Clicks - How many to be statistically significant?

Status
Not open for further replies.
sengseng

sengseng

New member
Jul 14, 2008
17
0
0
Chi
What's everyone's 'safe range' where your conversion data on your existing PPC campaigns is significant enough to make a decision on, but not too high where you could be throwing money away when you should have killed it. (ex. 300-800 clicks)

I ask because I do not want to err on waiting too long or killing a campaign too soon. Does anyone actually run statistical analysis on the minimum sample size?

Thanks all. Just doing my due diligence before I get ready to give AM the ol college try.
 


I usually do $500 / avg CPC.

So if I am paying an average of $1 per click after day one I'll determine the conversion data at 500 clicks. (I spread the budget out at 3 days)

What keyword matching type are you using when you begin? Broad?
 
productionhead said:
I usually do $500 / avg CPC.

So if I am paying an average of $1 per click after day one I'll determine the conversion data at 500 clicks. (I spread the budget out at 3 days)

What keyword matching type are you using when you begin? Broad?
Click to expand...

so you test the offer with 500 at $1 a click. Is this so you can get clicks quickly or is your cpc determined by your CPA. Would you do $1 cpc for a $10 lead offer? i'm assuming that once you determine if it converts, you will lower the bids right. When I test, my cpc is around .5 but I can never get enough clicks before my campaign dies out.
 
What's everyone's 'safe range' where your conversion data on your existing PPC campaigns is significant enough to make a decision on, but not too high where you could be throwing money away when you should have killed it. (ex. 300-800 clicks)

I ask because I do not want to err on waiting too long or killing a campaign too soon. Does anyone actually run statistical analysis on the minimum sample size?

Thanks all. Just doing my due diligence before I get ready to give AM the ol college try.
Click to expand...
I usually try to go for a 80-95% certainty (statistical phrase). If after I reach the 80 or 95% certainty and I'm "barely" in the negative, I'll let it run longer.

So if I am paying an average of $1 per click after day one I'll determine the conversion data at 500 clicks. (I spread the budget out at 3 days)
Click to expand...
You should spread the campaign out at least a week--you need to study the differences in the weekends and hours of the day. If it's too expensive/risky, decrease your daily budget.

And, sengseng, get a payout increase. ;)
 
aim high said:
so you test the offer with 500 at $1 a click. Is this so you can get clicks quickly or is your cpc determined by your CPA. Would you do $1 cpc for a $10 lead offer? i'm assuming that once you determine if it converts, you will lower the bids right. When I test, my cpc is around .5 but I can never get enough clicks before my campaign dies out.
Click to expand...

That was just an example but I usually set the cpc to achieve positions around 2-5 to get clicks quickly (this is usually for competitive niches, when im doing PPC with no advertisers or 1-3 i always go for #1). my initial cpc isn't determined by my CPA, i optimize to profitability after i get some data on conversion rates etc. if its a shit campaign that never happens.

and yeah I would do $1 cpc on a $10 lead offer .. if i can bring it to profitability ;}


uguxseo

yeah i'll try different time spreads, thanks
 
Statistically signifcant is a tricky thing to achieve as it completely depends on what you are looking to measure.

Here is one trick you can use to achieve statistical confidence but hard to do in your case:

ABC or ABA Split Testing

The idea is that you set up three tests:

Test 1) Control
Test 2) Experimental
Test 3) 2nd Version of Control

Basically Test 1 and Test 3 are exactly the same. After to send some basic traffic, you are basically waiting until Test 1 and Test 3 have about the same stats which they should because they are exactly the same. When you do, then you look at Test 2 to see how it compares.
 
As for your example, I would recommend the following:

1) Get some general conversion stats for the type of traffic you are running from affiliate manager/advertiser. Make sure you can trust the person and that they know what they are talking about.

For example: You are running PPC traffic through a gateway page. Your affiliate manager says they are seeing 5% conversion rate from this type of traffic

2) Run up to the following forumla (5-20 divided by conversion rate) before you see your first conversion. If you don't see a conversion by then, then it is time to retool your campaign. If you do see conversions but doesn't hit close to breakeven or profitability, you might think about optimizing campaign again.

In our example: I would run 100 clicks (5/.05) to 400 clicks (20/.05=200) before evaluating campaign. Depending on my financial situation and stats, I would retool if needed.

