Since you have the sale transaction times recorded, if the time of each click
was recorded (probably it is already) then very accurate keyword tracking could be done based on the correlation. (Assuming sale and click times are recorded accurately in real time).
All google (for example) has to do is include two time selection buttons to set the time range (like on the airline booking web sites) next to the existing day/week/month selector and every sale can be easily tracked to its originating keyword.
This would make a huge difference!
Don't you get reporting from your hosting?
you should clearly be able to see the referral and keyword used (and the time...there may a minute or 2 difference...) however unless you're taking huge amounts of traffic, you should be able to discern the info you're looking for...without too many gymnastics
Limitup: It works if you can see the clicks within (say) a 1 minute window. Unless you have a huge volume of clicks this will be enough to pinpoint the click that triggered the sale. Right now we only have a resolution of 1 day not 1 minute.
How do you plan on figuring out which visitor bought, based on the time they clicked your ad?
One click a minute is 1440 clicks a day in one ad group. I get about 1-100 clicks per ad group per day, an average of one click every 14 minutes or less. Should work fine.
Even if two clicks come close together and are conflated that's still way better than upto 100 clicks in a day that are conflated right now.
And what happens when someone bookmarks the site and buys 2 days later?
I could go on and on about why your proposed method will never work, but you don't seem to be interested in what I have to say so I won't bother. Good luck to you ...
If this is to complex or you don’t have the facilities you can always create multiple identical landing pages and then use the server stats to track the users that land on different pages and go onto make purchases.
