Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Randomizing Blocks


I have reviewed^ this but I want to check my understanding.

I want the same participants to complete two separate CBC exercises. I want to strive to counterbalance things so that some participants complete exercise A first and some complete exercise B first.

Can I do that by making blocks? I might be using this term "block" incorrectly.
But I want block 1 to be: [Exercise A Instructions Page, Exercise A CBC  tasks, Exercise A follow-up question page] and block 2 to be: [Exercise B Instructions Page, Exercise B CBC  tasks, Exercise B follow-up question page].

So do I create a new "set" with these two "blocks"? And the anchor questions are the instruction text and the follow-up quesitons? Would that accomplish what I'm going for?

asked Sep 20 by anonymous

1 Answer

+1 vote
You will have two blocks within your set.

Block 1: Exercise A
Block 2: Exercise B

Each block anchors will set ANCHOR 1 as their respective instruction page, and ANCHOR 2 will be set to the last respective follow-up question.

You can use a constructed list to randomise whether Exercise A is answered before Exercise B or vice-versa.

The constructed list will hold the information regarding the exercise display order. You can export this with your survey data and use it amongst your analysis.

May I suggest you run a very simple test where you have something like 3 questions for Exercise A and the same for Exercise B. They don't have to be conjoint questions, just something like an intro page and 2 select questions. Then set up your block and constructed list to randomise the blocks. Once you have this simple test working, you're up and running.

If having exactly the same number of respondents answering A-B and B-A is important, you may consider a least fill quota approach. If approximately the same will suffice, then the more simple randomising approach is the way to go.
answered Sep 20 by Paul Moon Platinum (94,725 points)
edited Sep 20 by Paul Moon
Thanks so much! Approximately the same will suffice. I will do the small-scale test, then add the CBCs.
Sounds like a neat plan!