Designing for the scent of information (Jared M. Spool, Christine Perfetti & David Brittan)

UIE Fundamentals

Designing for the Scent
of Information

User Interface Engineering
www.uie.com
Jared M. Spool

Christine Perfetti

David Brittan

Designing for the Scent of Information

i

UIE Fundamentals Reports
There are concepts every web designer should know. These are the concepts that
determine how people find and digest information. In other words, how they use the
web.
In the UIE Fundamentals Series, User Interface Engineering brings such core concepts to the surface. The series presents proven, time-tested ideas that drive today’s
most successful designs.
The series opener, Designing for the Scent of Information, reveals the techniques by which
the most successful designers draw users toward content. We believe the concept of
Scent will change the way you think about designing web sites.

Written By:

Jared M. Spool, Christine Perfetti, and David Brittan

Published By: User Interface Engineering
4 Lookout Lane, Unit 4D
Middleton, MA 01949
(978) 777-9123
www.uie.com
Cover Design: Victor Martello
Martello Studio
www.MartelloStudio.com
Copyright © 2004, User Interface Engineering. All rights reserved.
This report is protected by copyright laws. Reproduction is not permitted, except when the licensee
has purchased a group license, which allows duplication for other members of the licensee’s organization and storage on the organization’s intranet.

10 9 8 7 6 5 4 3

www.uie.com

©2004 User Interface Engineering. Reproduction Prohibited


1

The Scent of Information
Everybody knows that some web site designs are much better at helping users find
information than others. But few people know why. It’s tempting to assume that the
most successful sites have certain advantages—larger development budgets, more
sophisticated technology, or designers with more talent.
After watching hundreds of users navigate on scores of web sites, we are prepared to
say the most effective sites do have advantages, but not the ones people think. The
successful site designers all understand how users behave when they’re hunting for
information.
Those designers know something that eluded us when, years ago, we began observing users on web sites in our lab at User Interface Engineering. At first, we thought
users behaved randomly, like fish swimming in the ocean. Some users did well, some
didn’t, and we had no way of predicting how people would do on each site.
Soon, we began to notice patterns. On some sites, for example, we realized that
every user hit the Back button at the same point. We noticed that certain types of
links always worked better when arranged in particular ways. All the successful designs were subtly—but distinctly—different from the failed designs. We had to come
up with a name for this hard-to-see
difference.
A few years earlier, researchers at Xerox’s
Palo Alto Research Center theorized that
when people search a large information
space—a web site, for example—they exhibit
the same behavior as animals hunting their
prey. They become “informavores” on the
hunt for information. The PARC researchers
called their theory the “scent of information.”
The users we observed seeking content on
web sites displayed just such foraging patterns. Our users would search for a scent trail
and, once they had picked one up, would
follow it toward their content. As the scent
got stronger, they grew more eager. When
they lost the scent, they backtracked until they
picked it up again.
How do you recognize the scent of information? There’s not much to see. As you’ll come
to realize, good scent has certain visible signs.
These have to do with the placement of critical information, the length of links and pages,
and a few other perceptible elements. Usually,

www.uie.com

In this Report…

The Scent of Information............................... 1
Following Scent ............................................ 2
Things that Block Scent ................................ 5
Iceberg Syndrome.................................... 5
Camouflaged Links .................................. 6
Banner Blindness ..................................... 7
Links That Lie........................................... 7
Missing Words ......................................... 8
Cute Link Names ...................................... 8
Misplaced Links ....................................... 9
The Search from Hell .................................. 10
Information Masking .............................. 12
Byzantine Policies .................................. 14
Navigation-Panel Shell Game...................... 14
Categories that Belong on Jeopardy ...... 14
Inscrutable Jargon ................................. 15
Longer Is Better........................................... 16
Longer Links Say More .......................... 16
Longer Page, More Scent....................... 18
The Scent of Graphics ................................. 20
Navigation graphics ............................... 20
Content Graphics.................................... 21
Decorative Graphics ............................... 21
Designing for Scent..................................... 22
Content Must Rule.................................. 23
Measure Users’ Confidence ................... 23
The Tao of Scent ......................................... 25



2

however, scent is invisible. It is a product of how well the designers understand the
site’s users, those users’ needs, and how the users access the site.
In fact, the best way to detect scent is to measure the users’ confidence (a process
we’ll walk you through later on). When the scent is weak, users are not confident at
all. They doubt their choices. They tell us they are making “wild guesses.” They click
hesitantly, hoping the site will magically come through for them. More important,
they rarely find what they are seeking.
When scent is strong, however, their confidence builds as they draw closer to their
content. They traverse the site with little hesitation. Moreover, they find what they
are seeking.
Users display just such confidence on all the most successful sites—sites like CNN,
BBC, Amazon, Dell, eBay, Lands’ End, and the U.S. Bureau of Labor Statistics.
Those good designs didn’t happen by accident. They happened because the designers
understood the scent of information. The designers learned how to apply the principles of scent, how to make content pull users toward it like honeybees to clover. You
can too.

