The Truth About Asking too Many Questions
This is a great article by Dev Bhatia at ClickZ on the New York Times' attempt to collect more data from its online registrants. I've always heard and intunitively known the more questions, and more personal the questions, one asks (no matter how valuable their brand is, i.e. NYT) it doesn't make much web sense.

From the article:
Instead of asking the registrants for the answers to these questions, what if The Times had gone to a third-party data bureau and simply appended this data to its files? Demographic data of this sort is a commodity and can be had for pennies per record. Had The Times done this, it would likely have gotten more accurate data, because appended data is much more likely to be accurate than online-form data, which is notoriously unreliable. So, it could have expected to raise the price of future advertising even higher. In addition, The Times would not have suffered from the drop-off in conversion rates, which occurs when too many questions are asked.

In economic terms, The Times could expect to make more advertising dollars in the future by measuring its cost as the cost of appended data rather than that of lost registrants. It would be creating revenues without harming its asset. Clearly, The Times's online enterprise is valued by its total number of subscribers.