This is an interesting discussion! There are loads of hidden questions:

Are we all putting links in the same place? Does it make a difference?

Do customers really care (or know) about the 'aff_id=NN' code in a URL, or are they more interested in reason they'd click the link? (i.e. There's something they do or don't want at the other end).

Is it different in the (more tech savvy) Internet Marketing 'niche'?

Were you masquerading as the author of the content at the other end, or presenting as a partner/advocate at the other end of the link?

And for those that suggest they've got some test results...were they statistically valid? What confidence level do you have in them? Do the results suggest a general priniciple, is the result so context-ladened that it's a mere story - a one off?

There are loads of wierd practises around marketing on the web:

Long, long headlines - are these really better? (Our tests suggest NOT) Or are they just something that's grown out of a wierd evolutionary process.

Long form landing pages - is it better to have one long page, or to get customers to click through to multiple medium length pages? (I've seen stats that suggest the latter is better, but again, perhaps it's a context thing).

Johnson boxes - is a yellow box really the best way to present a testimonial?

Order boxes - are dashed lines really more effective?

Hiding the price at the bottom of the page - or requiring a click through to the order page...

So many more!

But of course you can 'boil the ocean' with tests, and never reach your main objective: Selling stuff.

We've had a motto this year at my company: 'Momentum over Optimization' - and it seems to be powering our way to much better things. Reserve optimisation approaches to areas where you have enough scale for confidence in the results and that yield enough value to justify your investment in the test.

The other approach is to 'phone a friend' - or ask a question on a bulletin board, or forum, or similar. But you've got to get really smart about how you frame the question, and how you interpret the responses for it to be any better than guessing yourself!