Email Answers: Does The “Rule of 30″ apply to impressions?
Cache (yeah that is his real name.. I think 
) wrote me an email:
Hey John – we are doing some ads and testing things. You seem to be a guru in this area, I wondered if you had an idea on this question.
I read a good article that stated statistically speaking 30 responses under the exact same conditions is accurate + or – 5%. This helps a lot to know statistically if we had the same keyword same add 30 people came to the landing page none converted then we try something else. I have another question and wasn’t sure if you knew the answer. I would like to know statistically how many impressions on a specific keyword/ad copy combo before it is unlikely to change. So if we don’t get 30 clicks how long do we wait before we change. I would think there is a similar statistic saying so many impressions under the same criteria means you will continue to get the same response. If we could know that then we know when to change things up so we aren’t moving to quick or too late. I am searching for an answer to this question if you can help that would be great.
To summarize:
If your landing page gets 30+ hits and no conversions you really need to change something (usually the landing page). I’d say that is fair.
As far as impressions go…. What I like to do is take my ads and keywords and split them apart in high CTR campaigns and low CTR campaigns. When I first start a campaign I just don’t know yet, but over time I get a good idea of CTR and move the higher CTR adgroups together in their own campaign.
If you are getting zero impressions but a high position then the keyword you have just doesn’t have much traffic. Leave it alone for a while.
If you are getting lots of impressions but a low ranking (say 10 or so) then you have a keyword that will work, but you really should fix your adcopy and/or raise your bids. Back to the original question….
So if we don’t get 30 clicks [on an ad] how long do we wait before we change.
My answer: You should be split testing with multiple ads (usually two). Then you use the “rule of 30″ for picking which ad you should replace with a new ad, then you wait for 30 more clicks on the ad and replace the ad with the lowest profit margin. Using this tactic over time will always yield good results over the long term.
PS: You should also be split testing your landing pages.
Update From My Comment:
“…if you are really looking for a target number. Try it this way.
1) What is your target CTR? Lets go with 1% (I see everything from .00% to 60+% CTR on ads). I would say take the average CTR across your similar ads/keywords as a baseline.
2) At a 1% CTR you would need 3000 impressions to get 30 clicks. Using the 33% rule take 1/3 of that and you get your answer. 1000 impressions and you know you probably have a dud.
3) A little Algebra and we have a formula for you.
Target impressions before it is likely a dud = Target CTR% / 10
Example 1:
Your CTR rate should be 1% so you if you get over 1000 impressions then you will never meet 1% CTR., Change something.
Example 2:
Your CTR rate should be 3% so you if you get over 333 impressions then you will never meet 3% CTR., Change something.
Example 3:
Your CTR rate should be .02% so you if you get over 50,000 impressions then you will never meet .02% CTR., Change something.”
Interesting. John is a genius at this stuff.
What if you have 2,000 impressions and still not 30 clicks. That is kinda the heart of my question, statistically is 500 impressions or 1,500 impressions a valid sample? I would think 2,000 is but I am curious is there an impression number mathematically that basically says your results won’t change much anymore from this point on – change something.
@Cache If you have 2000 impressions and zero clicks, I would wonder how many ads you have for that keyword. I am guessing your ad is not relevant to the people that are seeing it. Try adding multiple ads to the adgroup to see if any adcopy you throw at it will get clicks. But if you are really looking for a target number. Try it this way.
1) What is your target CTR? Lets go with 1% (I see everything from .00% to 60+% CTR on ads). I would say take the average CTR across your similar ads/keywords as a baseline.
2) At a 1% CTR you would need 3000 impressions to get 30 clicks. Using the 33% rule take 1/3 of that and you get your answer. 1000 impressions and you know you probably have a dud.
3) A little Algebra and we have a forumla for you.
Target impressions before it is likely a dud = Target CTR% / 10
Example 1:
Your CTR rate should be 1% so you if you get over 1000 impressions then you will never meet 1% CTR., Change something.
Example 2:
Your CTR rate should be 3% so you if you get over 333 impressions then you will never meet 3% CTR., Change something.
Example 3:
Your CTR rate should be .02% so you if you get over 50,000 impressions then you will never meet .02% CTR., Change something.
Hope that makes sense.