If the number of impressions is less than 5, it is not possible to accurately evaluate the click-through rate.
Strictly speaking, to make this judgment, a statistical phone number database uk hypothesis test to determine whether or not the result is significant is required, but this is a complicated procedure and varies depending on the conditions, so we will omit it here.
It is generally said that about 400 samples are enough to obtain reasonably reliable data.
Reason for inaccuracy #3: Not randomized
For example, if all users exposed to ad A had previously seen ad B, and all users exposed to ad B had never seen either ad A or B, the test would likely not be accurate.
As you can see, web advertising has its own unique characteristics, and there is a possibility that it may not comply with Fisher's Three Laws. How can you avoid such problems?
Solution 1: Use an AB testing tool
This is the easiest solution. Many "AB testing tools" have features to meet "Fisher's Three Laws."
Specifically, these features include "exposing users completely randomly," "notifying users when statistically significant differences emerge," and "cleansing data by taking into account past advertising exposure."
Solution 2: Set your ad rotation to "Don't optimize"
Google Ads allows you to disable the setting that displays the most suitable ads to users. This avoids "Possibility of inaccuracy 1: lack of local control" . However, it may affect the delivery results of your ads.
Solution 3: Test for a longer period
By taking a longer period of time, you can avoid possibility 2 of inaccuracy: insufficient repetition .