Following Scent
The best way to understand what scent is all about is to see it in action. Let’s look at
how scent helped Doug, a user we tested a few years back. When he came to our lab,
Doug had just purchased a Hewlett-Packard DeskJet 722C inkjet printer for $10 at a
yard sale. From the yard sale, he immediately visited his local office supply store and
bought the necessary ink cartridges for $55. Now he had invested $65 in the printer,
but he didn’t have the software to run it on his Windows 95 system.
We brought Doug to HP.com to find the necessary software
[Figure 1]. He scanned the page and noticed the picture of the CD
with the word Drivers above it.
We’ve noticed that people looking for information all exhibit
similar patterns. They first scan for their trigger words—words or
phrases they associate with the content they’re seeking—in an attempt to pick up the
scent. That is what Doug did on Hewlett Packard’s site. He clicked on the word
“drivers”—one of his trigger words—because it told him he was on the right track.
Oddly enough, Doug had never mentioned the word “drivers” before the page came
up—he kept referring to “software.” If Doug had not known that a driver was what
he needed, he would have failed on the site. But the designers at Hewlett Packard
probably figured that any user who could install a driver knew what one was.
When Doug clicked on Drivers, it brought up the HP Customers Service page [Figure
2]. This page demonstrates the power of trigger words. In spite of the clutter, Doug’s
eyes went straight to the trigger word in the lower right-hand corner, Printers.

www.uie.com



3

Figure 1: The HP.com home page when our user, Doug, visited to find his driver.

As soon as Doug identified the Printers choice, he clicked the Next button and
moved right on, with no hesitation. Our
research consistently shows that, on a welldesigned page, the user’s trigger words will
jump right out, regardless of how much
other content the designers provide.
At this point, the site started to bully Doug with a series of questions designed to
winnow down the list of possible drivers. What type of printer? What series? Which
model in the 700 Series? Finally, after all these questions, Doug specified the HP
DeskJet 722C Printer.

www.uie.com



4

Figure 2: When Doug clicked on the Drivers link, he saw this very busy page.

You might think a site would annoy a user when it badgers them with questions. Yet
these aren’t just any questions. These questions show the site is becoming more specific. Moving from general information to specific information gives users the
necessary confidence that they are on the right track. With each click, they feel
they’re getting closer to their goals, so they continue to search for their content.
When he arrived at the HP DeskJet 722C Printer
page [Figure 3], Doug encountered a couple of
glitches. First, he noticed the Drivers & Software
link in the center of the page.
Even though Doug had already informed the site he was interested in drivers, here
the site was asking him yet again to specify what he wanted. Did this put him off?
Not at all. As long as a site is bringing them to where they want to go, users seldom
complain that a site isn’t listening to them. They have the confidence to go forward.
The second bump in the road
occurred when the site offered Doug a
choice between two similar-sounding
drivers, both of which worked with
Windows 95.
What is the difference between a 720C Series Driver and a 720 Series Printer Driver?
Although Doug stumbled over this question for a moment, he recovered quickly. It
turned out that both drivers contained the software necessary to run the printer.

www.uie.com



5

Figure 3: HP.com’s DeskJet 722C Printer Customer Care Page.

In seven clicks, Doug had found exactly what he was seeking. This is what happens
when scent works.

Things that Block Scent
As Doug’s experience showed, there are many ways designers can enhance scent on
a page. Unfortunately, there are even more ways to prevent scent from working. Another user in our lab attempted to find a driver for the same printer, the DeskJet
722C, on the same site, HP.com, but with very different results. Let’s look at a few
of the scent blockers this user encountered.
Iceberg Syndrome
When our user—Linda was her name—arrived on Hewlett Packard’s home page, she
missed one critical piece of information that Doug had managed to find: the Drivers
link [Figure 4]. Even though the link is visible in our screen shot, on Linda’s screen it
fell just beneath the fold, the imaginary line on the page at which users need to scroll.
In general, there’s nothing wrong with placing links beneath the fold. The idea that
users won’t scroll is only a myth. Our research shows that users are more than happy
to scroll, so long as the designers do nothing to prevent them from doing so. (We’ll
talk more about scrolling a little later.) What prevented Linda from scrolling on this
page was the content in the center, which contained marketing messages about Hewlett Packard. Linda told us this content was just “marketing fluff” [Figure 5].

www.uie.com



6

Figure 4: Because Linda didn’t scroll, she missed the critical Drivers link below the fold on her 800x600
screen.

Linda assumed that any content below the fold would be even less interesting to her
than what was above the fold, and so she didn’t bother to scroll.
We call this phenomenon iceberg syndrome. It happens when users assume that the
above-the-fold content represents the tip of the iceberg, and that the unseen content
is more of the same. In this case, iceberg syndrome prevented Linda from finding
the Drivers link.
Camouflaged Links
Unbeknownst to her, Linda had another chance to find her driver quickly. She could
have clicked on the HP Services & Support link at the top of the page. That link would
have brought her straight to a page that listed the necessary drivers.

www.uie.com



7

Figure 5: Linda didn’t scroll because of “marketing fluff” that appeared above the fold.

One thing that may have stumped Linda is the link doesn’t look like a link. The web
has done an excellent job of training people to think that links are blue and underlined. When we hitch users up to an eye-tracking device, we see their gaze go straight
to the blue underlined text on a page. This HP Services & Support link is neither blue
nor underlined. Instead, the link appears as a graphic that looks more decorative than
functional. Linda might not have realized the graphic was something she could click
on—if she had seen it at all. That brings us
to another scent blocker.
Banner Blindness
Linda probably never saw the HP Services & Support link, because it appears in the top
60 pixels of the home page. What typically appears in the top 60 pixels of a home
page? Banner ads. Our research has shown that users typically ignore information
presented in that portion of a home page. Such banner blindness occurs because the
web has trained users that nothing of value ever happens in banner ads.
So, after scanning the page and not seeing anything that contained scent, Linda did
what pretty much everyone does: she clicked on the only link that seemed to offer a
prayer of getting her to her desired content: Search. What a rabbit hole that turned
out to be.
We’ll come back to Linda and her adventures in Search Land, which so fascinated us
that we want to go into them in some detail. Meanwhile, let’s look at some other
common ways in which designers throw users off the scent.
Links That Lie
As we’ve mentioned several times, there is a strong connection between a web page’s
scent and the user’s confidence. When the scent weakens, so does the confidence of
the user. It works the other way, too. When users lose confidence, they start to think
the scent is weak.

