Category Archives: User Behavior

Questioning the Power of the Influencer

First published October 2, 2008 in Mediapost’s Search Insider

Word of mouth is powerful in marketing. In the last two weeks, we’ve seen how the opinions of others can cause us to change our own beliefs to match. We’ve also seen how the speed at which the word spreads is a function not only of the structure of the network itself, but also the value of the message and its impact on the people in the network, as well as how much they stand to gain (or lose) by spreading the word.

Influencers: Our Connection to Opinion?

In the world of marketing, one of the most cherished concepts has been the idea of an influencer or opinion leader, the super-connected individual who acts as a hub in an information cascade, rapidly disseminating the idea to many. According to this theory, most of us (90%) play relatively passive roles in information cascades, meekly accepting the opinions of these influencers and following the herd. Katz and Lazarsfeld introduced the two-step influencer model in the middle of the last century, showing how media first influences these influencers, or opinion leaders, who then act as a conduit and “infection agent” for the greater population.

It’s Not the Influencer, It’s Our Willingness to be Influenced

For the past 6 decades, marketers have allocated a lot of effort in reaching these influencers, assuming that once you capture the influencers, you capture the entire market. The assumption was that information cascades depended on these influential hubs. Malcolm Gladwell’s “TheTipping Point” brought this phenomenon to popular attention.

In the past few years, a number of researchers, including Duncan Watts from Columbia University, have questioned the impact of influencers on information cascades. They’ve created several network models which have shown that in most cases, ordinary individuals are all that’s required to trigger a word-of-mouth cascade. We are not merely sheep following the herd. We are all influencers in our own right, but only when we feel strongly about something. The necessary ingredient is not a hyper-connected influencer or super trend-setter, but rather a group of people willing to be influenced.

Passion by Word of Mouth

Which brings us to Mel Gibson’s “The Passion of the Christ.” When promoting the film, Gibson knew the most receptive audience would be church-goers. So he arranged for private screenings and the distribution of free tickets in churches throughout North America. We had Watts’ ideal model, a low variance network (similar levels of influence) that shared a vulnerability to influence, given the nature of the message. Word spread quickly before the launch of the movie (which also resulted in a firestorm of controversy), making “The Passion of the Christ” one of the most successful movies of 2004.

This example also leads us to a possible error in analysis of information cascades that has perpetuated the “influencer” theory. It’s relatively easy, when looking in hindsight, to make the assumption that if a cascade happened, the individuals at the beginning of the cascade had to be unique in their ability to influence others. A proponent of the Influentials Theory could look at the example of “The Passion of the Christ” and say that it was the pastors and ministers of the selected screening churches that acted as the influencers, spreading the word to their congregations.

But Watts’ theory offers an alternate explanation. The everyday, commonly connected members of the audience were willing to be influenced, and once captured by the message, went and spread it within their other social groups. It was the willingness to be influenced that was the critical factor. To use the analogy provided by Watts in his paper, assuming some unique level of influence by the catalysts of a cascade is like assuming that the first trees to burn in a forest fire are somehow able to spread flames farther than other trees. Often, the fact that the tree was combustible in the first place is overlooked.

Starting a Brand Fire

So, when we talk about brand, what makes a tree ready to catch on fire? Here we have another important insight from Watts’ work. Too many marketers make the assumption that influencers are the critical component of success. Proctor and Gamble has made influencer marketing a cornerstone of its strategy. But the fact is, if “The Passion of the Christ” was an unremarkable movie that audiences couldn’t connect with, all the influencers in the world wouldn’t have caused the word to spread. It was a powerful message connecting with an audience primed to accept it.

Watts’ models show that the success of a cascade depends on the vulnerability to influence. If that is present, ordinary individuals can cause the word to spread as far and just as quickly as hyper-connected influencers. And the vulnerability to be influenced, the “combustibility” of the audience, depends on many factors, perhaps the most important of which is the back story of the brand.

The Combustible iPhone

Look at what has been one of the most successful cascades of recent times: the Apple iPhone. The iPhone is a tremendously combustible product. It’s not technology mavens causing the word to spread (although they do have influence. Watts is quick to point out that they have impact, but it may not as disproportionally large as everyone believes), it’s the person sitting next to you on the plane who says she loves it. And we’re receptive to that message because we have that magic connection of brand (Apple makes cool products) and a remarkable product. We’re ready to be set on fire.

I’ve spent the last few columns detailing the aspects of word of mouth because they have a tremendous impact on brand and how we create our own brand beliefs. And it’s these brand beliefs that are triggered when we interact with search results. Next week, we return to more familiar territory and see how this interaction plays out.

Traveling at the Speed of Buzz

First published September 25, 2008 in Mediapost’s Search Insider

What makes up buzz? And what determines how fast it travels? Last week, I talked about how important the opinions of others are in shaping our brand beliefs. Today, I want to look at one category of word of mouth, the juicy tidbit, recently christened “buzz,” and see what makes it leap from person to person.

