As Karon said, tracking and evaluating the tracking data is the only way to tell whether any promotional/advertising program is benefiting you.
As she indicated, many folks use "tracking URLs" or they add a little bit of unique code to the "destination URL" in their PPC program(s) to be able to figure out which visitors were driven to the site by PPC ads.
Here's (a simulation of) how we do it:
In the examples, you can see both "tracking URLs" ('fudge.html' for all fudge-related PPC search terms and 'nuts.html' for all nut-related terms) and "source IDs" (GGL=came from Google, FDG=fudge term, CAR=caramel).
Using this combination, you can examine your log files, or use a program to do the same, and quickly see (a) how many people came to your site as a result of a topical search (fudge,nuts) and (b) exactly which term they searched for to find your ad (caramel fudge,basic nuts).
Now that you can sort your visitors out, your next step is to figure out if they purchased anything while visiting, or whether the visit led to a potential sales contact of some kind.
Most tracking/statistics programs use the visitor's IP address to sort out the complete visit "path". It's not perfect, but it gives you a good idea of where these visitors went and how long they stayed on any particular page. There are much more precise means of identifying a visitor, but that can get pretty technical and code-y.
Once you can see the path each PPC visitor took through your site, you can see (a) how many of them went to your sales channels (shopping cart, contact form, etc.), (b) and how many completed the sales channel (by requesting your "thanks for the contact/sale" page).
You can also see which pages most people "landed" on, and which pages most people "exited" your site from. Very valuable information. (If everyone is landing on your index page, and 80% of the visitors are exiting from your index page ... then there's some work to do on the index page to bring that exit number down!)
Just like any advertising program, you are looking to determine your Return On Investment (ROI), where you add up how much you spent on PPC to make so many sales from PPC visitors. Spending more than you're taking in? Not a good ROI. Taking in 300% more than you're spending from PPC visitors? Good ROI (depending on your overhead).
So, build your PPC program to include tracking URLs and source IDs, learn to or grab one of the many free programs that analyse your log files, investigate the analysis, check your income, adjust your PPC spend to suit the result.
There are lots of factors in what drives a visitor to produce, including ad copy, page content, page layout, color choices, and even the position of your PPC ad in the results display. Tinker with your formula by making ONE change to ONE element (as suggested by Karon), so when you see a changed result you can more easily pinpoint WHY the result changed.
Also keep in mind that web traffic is not a constant stream. It is more like a series of waves. There will always be the deepest valley and the highest peak in your traffic rate, so try to find a range of traffic rate that you feel is a reasonably accurate expression of how your changes are affecting your traffic, and then go for it. It's only a little bit scientific.
It's fun AND crazy-making!