Need help selecting a sample size for a case study.

[seaniaaa]

Hello all. I'm trying to put together a case study. I want to compare the conversion rates of a group of 1,134 businesses based on one binary variable. It would be very time intensive to comb through the entire population, but I need the math to be solid enough to convince businesses of the value of that binary variable. I've been out of college for far too long and haven't run an analysis like this in ages. The internet left me far more confused than when I started. Any guidance would be appreciated.

 

[Phil Rozek]

@seaniaaa, it probably depends on how much time goes into sniffing out each of those businesses, to record whatever variable(s) you're trying to record. I've done a lot of those case-studies for my blog over the years, and the SOPs generally are as follows:

1. Pick a pool size that's big enough that I think, "Oh [BLEEP], this'll be a lot of work," but that's smaller than the maximum. I've found 500 businesses is a good size, and I typically start in that neighborhood.

2. "Do a few myself": I'll see how long it takes me personally to sniff out 20, 50, or maybe 100 of the businesses, just to see how long the process takes, how easy it is to record this or that, how many outliers are in there, etc.

3. Usually I have a helper, so at this point I can assign most of the rest of the heavy lifting to him or her, based on clear SOPs I've worked out (see point #2).

It should be a lot of work, in my view. More perspiration than inspiration. I think that's what makes these things valuable to readers.