The top task identification approach helps you separate the chocolates from the salads and discover what customers really want when they come to your website.

In traditional surveys and focus groups people will say they want salads, but when they're on their own on the Web they go for the chocolates. For the last nine years we've been testing methods to identify and separate the chocolates (top tasks) from the salads (tiny tasks).

It started with a card sorting study of a tourism website. On individual cards were tasks such as accommodation, vacation packages, planning a trip, top attractions, special offers, etc.

In workshops, I would go through a time-consuming process of getting people to put these cards into groups with the objective of helping this website design a more intuitive navigation. But the exercise was cumbersome and it was often hard to see clear results coming out of it.

I was looking for clear, repeatable trends and I wasn't getting them. Were there universal top tasks for tourism websites? I started experimenting. Instead of getting people to sort the cards into bunches, I started asking them to pick out the cards they thought contained the most important tasks.

Some trends began to emerge. Then I started limiting the number of cards they could choose to 10. Stronger trends emerged. Then I started giving people less time to choose. Even stronger trends emerged. Then I started getting people to vote. They had to choose the most important one, the next most important one, and so on. Very strong, clear trends emerged. The same choices rose to the top in New Zealand, Sweden, Switzerland, the United States, etc.

There were 146 cards to choose from. Again and again, people told me they couldn't possibly choose 10 out of 146, and certainly not in the time I was giving them. That's common sense, isn't it? Classic marketing research theory would say it's impossible. Classic marketing research theory is wrong.

It was amazing. After going through this process with over 1,000 people in 11 different countries, there was an extraordinary consistency of choice. Two tasks-'accommodation' and 'special offers'-got 21% of the vote. This needs to be placed in perspective. There were 146 cards to choose from. Two got as much of the vote as the bottom 108.

In order to allow people to record their votes for their top 10 choices I created a one page printout in order to allow them to vote. Certain participants started cheating. Instead of going through all the hassle of spreading the cards out and diligently sorting through them, they scanned the one page list and began to vote. I was annoyed. I told them they needed to go through the card sorting process because there was something special in this process that would get much better results.

People asked why. I didn't really have a good answer. So one day I decided to only hand out the list. To my great surprise, I found that the voting patterns were even stronger and clearer. The same top tasks of accommodation and special offers kept coming out in front. The process was much faster than card sorting and the trends were stronger.

[More from this author: Old Survey Methods are Broken]