metrics lets you base your winning tests on actual data.
- Create Facebook custom audiences based on interest and split them into small age groups. For example, add users aged 21-22 in group A, users aged 23-24 in group B, etc.
- Create two ads using the different headlines you want to test but keeping the same copy.
- Run the ads and test their significance based on CTR. You could also set up UTM parameters to track each variation’s on-site behavior.
Similar to what you can do with email subject lines, you can send one headline to 10% of your list and the other headline to another part. Once you’ve deemed a winner you can then revise the headline on your site and send it to the rest of your list, too. The same idea can be extended to the actual article itself (if you have enough traffic).
Use tools like Conductrics, Dynamic Yield, or KingSumo to employ multi-armed bandit algorithms.
Using bandit algorithms involves exploring all “arms” of a test and monitoring user feedback, such as clicks, for each arm. Real-time feedback data is then used to gauge the performance of each arm until the bandit converges to only serving the best-performing arm. Alternatively, you could employ bandit algorithms yourself instead of using tools to do it for you. For example, you could display both potential versions of a headline on your homepage or news feed, and measure the CTR for each headline. Once the CTR for one headline exceeds that of the other, you simply switch to the one with the highest CTR for all users.
Last edited by @hesh_fekry 2023-11-14T15:55:27Z