Overpromise and Underdeliver
Appetite for Destruction

Data, Delight, and User Experience Design


Many thanks to my friend Ric Ewing, who was a co-author of this piece.

Pick a card from the deck that I've fanned out before you. Memorize the card, then place it on top of the deck. I'll let you cut it a few times, before I reveal that it's… the Six of Spades!

What's the secret behind this magic trick? If I told you, then it wouldn't be a trick anymore—and you wouldn't have been delighted in the process.

As a magician, I can't reveal my methods. But as a UX designer, I have an obligation to point out that the trick isn't really magic. "Not even the best magician in the world can produce a rabbit out of a hat if there is not already a rabbit in the hat," says a character in the movie The Red Shoes. UX designers are ever seeking to identify, define, refine, and otherwise clarify what people expect to see and do when they visit a website or use a web application. The best sites fulfill these core user expectations, but the very best sites deliver an extra layer of "magic" that transcends those expectations through how data is surfaced through their site experience.

So how do you make sites that provide this extra bit of magic? Here's a few common-sense notions on the subject.


I was sure the Ace of Spades was your card…

Any system operation that feels like magic must provide a meaningful pattern to a user over time.

Users are great at finding patterns in any kind of system behavior. As Chuck Palahniuk said: "What we call chaos is just patterns that we haven't recognized. What we call random is just patterns that we can't decipher."

Matthew E. May, in his fascinating book In Pursuit of Elegance, illustrates this point beautifully with regard to how this winnowing effect occurs:

"Suppose I tell you that a sequence of letters has been written on a slip of paper, that the sequence has exactly three occurrences of the letter y… if I give you no more clues, there's no reliable way for you to determine the sequence… But now suppose that I give you two further hints. First the hidden sequence of letters is an actual word in the English language; second, it contains the smallest number of letters possible having exactly three ys."

Per Matthew's thought experiment, the only possible solution becomes the word syzygy—after he's constrained the provided pattern from an infinite possibility of made-up words to the finite probability of a set amount of data. As humans, we are hardwired to detect these patterns and exceptions to them—we literally sniff the scent of the information out over time.

And the best magic tricks function in a similar manner: moving from a set of 52 possible cards to a result that the audience finds meaningful. Magicians regularly manipulate this effect in card tricks by intentionally showing cards to the audience that aren't a direct match. Then the third card is revealed in a dramatic flourish—pinned to the ceiling, stowed in a manila envelope underneath Seat 12 in Aisle F, and so forth. The audience knows that, based on having observed a magic show before, the chances are high that the correct card will be revealed in due course. There's nothing random about this activity—unless the magician fails to reveal the correct card, over and over again.

You must determine the pattern for the user in advance. Chances are that clicking a button or initiating an interaction that is meant to be fun needs to have repeatable, consistent results. Users may initially feel that something on the screen before them is random (meaning: unexpected), but based on the provided information they will always establish patterns to explain the behavior. And these patterns emerge from every design decision that you make, whether you intend them or not.


I will now make this coin disappear…

The more context you withhold from a user, the greater the perception of gain on their part if they receive meaningful information in response.

The magician is holding a silver dollar. He waves his magic wand at his left fist—and now it's empty. He opens his hand, and a silver dollar appears out of nowhere. Cue applause.

Actually, that isn't what happened. This classic sleight-of-hand trick relies on misdirection for its success. If we were staring at his wand while the coin was being palmed, we would have seen the mechanism that made the trick possible.

Liken this to search engines. We regularly and willingly fill in a search query and hit "I'm Feeling Lucky" on Google, having no idea exactly what page will mysteriously appear in our browser window. This is a direct manifestation of Google's belief that their results are so good, the first hit will be the best result for your search.

This doesn't mean that the result returned on your search will be directly meaningful to you. But chances are they're more likely to be useful. This is the pattern that users come to expect when using the feature. There's a kind of magic in the yielded result—even if Google doesn't nail it the first time. The Google search box is correct often enough that it creates the correct perception of gain.

Contrast the Google experience with the "Random article" feature on Wikipedia.org. There is a pattern here as well: Users know they'll be directed to an article created and edited by site visitors. You'll click the provided link and load a page regarding "Democratic Party of Japan" or "Hummingbird" or one of three million articles across a site whose overall volume of content cannot otherwise be summarized without making your head explode. Clicking on this link is the only way you could ever randomly access a piece of content on Wikipedia without first thinking of a search term or clicking on a keyword that points to a more relevant article.

There is much less magic in this feature than in Google's feature. For Wikipedia, the lack of (direct) utility is intentional. Their "random article" feature fosters discovery and learning, much like throwing a dart at a three-story tall map of the United States. No matter where the dart lands, you'd find something of passing interest—as each article was written and edited by people who felt said article was important enough to include in the encyclopedia.

