If you have full tagging, and have separated the campaigns out so that you have adequate targeting control (fairly tricky when comparing Google Pages and Google Search Partners, for example - but again, something that analysing the referer_info can illuminate) then you can make better decisions about which target network should receive bid increases and decreases.

The problem with most web analytics packages is that extracting the proportion of misattributed data is external to the package, and hard to put back into the analysis. If your analytics package makes you have to rework your numbers, for a measurable cause, surely there’s a question you should be asking your analytics provider? :)

We have a couple of in-house tracking programs–one is a custom analytics program and the other is tag tracking in emailed referrals. (We have GA, but I rarely dig into it.) When I started trying to match up email tags against our in-house data submission logs, I realized, for instance, how randomly some people fill out the fields in our online forms. Now I need to dig into Google Analytics and try to set something up to provide cleaner data.

I find I’m wondering if Jeremy is suggesting some kind of Analytics filter in his last paragraph? I’m not sure how else to track bounce rates or ROI by tracking parameters.

The “referer” header is useful, but what would be seriously helpful would be to (optionally?) re-encode origin data in the request. Unlikely to happen for privacy reasons. There are good reasons to be able to with hold the referrer.

If anything, this shows how frustrating the non-webmail mail user interfaces are. I’ve plenty of new clients for whom we can correlate an increase in direct traffic to email campaigns - clients who thought their emails weren’t working, or that perhaps the email triggered people to use a bookmark. Show them how a click from a non-web email has no referrer, and that the URL can encode the origin, and they get useful measurable data.

And it would be helpful for the Web Analytics companies to identify, by default, streams of clicks from webmail. The servers for many of the services are quite identifiable in the referers, but rarely shown as a class of origin. Not exactly the question you asked, though ;)

To make the data more valuable, break up campaigns and mark them separately, so you have different tracking parameters accessible for each of the classes of origin (Google Pages, Google Partner Searches, Google Content Network, Placement/Site Targeted) etc. That way you can better track bounce rates, ROI, etc, by controllable targeting options.

What would be the ideal situation, if one is able to control everything downstream of the PPC click. I.e. what do you recommend for the technically capable to make the tracking more accurate?