The web is all nigh hyperlinks. But, yous may wonder, when presented with a bunch of links, how do users make up one's mind which link to click on and which to ignore? The respond is: information smell. Like food scent guides animals to their meals, information scent guides people to those webpages that are likely to comprise the content they're looking for.

Information scent is a central concept in the data-foraging theory — a theory essential for agreement how people navigate on the spider web and how they interact with different potential sources of data in order to satisfy a question or an information need. In simple terms, information technology says that, if people have a question, they will decide which webpage to become to based on their approximate of (ane) how likely it is that the folio will provide an answer to their question, and (2) how long it'southward going to accept to get the reply if they get to that page.

The estimate of how relevant the page volition be, if visited, is the information scent of that folio.

Definition: The data smell of a source of information (such as a webpage) relative to an information need represents the user's imperfect estimate of the value that the source volition deliver to the user, derived from a representation of the source.

Information scent is a relative concept — meaning that the aforementioned source of information may have unlike data scents for dissimilar information needs. For instance, a link titled Food will have loftier information aroma if you're looking for cheese, merely low data scent if you're looking for a facial cleanser.

You lot may wonder what the source of information stands for and what its representation might be. On the web, the source is usually a webpage — well-nigh often represented by a link. As the user considers whether to click on that link, the link characterization, the content that accompanies the link, the context in which the link appears, and any background noesis (including recommendations from others) that the user may have about the source all influence the data scent that that source emits and thus the likelihood that the folio will be visited or ignored.

In our analogy with food foraging, the source would be a food patch— allow'due south say a savannah. And the representation of that savannah would exist what the animate being actually sees or feels — for case, the sight or scent of antelopes.  The predator (like the human) will probable consider other factors such every bit any prior bad experiences with that savannah (e.g., the memory of a difficult chase or of fighting for the prey with other predators) earlier proceeding with the hunt. And fifty-fifty if information technology does proceed, the hunt may still prove unsuccessful despite all the enticing cues.

Thus, there are four concepts in play: there'south the actual source (the webpage or the savannah) and its remote representation (the link or the sight and smell of the casualty). These have a true value (unknown until afterward consumption, whether reading the page or eating the nutrient) and an estimated value (assessed earlier consumption and used to determine whether to proceed with clicking or hunting).

The data scent is the estimated value of the source derived from: (1) what the user sees (the analog of what the predator smells) — the representation of the source equally perceived from afar (e.g., from a search-results page); the representation includes, commonly, the link text, just too the immediate and more than full general context surrounding that link text; (ii) what the user knows about that source — whatsoever noesis that the user may accept about the search domain, the brand, the author of that source.

Permit'due south explore each of the factors that make upwardly information aroma.

What Makes Up Data Scent

We've seen before that in that location are 2 big components of information scent: what the user sees (which is given by the representation of the information source on other pages) and what the user already knows about the source. The kickoff component is largely controllable past designers: we unremarkably tin decide how a page will be represented on another pages (although perchance to a bottom extent if the other page is a search-engine-results folio). The second component is just indirectly controllable past designers, through the perceived value that they may have constructed in the past for that brand or source of data.

What the User Sees

The Link Label

Perhaps the most important component of the information scent, the link label is supposed to be a succinct even so authentic description of what the page is about. If this description feels relevant to the user's goal, the link will have high information scent for that user and her job, and she will be likely to click it.

That is the main reason for which we repeatedly argued that link names should be articulate and self-explanatory.If the link name is too obscure and vague, people might miss a skillful source of information. Even if the link label is precise and accurately describes the webpage it points to, information technology may even so miss the marker and have low information scent if it contains words that are not easily understood by the target audience. Jargon, branded terms, or simply likewise sophisticated words may terminate upwardly ignored and may not provide enough understandable cues for all your users.

Cisco.com: The headline above uses many branded terms that people who are not familiar with its technologies will not empathize. NGFW stands forNext Generation Firewall (Cisco'due south name for its firewall applied science),AnyConnect stands for its VPN service, and Duois an integrated multifactor-authentication service.

