2. WHAT IS A/B TESTING
The A/B test allows us to decide more
scientifically between any two strategies.
A company has changed the colors of the
button on its website. Accordingly, is there any
change in the time that visitors spend on the
site? Is there a difference between the
previous form and the current one? Based on
the output, the company makes the necessary
decisions.
3. HOW CAN WE DECIDE?
How do we evaluate the data as a result of the
change?
In the figure, the majority looks like Group B. But
for its accuracy, we must use statistical methods.
4. HYPOTHESIS TESTING
• H0 : There is a statistically significant difference between the change in
the site and the time users visit, if, p value < 0.05 H0 is rejected.
• H1 : There no statistically significant difference between the change in
the site and the time users visit, ifi p value > 0.05 H0 can't be reject
5. EXAMPLE
Problem
• A company recently offered what it called maximum bidding
average bidding, a new bid type as an alternative to the export
type introduced.
• One of our customers, company X, has decided to test this new
feature and wants to do an A/B test to see if average bidding
brings more conversions than maximum bidding.
6. HYPOTHESIS
• H0 : There is no statistically significant difference in purchasing between the
control group offered in the maximum bid campaign and the test group
offered in the average bid campaign.
• H1 : There is a statistically significant difference in purchasing between the
control group offered in the maximum bid campaign and the test group
offered in the average bid campaign
• Control and Test Data ;
• Control Data represent to maximum bidding
• Test Data represent to average bidding
7. NORMALITY ASSUMPTIONS
• Is there a statistically significant difference in purchasing between the
control group offered in the maximum bid campaign and the test group
offered in the average bid campaign?
• Normality Assumptions – Shapiro Test
• H0 : The assumption of normal distribution is provided
• H1 : The assumptions of normal distribution is NOT provided
10. RESULT
• First, we examined the normality assumption and variance homogeneity
in the sample. Since the assumption of normality and variance
homogeneity was provided, the T test was applied. As a result, there is
no statistically significant difference in purchasing between the control
group presented in the maximum bid campaign and the test group
presented in the average bid campaign. It can be continued with the
current method. It can also be continued with a method that is cost-
effective.