www.uie.com



8

Users are especially apt to lose confidence when the site lies to them. On an old version of the Discovery Channel site, for example, one of our users clicked on a link
for a “complete list of the Discovery Channel Home Videos.”
Imagine the user’s disappointment when the
next screen contained not the promised
complete list but just the most popular
videos. The user said, “Why can’t they just
tell me they only offer the bestsellers?”
We’re sure the site’s designers didn’t mean
to mislead. Nevertheless, when it happens,
it stops users in their tracks. We’ve watched
many users lose confidence when links don’t deliver the content they promise.
Missing Words
Confidence also plummets when a new page, appearing after the user clicks a link,
doesn’t contain the trigger words from the link that led to it. Have you ever found
yourself poring over a web page, wondering why the site took you there? Was it because the trigger words were nowhere in evidence? Have you ever had to use your
browser’s Find function to hunt for the trigger words on the page?
That happened when a user tried to renew his subscription to the
Ziff Davis magazine PC Week (the precursor to eWeek) on ZDNet.
He recognized the image of the magazine among the other
publications that Ziff Davis offers and clicked on it.
The subsequent page had nothing to do with PC Week. It was a
listing of more than 400 publications, in alphabetical order. The user had to click
three more times before finding any mention of PC Week. We’re glad to report that
eWeek subscribers are no longer subjected to any such rigmarole.
Cute Link Names
Many sites try to be clever in the labeling of links on a page. Unfortunately, this
strategy often backfires, making it difficult for users to know where to click.
At one time, the Library of Congress, afraid to appear boring, assigned cute names to
its online databases of legislative information: THOMAS and Will T. Bill. While we
can understand the library’s desire
to make things lively, it didn’t go
over well with our users.
The result: users didn’t know what the links were. If the associated text didn’t offer
at least some clarification, the users would’ve been completely lost.
It’s extremely tempting to use marketing names, clever buzz phrases, industry terminology, and cute nicknames to make things seem fun or interesting. But using
obscure terms in the link just weakens the scent and frustrates the user. Unfortunately, clarity always trumps fun.

www.uie.com



9

Misplaced Links
On some sites, links appear in
entirely the wrong context. We
watched Tim, a user on
Boston.com, the Boston Globe’s web
site, trying to find out when the
Boston Red Sox would next play the
Toronto Blue Jays. On the home
page, Tim clicked on the Red Sox
link, which took him to a page in
the sports section devoted to all
things related to the Boston team.
At the bottom of the screen, Tim
clicked on a Sports Calendars link, telling us he was “very confident” he was on the
right track. He said he expected the link to lead straight to the Red Sox schedule. But
that’s not where he ended up. Instead, he ended up on a page listing all the things to
do and places to go in the Boston area. Nowhere on the page was there any specific
information on the Red Sox. Tim was baffled.
When we talked with the designers
of Boston.com, they explained that
they hadn’t intended the Sports
Calendars link to be specific to the
content of the Red Sox page. On the
contrary, they pointed out that the
same three links appeared on every
page of Boston.com’s sports section.
Not only did Tim not know that, he
didn’t realize there was a sports
section. Clicking on the Red Sox link on the home page, he had encountered nothing
labeled “sports” or “sports section.”
Users and designers view the same site from
different perspectives. Designers see entire
sections, while users see only the pages they visit. The three links that made perfect
sense to designers made no sense to Tim, who had no concept of a section. The
links might as well have carried a warning label that read: “Chances are these links
don’t lead to what you’re looking for.”
Today, sports fans can have an easier time finding team schedules on Boston.com
[Figure 6]. They just click on the Schedule/Results link on the page devoted to the
team of their choice.

www.uie.com



10

Figure 6: Today, Boston.com visitors can easily find the schedule for their favorite teams with the
Schedule/Results links.

The Search from Hell
Now back to Linda. As you recall, she was having trouble finding a driver for the
DeskJet 722C inkjet printer on the Hewlett Packard web site. She thought she could
pick up the scent by clicking on Search.
The trouble is search engines are scentless. Regardless of the content that is behind
them, nearly all search boxes on all sites look the
same, with a box to type into and some button to
start the search.
When users arrive at a web page that contains a search window, they at first behave
the way visitors do on any site: they scan the page for their trigger words or any recognizable synonyms. Only when they can’t find their trigger words do they resort to
Search.

www.uie.com



11

Figure 7: Upon clicking the home page Search link, Linda found herself on this page.

