UPDATE:
Just analyzed results of A/B test on Fiverr and thought this might be useful for someone.
Most sellers (if not all) on Fiverr never actually A/B test. They indeed TEST by changing parameters here and there and trying to gauge results thereafter. However, this approach is way too subjective when you deal with multiple factors like
1. seasonal fluctuations (e.g. buyers spend more money in the beginning of the month than in the end)
2. random fluctuations (e.g. competitor unpublished his/her gig resulting in your gig selling more during some limited period of time)
3. other kinds of randomness
The only objective way to A/B test is to have two more or less evenly developed accounts offering the same service in a similar (but not identical) fashion.
In this case you can measure revenue RATIO between accounts.
E.g. Account A is a "control group"
Account b is an "experimental group"
Month 1: measure data on both accounts
Month 2: implement a change on account B
then compare account A revenue / account B revenue week by week
also Account X period N+1 / N revenue for each week (this means account A week2/week1, week3/week2 etc, same for account B)
also Account X period N+30days / N (this mean first week of current month against first week of previous month etc)
After you have got all this data written on paper, it will be SUPER EASY to derive RIGHT conclusions from your A/B test. Other ways of testing are WASTE OF TIME, never call them "tests", they are just mere experience/intuition based optimization.
Likewise, there's one part specific to Fiverr and it is super easy to analyze. It is ratio of orders with extra(s) calculated as
R= number of orders with extra(s) / total number of orders
I guess you don't need explanation why orders with extras are important, do you?
E.g. for one gig I got the following ratios:
week 14 - 9/28 R=0.32
week 15 - 17/41 R=0.41
week 16 - 14/36 R= 0.38
You can clearly see that changes implemented before week 15 resulted in increase of orders with extra(s) while changes made before week 16 haven't had any significant impact. (By the way, applying changes on Sundays is a good idea if you want to conveniently analyze results later). On week 16 ratio seemingly stagnated. However, considering that I switched from outsourcing work to doing everything in-house (thus increasing own profit margins) right before week 16, ratio stagnation can be considered a good result.
P.S. after getting some historical data and personal experience, you will be able to run 4 A/B tests per month with only 2 seller accounts.
P.P.S. ah yea, if you don't know how to run multiple Fiverr accounts:
a) Download FREE version of MultiLoginApp (yes, it is sufficient for up to 10 Fiverr accounts)
http://multiloginapp.com
b) Register a couple cheap companies (choose the cheapest option with low-cost maintenance) and create a few PayPal accounts with them. Alternatively, use your sister's or mom's credentials for secondary Fiverr acc, it's all good
