Randomization

Each respondent is given a set of random numbers, known as a “seed”. Seeds work via hidden fields. Create a hidden field named seed_N, where N is replaced with the number of arms you wish to randomize. For example: seed_2, seed_3, seed_4, seed_5,…, seed_100.

This hidden field will have the assignment of each user, which will be an integer between 1 and N. For example, if you made a hidden field called seed_3, each user will have a value of that field equal to 1, 2, or 3.

Now use the hidden field in your logic jumps. If, for example, you create a hidden field called seed_3, then create logic jumps such that:

if seed_3 == 1 do A, if seed_3 == 2 do B, if seed_3 == 3 do C.

Say you want to randomize twice in your survey, independently, what can you do?

You can create multiple seeds, with the same bucket size, by using the format seed_N_V where N is replaced with the number of arms, as before, but V is used to represent a different version. For example: seed_3, seed_3_1, and seed_3_2 will ensure that you can have three distinct random arms, each of which is independent from the other.

Random seeds can be tested just like any other hidden field, using the following format (testing seed_2):

https://m.me/YOURPAGEID?ref=form.YOURSHORTCODE.seed_2.YOURSEED

In this case, you should replace YOURSEED with either 1 or 2.

Every user is given a long integer which is their random seed (i.e. 2843167128). A calculation is done on this integer to determine their seed_N or seed_N_V. In order to analyze the data, you will need to recreate this calculation.

The formula for a basic seed_N is seed_N = seed % N + 1

For example:

1. seed_2 = seed % 2 + 1
2. seed_3 = seed % 3 + 1

For multiple seeds, it gets a bit more complicated. In particular, you will need to first hash the integer V times using Farmhash 32-bit fingerprint, and then you apply the previous formula.