What words do users type into the Search box? Why, their trigger words of course.
It’s as if they were attempting to create their own link. When users go to Search, they
are signaling to the designers that they aren’t picking up scent on the page. Let’s see
how Search worked for Linda.
Clicking on the home page’s Search link brought Linda to the site’s Search page, allowing her to search the entire site [Figure 7].
Interestingly, the Drivers link that got lost below the fold on the home page now appeared clearly above the fold. But Linda didn’t click on the link. That’s because of a
phenomenon we call “When Users Decide to Search, Damn It, They Are Going to
Search!” She typed her trigger words, “DeskJet 722C Driver,” into the Search field.
Now, from our earlier user, Doug, we knew that there were two Windows 95 drivers
for the DeskJet 722C Ink Jet Printer. Which is why everyone was surprised when,
after Linda typed “DeskJet 722C Driver” into the Search box, the HP site returned
24,309 results [Figure 8].
Search engines seem to go out of their way to return the largest number of results
they can. Certainly, 24,309 results is a very large number. However, Linda didn’t
want a large number of results. She wanted just one result—the driver for her
printer.

www.uie.com



12

Figure 8: Instead of returning Linda’s two desired drivers, HP.com’s Search returned 24,309 results.

The site’s search technology tried to help Linda choose from among the 24,309 results by providing a relevancy rating. Unfortunately, our research shows that users
really don’t understand this rating. After all, how does a user know the difference
between 88, 86, and 85 percent relevant?
Linda did what most users do in this
situation: she scanned for the most
likely link—the link with the best scent. To Linda, however, no link offered very
good scent. She finally chose the link with the phrase “What Ships in the Box,” reasoning that the drivers must come in the box with the printer. How confident was
she that the link would take her to her driver? Linda’s words: “It’s a wild-ass guess.”
Scent is about confidence. When users lose confidence, they are telling us the scent is
weak.
Information Masking
When Linda clicked on “What Ships in the Box,” the site presented the various
components that come with the printer [Figure 9]. The page contained a Drivers &
Software link. If Linda had clicked on it, she would have found exactly what she was
looking for. Instead, she fell prey to a phenomenon called
information masking.

www.uie.com



13

Figure 9: Linda missed the Drivers & Software link in the left navigation bar because of a classic case
of information masking.

Web pages are busy visual spaces. When users are having trouble finding what they
want, their brains tend to mask out sections of the page that never bear pertinent
information. For example, if useful material always appears in the center column, the
user masks out the other columns.
That’s what happened to Linda. Having repeatedly found links at the center of the
screen, she stopped looking at any links that appeared on the right. Therefore, she
completely overlooked the Drivers & Software link.
What Linda did notice was the Printer Software link in the center of the screen. Since
this link contained her trigger words “printer” and “software,” her confidence soared
and she willingly clicked on.
Silly her. All she got for her trouble
was this text, which described what
software comes in the box but does
not link to the drivers. In fact, it
doesn’t link anywhere.
Linda had hit a dead-end. She had followed what, to her, seemed the best scent
available and, because of a slight wrong turn on the home page—her failure to scroll
down to the Drivers link—had ended up losing the scent entirely. Users’ only options
at this point are to backtrack until they pick up a better scent or to give up. Most users who encounter this type of failure choose the latter.
Byzantine Policies
After Linda’s experience, we asked the developers at Hewlett Packard why the
Printer Software page did not link to any software. They blamed a geographic restric-

www.uie.com



14

tion. It turns out the team that developed this page worked at HP’s office in Vancouver, Washington. The team in charge of the page that lists the drivers worked out
of Roseville, California. We wondered if Hewlett-Packard corporate policy prohibited people in the State of California from talking to people in the State of
Washington—or was it a question of Oregon State Law?
On further investigation, we learned that, at
For more information on how users interact with
the time the teams created these pages, HP
Search, see the UIE Best Practices Report
had a policy against “deep linking” (linking
Getting Users to What They Want
available on www.uie.com.
to a page other than a home page) from one
part of HP’s site to another. Operating independently, the teams couldn’t control when each other’s pages would move, creating the dreaded 404 (Object Not Found) error. Since deep linking was restricted,
corporate policy prohibited the Vancouver folks from linking to the actual software
that was available on the site. They could only describe it.
As is typical, the site’s search engine prevented Linda from picking up a scent. It’s
possible, of course, that HP’s designers could rejigger the search engine to provide
exactly what Linda wanted. Should they? Simply making the link to the drivers more
prominent on the home page would solve the same problem at a fraction of the cost.

Navigation-Panel Shell Game
Search is not the only site feature that explicitly prevents scent from getting through.
Navigation panels can often have the same effect. Designers understand that each
link in the navigation panel represents a large amount of information, so they make
each link as general as possible. The more general and the less descriptive the link,
however, the less likely it is to contain the user’s trigger words.
One user likened the categories in navigation panels to playing an old-fashioned shell
game. The design forces the user to guess which shell contains the magic pea. In the
shell game, the goal is to make each shell look identical so that the user can’t guess
correctly. Designers need to approach the problem from the opposite direction: each
shell should be, in effect, transparent, letting the user see the content right through it.
Categories that Belong on Jeopardy
Ken came to our lab looking for the first Tom Clancy novel that featured the fictional hero Jack Ryan. When Ken visited Amazon.com, the navigation panel didn’t
give him enough information to decide which link to click.

