You can use "random sample" filters in flow steps. you will need 20 out of 1200 samples, which are 1.66% of the total. the hard point is to make sure that leads are not selected twice a couple of weeks apart. The trick is to create lists to store the samples.
To elaborate more on this:
- Create as many static lists as you need to store many weeks in advance
- use a batch smart campaign with a "add to list". In the add to list, add choices with the "ramdom sample" filter and set the value to 1 or 2 % (you cannot set it to 1.67%, only integers are allowed, AFAIK). setting it to 1% will select 12 leads, setting it to 2% will select 24 leads.
You also can to it with 2 static lists : one with the week sample, one with the previous weeks samples. You will have to empty the first one every week. you run the batch campaign every week, it selects some leads and add the to the weekly sample and the "all time sample". In this case, the trick is to use the "all times samples" as an exclusion list for the batch, so that anyone who has been select a previous week cannot be selected again. The point is that, as week after week, the source list will decrease in size, unless you have new leads flowing in. so you will have to come back and increase the random %