Blue Nile has a handy way of answering this question for its diamonds with this chart:

Users can initiate a search for similar diamonds directly from the chart.

In the page above, they've displayed all the values for the first facet (Main Ingredient), which pushes the remaining facets (Meal/Course, Type of Dish, Season/Occasion, Recipe Category, Preparation, Dietary Consideration) down to about 2800 pixels "below the fold."  That's a long way to scroll--and of course it significantly reduces the discoverability of these options.

How could Epicurious have avoided this? One option would be to initially present all the facets closed. This has the disadvantage of requiring two clicks to select a "refine by" option (one click for the facet, one click for the value).

Another option would be to initially display a facet that has relatively few values. My choice here would be Meal/Course:

Another approach is to use more/less links, as shown in this example from Buzzillions:

The top brands are shown initially, with the option to see "more":

One click does it

The fastest and easiest opt out is the "one click" experience. Just click the "unsubscribe" at the bottom of your email and you're done.

Simple single-click.

Single-click with link to re-subscribe.

Two clicks

Several sites used a two-click  "are you sure you want to do this?" approach.  The example below has clear primary action button that is labeled specifically "Unsubscribe."

Three clicks

Slightly more arduous is the requirement to express your preference using a check-box or radio button.

I suspect the example below has a high error rate, with people overlooking the check box before submitting:

The example below, while still requiring a total of three clicks to unsubscribe, probably has a low error rate because the radio buttons are left-aligned with the submit button, making them more discoverable.

Consider the options

A few sites took the "please consider the alternatives" approach to opting out. It requires a bit more thinking (or at least reading through the options.)

Find it yourself

The most frustrating experiences were companies that did not provide an "unsubscribe" link in their email. Instead, they instructed me to log into my account and edit my "communications preferences." In some cases, this was easy to find; in others, not so much.

A certain reunion site makes you hunt down your email preferences, which once found, are presented in a somewhat dizzying array of options.

After de-selecting all the options I was puzzled as to why I was still receiving emails. Alas, the small print:

It can take up to 10 days to process new or canceled subscriptions.
You will occasionally receive emails that tell you about updates to your account or announce [site] improvements...

Some buttons on this page are obvious, for example "Click to Talk." What is less obvious is the function of dimensional-looking yellow bar across the middle of the page. The first set of text ("Looking for more than a ballpark summary?") is not linked but the second set of text ("Contact us now") is.

While dimensional effects can add visual interest, non-clickable "raised" elements have the potential to be confused with buttons. We've seen this in testing where participants try clicking on anything with a beveled or drop-shadow treatment.

On Foodnetwork.com the selection control is contained within the search field:

I suspect the location in this position is more discoverable because it places the control between the typical starting point for a search task (the search field) and the end point (clicking the search field).

• Dynamic filtering updates results automatically as a user makes each filtering selection.
• User-initiated filtering where the user makes some selections and clicks an update button to see new results.

Which approach to use often comes down to data and performance issues (sometimes dynamic filtering isn't feasible or advisable from a technical standpoint, particularly where users are likely to select multiple filtering criteria).

Endless Shoes and Handbags is an example of dynamic filtering; Hotels.com employs user-initiated filtering.

Interestingly, Tripbase gives the user a choice. By default, the site uses dynamic filtering. But users can override that default by de-selecting the "auto-submit" option.

Rather than using the conditional logic to create a separate error message for each case, PayPal  makes the user wade through the IF statements to determine which case applies. This makes it particularly difficult to find the resolution. Errors are never a happy circumstance for users; by making error recovery more difficult than it needs be, the aggravation is compounded.

I would be interested to see eye-tracking data to understand  how quickly users notice this type of display. The New York  Times uses a drop-shadow treatment at the bottom of the bar, which visually brings the top bar  forward on the page. Twitter uses a slight embossing treatment and has plenty of empty space around the bar. Both treatments probably aid discoverability.

Twitter uses what I think is a subtle but effective display to indicate when the "Election 2008" bar is minimized:

After clicking the "No thanks" link on the New York Times top bar, I couldn't find a way to re-display it--and wasn't sure where else I could look to if I was interested learning more at a later time.

• a term's current rank
• whether its popularity is on the rise or on the wane
• its rank last week
• the total number of weeks it has been in the top 50 list

The table format is an effective way to communicate differing aspects of popularity (current, trend up or down, change from previous week, and popularity endurance).