Ken couldn’t decide whether the content fell under Bestsellers, Featured in the Media,
or Award Winners. He knew that many Tom Clancy books were top sellers. At the same
time, he knew the books received a lot of press and could very well have won

www.uie.com



15

awards. Baffled, Ken did what most users do when they run out of scent: he went to
Search.
Amazon’s latest design narrows the choice down to Bestsellers. Unfortunately, since
this link shows only the most recent best-selling books, it’s unlikely it would go back
to the first Tom Clancy book to feature the Jack Ryan character.
Another user, Janice, arrived at our lab in the throes of a strange stomach condition.
Gastroparesis, her doctor had called it. We watched as Janice went to Dr. Koop’s site
to find information about this mysterious ailment. Most of the navigation links
sounded the same: Health Topics, Healthy Body, Healthy Lifestyle, and My Health Page.
What was the difference between them? These categories could be the first round of
Jeopardy. (“I’ll take ‘things that ooze’ for $600, Alex.”)
Janice clicked on Diseases & Conditions, believing the page would describe gastroparesis, but, alas, it did not. The definition of gastroparesis was hiding under the link
Resources.
When we asked the designers at Dr. Koop why the definition for gastroparesis
wasn’t under Diseases & Conditions, they initially told us it was because gastroparesis
isn’t actually a disease or a condition—it is a symptom. While that may be true, we
pressed a little harder and discovered a deeper reason.
It turns out that the content for the definition did not originate at Dr. Koop. Instead,
the site had purchased a dictionary of medical terms from a third party. The designers had decided that all third-party content, including this dictionary, should appear
under Resources.
We can’t say whether placing a site’s third-party content under a single link is a good
policy or not. When designers make policy decisions about content placement, however, they have to communicate those decisions to the users. Designers communicate
decisions through links. If the links are not effective, users won’t know the policies
and won’t find the content.
Inscrutable Jargon
This is the navigation bar from an early version of Fidelity.com:

Now, we’re not even going to mention the two links in the upper left—Fidelity.com
and Home—that went to completely different pages. Instead, we want to tell you about
Matthew, a hardworking 26-year-old who, two years earlier, had landed an excellent
job fresh out of college. By living conservatively, he had saved up a decent nest egg.
Friends and family were starting to pressure him to think about putting some money
away for his retirement, but he had no idea where to start.

www.uie.com



16

That’s where things stood when Matthew came to our lab as a user. He wanted to
research potential options to start planning for his retirement. So, imagine his confusion
when presented with three separate links: Research, Retirement, and Planning.
If you work in financial services, you know exactly what these three terms mean. Unfortunately, outsiders don’t. Industry terminology can be a helpful shortcut, so long
as you can count on users’ understanding it. When your users come from outside
your industry, jargon can block your scent.
Fidelity has since purged its navigation bar of such jargon. Testing has shown that
their designers have significantly improved the scent since those early days.

Longer Is Better
We often tell our clients something they find counterintuitive: Don’t worry about
keeping things short. Focus on ensuring that all the necessary trigger words are on
the page. That often means making both links and pages longer. For many, this advice goes against everything they’ve learned about being concise. Yet we’ve
repeatedly found that longer links and pages make for better scent.
Longer Links Say More
If you look closely at the links in the navigation panels we just studied, you will notice that they are all between one and three words long. While this is extremely
common for navigation panels, it also is indicative of problems we consistently see
across many sites.
In our research, we’ve collected hundreds of clickstreams (sequences of clicks that
users make when trying to accomplish their objectives). We sorted the clickstreams
into two groups: the clickstreams where users succeeded and the clickstreams where
users failed. Then we looked for patterns in each group.
One of the first patterns to emerge had to do with the optimal number of words in a
link. The best-performing links were between seven and twelve words long.
In the chart in Figure 10, the horizontal axis represents the number of words in each
link. The vertical axis shows the likelihood that a link will bring users closer to what
they are seeking. (As you would expect, a zero-word link has a zero chance of succeeding.) While the average success of all links was about 43 percent, links containing
seven to twelve words did substantially better.

www.uie.com



17

Figure 10: Our research shows that links seven to twelve words long are more likely to take users where
they want to go.

Why did links with six or fewer words do less well? The answer is that it’s a game of
probability. A seven-word link is more likely to contain the right trigger words than a
one- or two-word link. The more likely a trigger word is to be present, the more
likely the link is to give off good scent. This doesn’t mean that a one-word link never
gives off good scent. In the Hewlett Packard example, the word “Drivers” gave off
great scent for the users who saw it.
Most often, though, short links work only when trigger words are nearby. CNN.com
is adept at that practice. The site uses many short links, like U.S. and WORLD. What’s the
difference between U.S. and World? Sometimes, it is not clear.
For example, a user wondering about the spread of SARS in North America might
not know where the editors of CNN have decided to put the SARS-related stories.
After all, something like SARS could be either a U.S. or a World story.
CNN strengthens the scent
of short links by following
them with lists of longer
links that are heavy on the
trigger words. Noticing the site already lists SARS in a story under WORLD, the user
knows to start there, even though the page doesn’t mention SARS in North America.
If the optimal length of a link is a game of probability, why don’t links that are 13 or
more words work well? At some point, the trigger words get lost in the surrounding
text. Too many words in the link and the users can’t see the individual trigger words
they need.

www.uie.com



18

Longer Page, More Scent
Many designers fall for the common myth that users don’t want to scroll. As a result,
they build short web pages with only a small amount of content and a few links, but
with the supposed virtue that they require no scrolling.
The problem with short pages is they block scent. They limit the amount of content
that is immediately available to users. To see more, users must pick a link in hopes of
finding useful information on subsequent pages. If they don’t succeed, they resort to
jumping back to the previous page.
Our research shows that longer pages increase the likelihood users will find the content they want. Users on CNN, Yahoo, or Amazon have no problem scrolling the
long pages on those sites to find the content they were seeking. We’ve seen that users are very willing to scroll as long as the content above the fold tells them it’s
worth scrolling.
As we suggested earlier, careless page design can deter users from venturing below
the fold. This is evident on an old version of the Sprint home page [Figure 11].
Users in our studies didn’t realize that this page contained two more screens beneath
the fold. The Sprint designers had made certain choices that suggested to users there
was no point in scrolling. First, they included a horizontal rule at the bottom of the
page.

