Traffic Distribution, Holdouts & A/B Testing

This pages talks about you can add traffic distribution to your Fibr personalization campaigns to determine the% of traffic eligible to see the variants created and conduct A/B Experiment

While you generally want to include as much traffic as possible in your Fibr personalization, there are times when you might want to limit the amount of visitors that see your experiences, and essentially conduct an A/B experiment with your existing page to see what is working better for you. Traffic distribution allows you to do this.

In this doc, we will explore:

Defining & Using HoldOuts

What is a HoldOut?

While creating any personalization campaign, you can add holdouts as an attribute. This allows you to set a percentage of traffic that you do not want to show the personalized varaint of the landing page

For example, let's say you are building an experience to message match your homepage headline with a specific ad. Where we changed the H1 heading to the ad, instead to the exisiting H1. Since this is a highly trafficked page and the impact of the test could be huge, you might want to run your split test with only 60% of all visitors hitting the homepage from that ad. Here, you can set your hold out to 40%, meaning that 40% off all audience will be "held out" of the personalization, and only 60% of the all the traffic sees the personalized version while 40% sees the orginal homepage.

How do I create a holdout?

In the example above, you would start out by creating a new personalization campaign. This is how the screen looks -

To limit the test to only include 60% of traffic, you would need to set the holdout to 40% (since you are holding out 40% of the traffic from the personalization)

You are done! Publish your personalization campaign, and when you launch 40% (or whatever you specified) of your traffic will be held out from your personalized campaign.

How to run an A/B test for all site visitors

A/B experiments, in the context of Fibr, involves testing whether your personaliazed variation of a landing page is performing better in terms of engagement and conversions as compared to your original page.

Marketers usually leverage A/B experiments with Fibr to test out different messaging, images, CTAs to optimize user experience and enhance conversion rates by comparing different page variations and identifying the most effective elements for personalized content delivery.

Note: In general, for A/B experiments to be statistically significant, we recommened conducting them on high traffic pages, recommended bar to conduct meaningful experiments is if you have >250K website visitors.

Creating an A/B Experiment

To create an A/B experiment, simply create a holdout by following these steps. Set the hold out percantage to 50% for equal distribution of audience!

You are done! Publish your personalization campaign, and when you launch 50% (or whatever you specified) will go to the new personalized variation, and the rest 50% to the original experience.

Seeing results of A/B Experiments

You can check the results of the A/B experiment in the reports section.

For Example: You want to check the difference in CTA clicked after changing the CTA text as compared to the original one. You will have to set tracking CTA clicks as an event and Fibr will start tracking it.

Fibr includes a parameter "Conversion Lift" in reports for every campaign that has a holdout set.

What is Conversion Lift?

This number represents the lift in conversion rate for all visitors who have seen the personalized variant of the landing page as compared to the conversion rate for all visitors who have have seen the original page.


Don't be a stranger

Last updated