Split test Traffic Distribution

Hi @adamwrethinkfirst

Thanks for your message. Currently its not possible to distribute traffic per split test by %.

Also, in a regular split test as well, it would also not show evenly. There are a few reasons split test variations won’t show evenly across:

1. The mathematics of randomness. The code isn’t going: “Show A. Show B. Show A. Show B.” Rather it’s based on a “random number generator.” Essentially the code randomly picks a number. With a vast collection of data, assuming every other variable is equal, then yes, the numbers should eventually close to 50/50. But there are other factors involved, such as #2 and #3 below.
2. Not every visit will meet the conditions to show your campaign; therefore, a particular split-test variant may not show as much as another variant and thereby skew your analytics results. If Campaign A shows immediately, whereas Campaign B shows after 20 seconds, your Campaign A will have a significantly higher impression number and will appear to “show more.” With that said, your Campaign B may have a better conversion rate compared to Campaign A.
3. You could have different start times. If the primary Campaign was started on January 1, but your Split- test didn’t start until January 15, then (assuming everything else being equal), your Primary Campaign should be showing nearly three times as often as your split-test variant when you review your January statistics.

To illustrate the variance of how much each?campaign split-test variant can show, check out this JSFiddle created by our lead developer that illustrates how the random generator will not give perfect 50/50 stats all the time but can fluctuate. Press Run Test?and look at the Distribution Numbers that show: https://jsfiddle.net/channeleaton/f65osrw5/4/show/.
(You can repeatedly press Run Test?to see the different possibilities).

I hope that helps.

Hi,

I hope you have everything you need on this. Please feel free to create a ticket if you need any other assistance.