Figure 11: Without realizing it, Sprint’s designers prevented users from scrolling by choosing scrollstopping design elements, such as a horizontal rule and large margins.

A horizontal rule acts as a border on the screen, a signal to stop scrolling. Second,
the designers inserted a large margin after the horizontal rule, a margin that just happened to fall on the fold boundary. This unfortunate coincidence suggested to users
there was nothing further to see. Had some content worked its way up on the page
and eliminated the margin, users would have known to continue scrolling.

www.uie.com



The current Sprint site remains wedded to using a
large number of horizontal elements, any one of
which could—depending on the size of the user’s
screen—signal the end of the page [Figure 12]. If
asked to test their most recent designs, this is something we would certainly be looking for.

19

For more information about how users
scroll, see the article
As the Page Scrolls
on www.uie.com.

Figure 12: The current Sprint site still contains elements that could prevent users from scrolling.

Designers need to be aware of these scroll stoppers and ensure that they aren’t accidentally putting them in the wrong place. With adequate attention to maintaining a sense
of flow, designers give the user “permission” to scroll. In so doing, they give themselves permission to build longer pages—pages that emit more scent and give users a
better shot at finding the content they want.

www.uie.com



20

The Scent of Graphics
Over the years, User Interface Engineering has developed a reputation for being
down on graphics. We earned some of this with our 1997 publication Web Site Usability: A Designer’s Guide, in which we stated, “Graphics neither helped nor hurt a web
site.” Since then, we’ve investigated the topic more thoroughly and concluded there
isn’t just a single type of graphic. Instead, there are three types: navigation graphics, content graphics, and decorative graphics.
When we look closer at the three types of graphics, we can see that navigation and
content graphics can help users a lot, while ornamental graphics contribute little to
usability.
Navigation graphics
Certain types of images give off scent because they help users understand what they
will see once they click. These are what we call navigation graphics.

In the image above, from the U.S. House of Representatives web site, Arizona residents interested in finding their state representative need only click on the portion of
the map labeled AZ. Anyone familiar with U.S. geography will easily find the right
link.
However, designers must use navigation graphics with care. One problem with using
graphics, even graphics of text, is that they often don’t look like links.
We’ve seen sophisticated users wave their mouse around the screen, waiting for the
browser to send them a sign. We’ve also seen many users miss the links entirely.
Navigation graphics work best when users can clearly identify what is a link.

www.uie.com



21

Content Graphics
Sometimes the best way to communicate information is with a picture. Content
graphics are images that communicate content directly.
Amy came to our lab interested in buying a new car.
She knew she wanted one of GM’s sporty two-door
coupes, but she didn’t know its name. That’s understandable. Many GM cars have similar names:
Firebird, Skylark, Skybird, Sunbird. She just didn’t
know. Photos of cars on Edmunds.com helped her
tremendously.
The moment Amy saw the picture above, she exclaimed, “That’s it!” The picture,
while quite dull when viewed out of context, was very exciting for Amy. As a content
graphic, it did its job.
One day, we had the opportunity to watch a user,
Mitch, spend 2½ minutes downloading this image
on the equivalent of a 14.4 kbps modem.
For what seemed like ages, Mitch sat entranced by
his browser as the image slowly materialized. As
soon as the entire image had loaded, he shouted,
“That’s fabulous!” and proceeded to place a $38 bid
on the tin. (We don’t think it cost $38 when it was
new and filled with fresh coffee.)
Mitch was an avid restorer of antique coffee cans. He regularly purchased cans just
like this, painstakingly restored them, and resold them for hundreds of dollars. The
image, albeit tremendously slow to load, told him everything he needed to know to
bid on this item.
By the way, the image so mesmerized Mitch that it
never occurred to him the modem was slow. This is
a common behavior we see—users don’t mind slow
downloads when they are getting what they are
looking for.

For more information on how people deal
with slow downloads, see the article
The Truth about Downloads
on www.uie.com.

Decorative Graphics
Designers often turn to images to make a site feel fun, authoritative, or professional.
Such ornamental graphics may provide a site with visual flair, but we’ve found that
they neither help nor hurt the user experience. We can measure users’ perceptions of
sites as being fun, authoritative, and professional. What we’ve learned is that the
presence of a decorative graphic usually has no
effect on those attributes.
Sometimes, however, decorative graphics can backfire on the designer. An image from the site of
National Public Radio’s popular “Car Talk” program depicts the show’s co-hosts in surgical garb

www.uie.com



22

operating on an engine block. The site has many similar caricatures, obviously meant
to bring the same whimsical elements of the radio program to the site.
One of our users was having a hard time scouring the Car Talk site for information
on tire rotation. When she encountered the surgery cartoon, she turned to us and
said, “I’d be happy if the designers spent as much time organizing the information as
they did making these cute graphics.” She was not amused, she was frustrated. Our
research has proven repeatedly that if a site’s organization isn’t working, users are
unlikely to rate the site as fun, authoritative, or professional, no matter how great the
graphics are.
We’re not saying that designers shouldn’t use decorative graphics. We’re only suggesting that they focus their efforts on the elements of design that matter most to their
users.

