How to run multiple projects at the same time without collision

If you want to run multiple A/B tests that you do not want to affect each other, you need to set up Collision prevention statements in Audience for these projects.

You need to do this before you start any of the projects.

 

Running two projects with the same Audience at the same time without collision

Let's say you are going to run two projects at the same time. You want these to get 50% of the traffic each. If these two are on the same Audience it is a fairly easy setup;

As projects are evaluated in the creation order, you need to set the traffic allocation to less than 100% total in the first project to be able to send some of the traffic to the other project. In this example we will set it to 50% total to get an even split for the two tests.

Notice that for the second project you want the total traffic allocation to be 100% because you want 100% of the visitors that aren't allocated the first project to be allocated to the second project.

 

mceclip0.png

First, open the editor for project 1 (created first of the two). Then click on the Settings button. Choose Traffic allocation.

  

mceclip4.png

Then set the traffic allocation for the variants. If you wish the two projects to have equal traffic you want this setting to be a total of 50% in the first project. So set the variations to get 25% each.

 

mceclip2.png

Then open Audience settings and add a Collision prevention statement.

 

mceclip3.png

Now select the project you wish to set up collision prevention against in the dropdown. In this example that is the one named test 2. Hit Apply & Close to confirm this and then Save your Audience. Ok, now the setup for the first project (named test 1) is done.

 

Now we need to set up collision prevention for the second project (named test 2 in this example). We do that in the same way but here we choose the first project (test 1) in the collision prevention statement dropdown and set this project's total allocation to 100% (50% for each variation). This will allow the project to get all visitors that aren't allocated the first project.

Example allocation with the two projects that has two variations each;
Test 1: 25/25 (50% total)

Test 2: 50/50 (100% total)

 

Running two projects with different Audiences at the same time without collision

If you want to run two projects at the same time that does not have the same Audience it's a bit more tricky. You need to figure out which test is going to have the most visitors. In other words, which project is more likely to have visitors fulfilling the Audience first.

Let's say you will be running a Start page test and a Product Detail Page (PDP) test at the same time. If you do a lot of product specific advertising it might be more likely the visitors will land on/visit a PDP before the start page. This will leave less visitors left to the start page (if the PDP test is set to 100% of total allocation)

If you do not have a lot of product specific advertising the visitors might be more likely to enter your website on your start page, which would leave less visitors to the PDP test.

With this in mind you need to decide which Audience is more likely to be fulfilled first, the Start page or the PDP.

When this is decided you will want to set a lower total traffic allocation on the project that you believe most of your visitors will be getting first.

Example:

If you have a PDP driven website and most of your visitors land on (or at least navigate to) a PDP first, you should set the total allocation for the PDP test to less than 100%. This way the visitors that aren't allocated the PDP test have a chance to get the start page test (if they navigate of land on that page). The start page test should then be set to a total allocation of 100%.

Example allocation with two projects that has two variations each;
PDP test: 10/10 (20% total)

Start page test: 50/50 (100% total)

Notice that this is a rough estimate and you will need to set it according to your calculations.

 

 

 

 

 

Was this article helpful?
0 out of 0 found this helpful