Note. In UX we often use the phrase "a label (such as More or Learn more) has low data scent." What nosotros mean by it is simply that, whatever your information need may be, it'south hard to guess what the link may atomic number 82 to. We likewise sometimes say that "a label (such as News) has high data scent" when information technology accurately and comprehensively describes what it leads to. Technically, the second usage is incorrect — the characterization will just take high information scent for those users who are looking for news, just depression information scent for people who search for something else. However, even for those people with a different information need, a good label description is useful and valuable considering it saves them the endeavor and thwarting of clicking on a page only to find out that information technology's not what they demand.

Content that Accompanies the Link

Often, next to a link there may be a short text snippet or a thumbnail intended to nowadays the user with additional information. Even though users may not read all the text associated with a link, they may still scan information technology and glean additional cues from it. Those cues will augment the information smell of that link.

This fact has ii big implications:

  1. Summary text for an commodity or page should convey the gist of that information source and add detail to the link label.

    A poor summary is a wasted opportunity for the site and wasted time for the user: the site misses the chance to tell the user whether the article volition be relevant. As yous tin can run across in the Cisco example above, the summary of the headline adds cipher to the obscure terms in the title.

  2. The image associated with a link should always exist descriptive and representative for the page content or for the category it stands for.

    Likewise often sites choose generic images that are but loosely related to the content of the page.

    Sutterhealth.org: The images associated withNotice Walk-in Intendance and Find Urgent Care are generic, purely decorative, and don't add any extra cues to the label. The summary text under these links, however, does explain the departure between walk-in care and urgent care (and shows the keywords in bluish).

    Only even choosing a related image is not always good enough — especially if the prototype stands for a category of objects. The academic literature on categorization (going back to Eleanor Rosch's studies in the 1970s) shows that not all members of a category are created equal. Thus, people take longer to interpret images of chickens equally birds than images of robins because chickens are less representative of the "bird" category than robins. And so, when choosing images for your categories, don't utilise aesthetics or convenience as your only criteria. Call up of a category member that truly illustrates the set of objects that it's supposed to stand for.

    Costco.com: The image chosen to stand for the categoryJava & Sweeteners is hardly representative of what most people call back of when they recall of coffee. Thus, for a user searching for that category, the prototype adds petty data scent (unless the user happens to exist looking for the exact product shown in the epitome).

Context in Which the Link Appears

Frequently, what else is on the page volition also influence how a link is perceived (or if information technology seen at all). For example, for the same information need, the word "Christmas" might have different information olfactory property on two unlike websites such as HarryandDavid.com and Williams-Sonoma.com. Even if you've never heard of the sites before, if you're looking for Christmas table settings, the olfactory property will be loftier on Williams Sonoma's site considering of the other related  content also visible on the folio, simply information technology volition exist low on Harry&David's site for the same reason.

If y'all're looking for Christmas tabular array settings, the aforementioned link labeledChristmas will accept unlike data scents on ii different sites. On Harry& David'due south site (top), the other images on the page, as well as category names such equallyBakery, Gift Baskets & Towers, or Pears & Fruit will make the information aroma ofChristmas depression. But on Williams Sonoma's site, the images of dishes and tea kettles will increment the information scent of the link.

Although context is usually a stiff cue, designers often make the mistake of relying too much on it. Especially on small screens, the context is non ever entirely visible or may get ignored (peradventure because people may curl quickly by it in search of something relevant). So, information technology's ever good to be as specific equally possible with the link names instead of relying on the context to provide the extra cues.

On mobile, it's harder to empathise what the word Christmas on Williams Sonoma's site (left) is about because a lot less of the context is visible. Users must recall what they've seen before or use any knowledge they have near the visitor. Harry & David (right) changed the name of the link on mobile toChristmas gifts to increment the label's specificity. (Whether different category names on mobile vs. desktop is a proficient thought is a topic for a different article.)

The other frequent mistake that designers brand is not providing sufficient context soon plenty. Too often we see landing pages with very footling text and a big image on the kickoff screenful. Fifty-fifty when those pages practice contain the correct data for the user, in many situations they don't provide enough context to tell people whether they are on the right track. As a result, users don't bother scrolling anymore in search of the right data, nor exercise they click on whatsoever visible link: they chop-chop decide that the page is not worth exploring whatsoever longer and simply leave. Thus, the proficient information aroma that the link to the site may have provided (for example, on a search-results page) is wasted by the poor context provided past the actual page.

