|Bad inbound links CAN damage you|
But only if you don't pay attention
I think we can agree that developing large numbers of low quality links develops largish numbers of low quality visitors. This can make a mockery of some pay per click calculations. Low quality visitors will have a significant affect on your perceived ROI. If site A gets twice as much traffic as site B, but only 10% more revenue, then they will erronously make a judgement about the average value of a visitor.
This value per visitor can pervade through your strategy, limiting your PPC bids, determining your banner ad strategy (what are they then?) and all other campaigns. If you haven't got your act together, the guy with half the traffic can easily start beating the other guy in terms of profits, by simply bidding on PPC correctly.
The trick, of course, is to then measure ROI by campaign or by search phrase, but this breaks down in the extreme. When you run hundreds or even thousands of search terms on PPC, breaking them all the way down to measuring ROI by search term is unrealistic, as the majority will show nil ROI on perhaps a dozen clickthroughs.
So - where should we go? how many visitors in a campaign of any sort is needed to have a statistically accurate sample size? and how do you segment your visitors into campaigns that allow you to work out proper ROI's?
Good points, agreed. What you need to do, imho, is to track *everything* and then analyze the macro trends that impact the biggest portion of your ad spend first, and then go down the list, till the analysis will take too much time to justify either positive or negative ROI.
At that point, if you are automating some aspects of the process, then the analysis will not cost anything in terms of real dollars spent (as it will be part of a package or what not) however, if it is something done by hand, then you need to be careful with your time versus your return.
It's a matter of ROI & dependent on your adspend analyze the 'big ones' first, and then go down the food chain.
As for the 'statistically significant' bit of your post, I would say that you need to determine your desired CPA or CPO before you start doing any sort of ROI analysis - because if you don't assign a value (even if you adjust it later) then you have no idea if it was 'positive' or 'negative' ROI.