Companies like Google are so interested in fostering this kind of delight that they're willing to lose millions of dollars in advertising revenue simply to create this impression of magic. And other websites take this tack as well. Look at Flickr's "Interestingness" feature:

"The photos you see here are a random selection of some of the interesting things discovered on Flickr within the last 7 days. If you click the RELOAD! button you'll get another set of random sprinkles."

Sorry, Flickr—the provided photo results aren't random. Per Flickr's proprietary algorithm, the user is being presented with a range of images and associated metadata that matches search criteria that's invisible to the user. By all appearances, however, it's random to the user—meaning that for any user that loads this page or conducts a search filtered by "interestingness" across the site, there is no reason for it to appear other than the images having been of interest to others.

This feature is just as much of a black box as Wikipedia's "Random article" or Google's "I'm Feeling Lucky." Much like those two other sites, the chances are that these images will entice you to click and view more information about the photograph, the photographer, associated tags, groups, people, and so forth.


Will someone from the audience please step forward…

In all the cases that I've provided so far, the most desirable result—on the part of both the user experience designer and business stakeholder—is that the user continues to explore content by walking through various tags and/or related links to more content of interest. We are more emotional than rational in our behavior, and these features cater more to our whim than our desire to fulfill a task-based activity. As a result, the website feels capable of fulfilling our inarticulate desires. We're not sure what to expect, except that it will feel magical.

Making any elements of your user experience contingent on "magic" requires you to provide stringently controlled contextual information around said activity. At each place in the "world" of your website, users need to expect the result of what will emerge—or at least be cued to its impact on their browsing experience.

As I noted above, users seek implicit symmetry across any web experience: in information integrity and consistency, in the kinds of data that we are provided upon visiting a Web site or app, and in how websites behave over time. And as designers and developers, we often find ways to simplify how those complex patterns are rendered so that they don't overwhelm our users and send them spinning into confusion—usually in some form of a black box feature.

Both Google and Wikipedia are exploiting what Jakob Nielsen noted in a post on Useit.com as "informavore navigation behavior." When users are confronted with links, navigation, or other elements that imply a potential pattern, they have to make

"…tradeoffs based on two questions:

  • What gain can I expect from a specific information nugget (such as a Web page)?
  • What is the likely cost to discover and consume that information? (Cost is typically measured in time and effort…

Both questions involve estimates, which users can make either from experience or from design cues. Website designers can thus influence the analysis by designing to enhance user expectations of gains and reduce their expectations of costs. Ultimately, of course, what the site actually delivers is more important, but you'll never get experienced repeat visitors unless their first encounter is fruitful."

I would place emphasis on Nielsen's use of the word fruitful. Both "I'm Feeling Lucky" and "Random article" imply through their behavior that there is a very low cost to discover and consume information of potential interest. If I was searching for the best camera under $200 for my best friend's birthday, I might utilize "I'm Feeling Lucky." I save the cost of having to wade through search results, but I also have lower expectations regarding the most relevant result.

In an article on Johnny Holland called "Applying Curiosity to Interaction Design: Tell Me Something I Don’t Know," Stephen Anderson outlined three principles for how patterns of information are formed in the eyes of the user confronted with a mystery:

  1. Some tiny bit of information makes us aware of something that is unknown.
  2. Context provides some relevance.
  3. Enough clues are given to help us make a judgement about the personal value of that unknown information.

This isn't a simple, linear process we go through. We continually re-calculate and re-judge information, sometimes within microseconds, to determine the correct path forward through a website. Author Nassim Nicholas Taleb notes in the book Fooled by Randomness,

"…just like the helical propulsion of an actor into stardom, people patronize what other people like to do. Forcing rational dynamics on the process would be superfluous, nay, impossible. This is called a path dependent outcome, and has thwarted many mathematical attempts at modeling behavior."

When you add in the input of other users as part of the provided information—like the Flickr feature I'd noted above, whose data is derived from community activity—even more emotion influences the user's behavior as they progress through their site experience.


And for my final trick, I will now vanish…

I've been mulling over the content in this piece for almost a year, and I'm starting to suspect that after all of this thinking about making sites feel more magical, this is the most important principle of all:

Users perceive the most magic in what seems physically impossible and personally relevant. If I watch a woman being sawn in half, I squirm in my seat because I'm imagining myself being hacked apart. And thankfully, I can hit the Back button if things start to get too messy.


David Sherwin

Check out this response by Joshua Blake that builds upon the ideas on this piece:


The comments to this entry are closed.