Designing for Scent
Scent may be a natural phenomenon, but designing for scent doesn’t come naturally.
When we interview designers who have created sites with excellent scent, we invariably learn they have undergone a fundamental change in their approach to design.
Conventional web designers
approach a new site by starting with
the home page. To them, the home
page is the center of everything.
When designers focus on the home
page, they are thinking in terms of
navigation.
Consequently, they ask the wrong
questions: Where are all the places a
user might want to go? What clever
names do we want to call our
sections? How can we make sure
users understand our agenda?
Such designers are thinking about how to push the user off the home page to different parts of the site. This approach often forces designers to think about populating
the home page before they even understand what the site’s content will be. Teams
can spend weeks arguing over the exact wording of inscrutable links in unfathomable
navigation bars.
This approach is popular and largely unquestioned, but it often fails to yield a usable
site. The failure comes from the designer’s focus on structure instead of content. Users
don’t care about the site’s structure. They are interested purely in finding content
that lets them achieve their goals. The conventional approach to design forces the
designer in the wrong direction.

www.uie.com



23

Content Must Rule
To design for scent, you have to start with the content pages and worry about the
home page later. You start with the most important content and determine the most
effective trigger words. You then place those trigger words on the pages where users
will need them. Now that content is key, you find yourself asking an entirely different set of critical questions:
Why are users coming to the site?
Which page is most important to the user?
How will users find this page?
What are the users’ trigger words?
Where are they likely to look for
those words?
By focusing on the most important
content, you will end up creating a
structure that reflects the priorities of
the user. Most of the highly successful
sites on the Internet share this
approach. Attend to the content and
the structure will take care of itself.
Measure Users’ Confidence
Attending to the content means, among other things, making sure that each link you
create and each page you build is emitting the right scent. Moreover, as we are fond
of repeating, the only reliable way to judge scent is to measure users’ confidence.
One way to do that is to look at clickstreams, just as we did. Is your design forcing
users to hit the Back button? Do they often resort to using the site’s search engine?
Users often turn to these devices when their confidence is low, hoping they can pick
up the scent some other way. Detecting these subtle clues can help turn a poor site
into a great one.
The best way to measure confidence, however, is to engage with actual users as they
navigate. In our research, we’ve found there are two critical times to look for confidence: before the user clicks and right after. Before clicking, users are trying to
decide which link is the right one for them. We ask them, “How confident are you that
this link will get you to what you’re looking for?” If the users tell us they are very confident,
then we know the page probably has the right trigger words and the layout isn’t confusing them with too many similar words or visuals.
On the other hand, if users indicate they have little or no confidence—they are in
fact guessing—then we can focus our improvement efforts on this page. We make
sure trigger words are prominent, the layout communicates the options to the user,
and the graphics are adding to the scent of the page.
The story doesn’t end there. When the new page appears, it needs to do one of two
things: it should either contain the user’s target content or give that user a stronger

www.uie.com



24

scent than the preceding page did. Users expect scent to grow stronger until they
reach their goal. If the scent weakens, they lose confidence.
As soon as the new page displays, we ask users, “Is this page getting you closer to your goal
or farther away from it?” The answer tells us if the scent is good and the site is working.
If users feel they are moving away from their goal, then either the new page is not
communicating properly or the preceding page is giving off the wrong scent.
Does that sound like a lot of effort? It is. Designing for scent requires both collecting timely information about what users will find important and constantly testing
your assumptions along the way. But all that hard work pays off when you stand
back and realize you have created a web site that emits fabulous scent.

www.uie.com



25

The Tao of Scent
The most successful web designers understand how to create pages that lure users
toward the content they need. Here is a summary of what every designer should
know about the scent of information.
The design communicates scent through links. Each link needs to have
strong scent for the content that lies beyond it.
Trigger words are the words or phrases that cause users to click. Scent is
strongest when the page presents users with the right trigger words.
Links work best when they are between seven and twelve words long. That is
long enough to boost the chances of a user’s trigger word appearing in a link,
yet short enough to let the user find the word easily.
Trigger words are extremely powerful. They often jump out of very busy
pages that we’d otherwise expect users to struggle with.
Users expect each click to lead to information that is more specific. They do
not mind clicking through large numbers of pages if they feel they are getting
closer to their goal with each click.
Users don’t mind scrolling. They fail to scroll only when design elements
such as horizontal rules suggest there is nothing more to see.
Links that appear in the top 60 pixels of the page fall victim to banner blindness. Banner ads have done an excellent job of training users to ignore links at
the top of a page.
Users gravitate to links that are blue and underlined. To design links differently is to defy what the web has conditioned users to expect.
Links should accurately describe what the next page contains. Otherwise, users lose confidence in the site.
When users click on trigger words, they expect to find those words on the
next page.
Trigger words need to be readily understandable. Jargon and cute marketing
terms confuse users.
Users go to Search only when they can’t find their trigger words on the page.
Relevancy ratings from Search do not communicate scent. Users rarely understand them.
Users often notice only the section of the page in which they last found
something useful. The brain’s information masking causes them to miss critical
links in other sections.

www.uie.com



26

When users run out of scent, they often backtrack in hopes of picking up
better scent. This rarely works, however, since they chose the best scent to
start with.
Links in navigation panels are often too short to emit much scent. They
rarely contain the user’s trigger words.
Two types of graphics emit scent: navigation graphics, which help users decide
where to click, and content graphics, which convey essential information. Decorative graphics—images that merely jazz up a site—are usually scentless.
Designers create good scent by focusing on content pages instead of the
home page. When they understand which content is most important and
how users will access it, the rest of the site falls into place.
We can’t measure scent directly. Instead, we have to look at whether users
are confident as they click from page to page.