Of course these are very rough rules of thumb. You have to see how the specifics of the campaign look like and how you would optimize or test other tests. Days of the week matter. Keywords matter. Position matters. So on.
 
productionhead said:
I usually do $500 / avg CPC.

So if I am paying an average of $1 per click after day one I'll determine the conversion data at 500 clicks. (I spread the budget out at 3 days)

What keyword matching type are you using when you begin? Broad?
Click to expand...

I haven't actually started yet. Just doing preliminary research before I get involved with affiliate programs. I'll finally have the time to dive into it in a couple weeks, so I want to hit the track running and make some damn money.

If I am paying for clicks though, I am going to want as targeted a user as possible so there is a better shot at converting that person, so I'm looking at phrase matching as a starting point.

Thanks for all the responses. I am planning on running some statistical analysis to determine the optimal sample sizes of clicks. (putting college education to work for once.) I'll put together a template once I get some experience working with it and post it for everyone to use.
 
Last edited:
aim high said:
i'm assuming that once you determine if it converts, you will lower the bids right. When I test, my cpc is around .5 but I can never get enough clicks before my campaign dies out.
Click to expand...

.50 CPC for testing is too small for most niches, IMO. i doubt you'll ever get better than position 8 (even first page) on that amount unless the niche is virgin territory. :) your campaigns probably die out because of that.
 
NoelKuhl said:
.50 CPC for testing is too small for most niches, IMO. i doubt you'll ever get better than position 8 (even first page) on that amount unless the niche is virgin territory. :) your campaigns probably die out because of that.
Click to expand...

hm. i guess you're right. my campaigns always die out, then i get sad, lose faith. its a vicious cycle. So you would recommend testing with higher bids just to get the clicks and a set of data?

productionhead said:
That was just an example but I usually set the cpc to achieve positions around 2-5 to get clicks quickly (this is usually for competitive niches, when im doing PPC with no advertisers or 1-3 i always go for #1). my initial cpc isn't determined by my CPA, i optimize to profitability after i get some data on conversion rates etc. if its a shit campaign that never happens.

and yeah I would do $1 cpc on a $10 lead offer .. if i can bring it to profitability ;}

do you run landing pages on your tests? or does it depend on the offer.

uguxseo

yeah i'll try different time spreads, thanks
Click to expand...
 
aim high said:
hm. i guess you're right. my campaigns always die out, then i get sad, lose faith. its a vicious cycle. So you would recommend testing with higher bids just to get the clicks and a set of data?
Click to expand...
Perhaps you're better off chasing the longtail in the beginning. Cheaper clicks, and a good foundation for a larger campaign.
After you're sure the offer itself converts and you fine tune your LP, you can expand.
God. Can't believe I'm giving advice out for free to a UofM fan.
 
xmcp123 said:
Perhaps you're better off chasing the longtail in the beginning. Cheaper clicks, and a good foundation for a larger campaign.
After you're sure the offer itself converts and you fine tune your LP, you can expand.
God. Can't believe I'm giving advice out for free to a UofM fan.
Click to expand...

when i go long tail, i never get any clicks =\ .Especially not enough to have a good statistics pool.
 
aim high said:
So you would recommend testing with higher bids just to get the clicks and a set of data?
Click to expand...

Definitely do if you can bankroll it. You'll be surprised by the difference. I work on a tight budget as well so I mostly test 1 campaign at a time but bid high enough to get decent positions. I do cut my losses pretty quick though to compensate - like, no conversions on a keyword after 20 clicks, kw's gone.

I'm sure I lose plenty of opportunities that way but that's pretty much how working on limited funds work for me. I don't favor longtails either, only main kw's to test. I look for longtails afterwards.
 
NoelKuhl said:
Definitely do if you can bankroll it. You'll be surprised by the difference. I work on a tight budget as well so I mostly test 1 campaign at a time but bid high enough to get decent positions. I do cut my losses pretty quick though to compensate - like, no conversions on a keyword after 20 clicks, kw's gone.

I'm sure I lose plenty of opportunities that way but that's pretty much how working on limited funds work for me. I don't favor longtails either, only main kw's to test. I look for longtails afterwards.
Click to expand...

This isn't gambling or drug dealing. You don't need a bankroll. You just need decent credit. If you are paying cash for your ppc bills, you should really consider the benefits of doing it on credit. Only spend what you can afford to pay back, and get weekly payments to make sure you don't get screwed by your network.
 
Status
Not open for further replies.
Top Bottom
Back