When running an ad, how many impressions do you guys feel is necessary to find the more or less "true CTR" of a banner?
Is it 5k? 10k? 20k? Even more?
I know this is kind of a vague question, but Im just sort of looking for a rule of thumb when calculating CPC.
Thanks
It's difficult to say as it depends on various factors. More is always better when talking about data samples.
It's important to make decisions using data with statistical significance.
How confident are you that the results were not caused by random chance? Make sure your sample size isn't too small. For example, 1 click with 100 impressions gives you a 1% CTR. If the 101st impression also clicks on your ad, then your CTR increases to 1.98%, almost double.
Therefore it's important to have enough data to be sure that the results haven't been significantly affected by one or two clicks.
It's also important to consider the time frame rather than just volume. For example don't start a test one day one and finish it on day two. I'd say at least a week of data is needed to take into account fluctuations during the day, and over weekdays and weekends too.
Cheers!
Evelyn - TrafficStars Team
I use a statistical significance calculator for this. You can use the same binomial confidence interval calculator you probably use for calculating conversion rates to calculate CVRs:
https://stmforum.com/forum/showthrea...nificance-quot
So, to take an example, if you have an ad with 2k impressions and 14 clickthroughs, just plug it into the calculator:

You'll see that the "true" CTR is between 0.38% and 1.17% .
That'll narrow down as you give it more data. So I just keep feeding it data until I have the accuracy I need!
(As Evelyn says, timeframe is also important. Always capture data over a couple of days.)
Around 8-10k it's usually stabilizing, at 20k the number should be pretty much final. It will always go up and down a bit, depending on the bid of your competition and time of the day or day of the week and it's also a bit different from one network to the next. In case of CPC networks, you need to ignore the initial burst of impressions in a way, because the algos usually prioritize new banners in order to test them quickly. Once that testing period is over, the CTR usually goes down.