www.uie.com



27

About the Authors
Jared M. Spool, Founding Principal (jspool@uie.com)
If you’ve ever seen Jared speak about usability, you know that he’s probably the most
effective, knowledgeable communicator on the subject today. What you probably
don’t know is that he has guided the research agenda and built User Interface Engineering into the largest research organization of its kind in the world. He’s been
working in the field of usability and design since 1978, before the term “usability”
was ever associated with computers.
Jared spends his time working with the research teams at the company, helps clients
understand how to solve their design problems, explains to reporters and industry
analysts what the current state of design is all about, and is a top-rated speaker at
more than 20 conferences every year. He is the conference chair and keynote speaker
at the annual User Interface Conference, is on the faculty of the Tufts University
Gordon Institute, and manages to squeeze in a fair amount of writing time.
Christine Perfetti, Director of Instruction (cperfetti@uie.com)
When you talk with Christine, it doesn’t take a long time to realize that you’re talking
to someone who knows a whole lot about designing usable web sites. As one of the
most requested instructors at User Interface Engineering, she has worked with dozens of companies on their toughest web design problems. It would be very difficult
to find someone who knows more about task design, user recruitment, and usability
testing practices than Christine. She’s regularly a top-rated presenter and keynote
speaker at conferences all over the world.
Christine came to us from Fidelity Investments, where she was an influential member of their highly talented instructional design team. It doesn’t hurt that she was one
of the best students in Brown University’s graduate psychology program. Did we
mention that she taught Human Factors at the prestigious Tufts University Gordon
Institute for Engineering Management?
David Brittan, Director of Publications (dbrittan@uie.com)
David Brittan is a journalist and editorial manager who specializes in bringing technical subjects to life for a wide audience. He has been an editor of such magazines as
MIT’s Technology Review, IBM’s Think Research, and Issues in Science and Technology. His
writing has appeared in New Scientist, The New Republic, Ed. (Harvard Graduate School
of Education), Contract Professional, and many other publications.
Before entering journalism, David was an award-winning classical accordionist. He
received a Bachelor of Music degree from San Diego State University and performed
graduate studies in musicology at Harvard University.

www.uie.com



28

The UIE Fundamental Series is published by
User Interface Engineering
4 Lookout Lane, Unit 4D
Middleton, MA 01949
(978) 777-9123 or (800) 588-9855 (within the U.S. and Canada)
www.uie.com

Company Background
User Interface Engineering is a leading research, training, and consulting firm specializing in web site and product usability. Jared M. Spool founded the company back
in 1988 and has built User Interface Engineering into the largest research organization of its kind in the world. With our in-depth research findings based on user
observation, we empower development teams to create usable web sites that increase
customer satisfaction and loyalty.
Research
We conduct original research on all aspects of web site usability. In addition to our
groundbreaking studies on the Scent of Information and Online Branding, we’ve
identified best practices for creating e-commerce sites that increase bottom-line results.
Publications
If you are like every other web designer we’ve ever met, then you want to build the
best possible sites you can. We’ve designed our UIE Reports to tell you everything
we know about what it takes to create the best web sites.
Training Events
We offer the only research-driven, in-depth training courses designed to teach web
designers exactly what it takes to build usable web sites and products. UIE offers
training events designed for the entire development team. Topics cover how to design for web-site usability, designing for ROI, paper prototyping techniques, and
tricks for including usability testing in the product design process. Events include:

UIE Roadshow: Know Your Users—Denver, Washington DC, Seattle, and Minneapolis
Winter, 2005
When it comes to your users, do you really know what they want? Are you unsure
whether you’re getting the most out of your user research? Many teams tell us they
are missing key information to get into the mind of their users.

To answer these critical questions, we’ve found four of the world’s most renowned
experts on delivering successful content: Kate Gomoll, Ellen Story, Kim Goodwin,
and Ginny Redish. This winter, we’re bringing these four experts with us on the road
for our third Roadshow, which we’re calling Know Your Users. In this 3-day event,
you’ll have the opportunity to spend a full day with each presenter.

UIE Roadshow: The Secret Design Strategies of Highly Successful Web Sites

Summer, 2005
In this 2-day event, you’ll have the opportunity to attend two information-packed
events, learning the design and usability strategies used by today’s most successful

www.uie.com



29

web sites. Jared M. Spool and Christine Perfetti—two of the world’s most knowledgeable web usability experts—will unveil these secrets to you, complete with a
boatload of examples, entertaining commentary, and inspirational insights.

User Interface 10 Conference—Cambridge, MA—October 10 – 13, 2005

Program details available in April 2005
Take your organization’s web design and usability practices to the next level. At User
Interface 10, the most knowledgeable experts we know will offer practical design
tips. You’ll spend 4 days learning advanced techniques from top design, information
architecture, and usability experts.
Learn more at http://www.uie.com.

www.uie.com


Designing for the scent of information (Jared M. Spool, Christine Perfetti & David Brittan)

More Related Content

What's hot (20)

Similar to Designing for the scent of information (Jared M. Spool, Christine Perfetti & David Brittan) (20)

More from Nguyễn Thượng Đan (20)

Recently uploaded (20)

Designing for the scent of information (Jared M. Spool, Christine Perfetti & David Brittan)