PosturePillow: The first screenful of the landing page contains a big paradigm of a bed. The company name (present in the logo) and the somewhat ambiguous image are the only cues to the blazon of products sold on this site. For the hurried visitor coming here from a search engine, these may not exist enough to bother to scroll downwards on the page or click on one of the links.

The context can too include the position of an item on the page. Often information presented in the right rail may be interpreted incorrectly as an ad, even though the link description is explanatory enough — just because people take learned that ads are hosted in the right-manus side of the page, and and then whatsoever information is there will lose in relevance.

What the User Knows: User's Prior Experience

Another component of information scent is the knowledge that the user has accumulated in the past either directly — over her own previous experiences with the company, with the same type of content, or simply from using the web — or indirectly, from word of mouth or recommendations from friends or strangers.

Here are some elements that are part of the user's prior knowledge:

Familiarity with the brand and trust in it.  If y'all already know of Williams Sonoma and mayhap take interacted with the brand before, you will be able to understand what "Christmas" on its website stands for even in the absence of other page context. Or, if in the past you have had good experiences with Cisco, you may click on the link to ane of its products, even though the link characterization is non very descriptive.

Familiarity with the domain.  If someone is applying to college equally an undergraduate, he may know that the common data set up (which contains statistics that universities publish about each year's class of incoming undergraduates — detailing things like range of exam scores, GPAs, as well as different admission criteria) is normally in the section of the academy site that hosts finance-related information. So, for those people, a link label such as Office of Vice Chancellor of Finance may have loftier information scent, simply for someone who is hearing well-nigh the mutual data ready for the offset time, that link proper noun may not be at all transparent. (And neither volition the name of the "mutual data set" itself. Thus, a novice user must overcome two levels of poor data scent, making it extraordinarily hard to find the information.)

The common information fix, which contains statistics well-nigh the admitted freshmen likewise as information about admission criteria, is often hosted in the finance section of the university website, as seen above on UC Berkeley's (meridian) and Caltech's (bottom) sites. But this type of background information won't be available to someone who has no prior knowledge well-nigh college admissions.

Social Foraging: Word of Rima oris and Recommendations from Others

The social-foraging theory is an extension of information-foraging theory that explains how networks of people collaboratively forage for information. These networks could exist organized (e.g., communities of scientist working together on the same trouble) or advertizing hoc (e.g., Wikipedia contributors, reviewers on Amazon, taggers on a collaborative tagging system). The idea is that every bit people look for information or interact with data, they get out traces for others about the quality of various information source. These traces effectively augment the information scent for other users.

Thus, allow's say you are looking for a hair straightener on Amazon. Y'all choose i and read through the reviews. One extensive review says that this pilus straightener is good, but not as adept as another i. As you are going back to the page of search results, the information that you now take about the 2d product will often bias yous to click on it and consider information technology, even if information technology's more expensive than the showtime. That's because that product carries now additional information olfactory property, offered by the reviewer of the get-go pilus straightener.

A review indicates that some other hair straightener is ameliorate than the one that the user is currently looking at (top). That review adds to the user's knowledge and makes it more likely that she will select the recommended straightener from the production list (bottom).

Why Click Bait Doesn't Work in the Long Run

Now that you lot've understood the dissimilar components of data scent you may be tempted to game them in order to attract users to your site even though it may not be fully appropriate for their needs. (This approach is often driven by vanity metrics such equally number of clicks.)

For instance, y'all may be tempted to come up with an intriguing title that matches a trendy topic for a tiresome article that is non even remotely connected to the topic.  But this arroyo can backfire. Yep, you lot will get your click, but at the same time, you will employ upwardly your visitors' trust. Like in the story of the boy who cried 'wolf' too many times, next fourth dimension when you will really have relevant content, people will be less likely to click on it knowing that they've been burned in the past. And even worse, although they may not have been burned personally, they may read complaints from other people who had a bad experience. So you may not fifty-fifty get that first click.

The other side of the coin is that, if your brand is strong and people can trust you lot, you lot have slightly more room for error (not a lot — many mistakes will ultimately erode your brand, because brand is feel in interactive media). But basically if you can increase the expectation that users will find what they need on your site and that, whenever you promise something with a label or an epitome, you will deliver — then people will be more probable to give you the benefit of uncertainty for the occasional misstep.