Buzz is Nothing New

For some reason, we think buzz is a new thing that lives online. In fact, it’s as old as human behavior and has its roots in our very social fabric. We need to pass on information. We’re driven to do so. We gossip because it’s inherently satisfying, both to ourselves and to the recipient. But the spread of gossip through a social network is neither uniform nor consistent.  In the ’70s, Mark Granovetter discovered that, like many things, social networks are patchy, made up of tightly linked clusters of people who spend a lot of time together (families, friends, co-workers) which are loosely connected to each other through “weak ties,” more distant social relationships. The survival potential of a viral piece of information (Richard Dawkins first coined the term “meme” as a cultural equivalent of a gene in his book, “The Selfish Gene”) lies in its ability to jump Granovetter’s weak ties.   If the meme doesn’t jump out of a cluster, it ceases to propagate itself and can die an isolated death.

It’s Not Just the Network

In 1993 Jonathon Frenzen and Kent Nakamoto launched an interesting study showing that the ability of a “meme” to spread through a social network depended not only on the structure of the network (the main point of Granovetter’s work) but also on the impact of the meme’s message on the carrier (akin to the idea of a phenotype in genetics) and the value of the meme itself.

Frenzen and Nakamoto worked with three different variables: First of all, they altered the value of the message. In the first variation, it was news of a 20%-off sale, in the other variation; it was the more valuable news of 50% to 70% off. Secondly, they varied the amount of product available at the sale price. In one case, there was unlimited inventory. In another, the supply was very limited. Finally, they varied the structure of the network itself, in one case having a network of strong ties, and in another, strong tie clusters linked by Granovetter’s weak ties.

What they found was that the value of the message (20% off vs. 50% to 70% off) has a significant impact on the rate in which the word spread, as did the availability of items at the sale price. The second factor introduced a moral hazard aspect. It made spreading the news a zero-sum game: if I tell you, I might lose out.

Frenzen and Nakamoto also found that in strong tie clusters, word seemed to spread relatively quickly regardless of the nature of the news. There were variations, but in all cases, the majority of the strongly linked network came to know of the news fairly quickly.

Social Speed Traps

If the discount was fairly low, the news tended to get stuck within clusters and had difficulty jumping the weak ties. If the news was valuable (50% to 70% off) and supply was virtually unlimited, the news was much quicker to jump the weak ties, spreading through the network very quickly. But, if the discount was large and the supplies were limited, suddenly the news tended to get trapped within the strongly tied clusters. People were reluctant to spread the news because the more people that knew, the more it was likely that they and their close family and friends (the people within their strong tie clusters) would lose out on a great deal.

Weak Ties on the Web

In both the online and offline worlds, the speed with which buzz will spread depends on the value of the message (is the gossip juicy? Is the price unbelievable?) and how much we stand to gain or lose (does sharing reduce the chances of me and my close circle getting ahead?). Gossip’s primary purpose is to create social bonds, and the sharing of intensely interesting information is something we’re programmed to do. Similarly, we’re programmed to share opportunity with those closest to us, either through kin selection (we want those with whom we share the most genes to get ahead first – W.D. Hamilton did the foundational work on this) or reciprocal altruism (doing a favor for a friend knowing that at some point, we’ll benefit from the payback — Robert Trivers is the name to search for if you’re interested). In most cases of online buzz, there is no moral hazard. In fact, unless a meme has what it takes to jump the weak ties in a real-world social network, it will never make it onto an online forum. Posting on the Internet is, by its very nature, a weak tie, a reaching out from ourselves to everyone.  We don’t publically post memes if it costs our strong ties the opportunity to capitalize on them. Similarly, we’re less likely to post unremarkable news, although I’m still trying to reconcile Twitter and Facebook status updates with this theory.

So, in the world of social networks, some people have more influence than others, right? Some are mavens, or super connected hubs, or natural salespeople (borrowing from Gladwell’s “The Tipping Point”). Not so fast, says Columbia University’s Duncan Watts.  But more on that next week.

False Memories: Was that Bugs Bunny or Just My Imagination?

First published September 11, 2008 inn Mediapost’s Search Insider

I’ve talked about how powerful our mental brand beliefs can be, even to the point of altering the physical taste of Coke. But where do these brand beliefs come from? How do they get embedded in the first place?

A Place for Every Memory, and Every Memory in its Place

Some of the most interesting studies that have been done recently have been done in the area of false memories. It appears that we have different memory “modules,” optimized for certain kinds of memory. We have declarative memory, where we store facts. We can call these memories back under conscious will and discuss them. Then we have implicit memory, or procedural memory, that helps us with our day-to-day tasks without conscious intervention. Remembering how to tie your shoes or which keys to hit on a keyboard are procedural memories

Declarative memory is further divided into semantic and episodic memory. In theory, semantic memory is where we store meaning, understandings and concept-based knowledge. It’s our database of tags and relationships that help us make sense of the real world. Episodic memory is our storehouse of personal experiences. But the division between the two is not always so clear or water-tight.

The Making of a Brand Memory

Let’s look at our building of a brand belief. We have personal experiences with a brand, either good or bad, that should be stored in episodic memory. Then we have our understandings of the brand, based on information provided, that should build a representation of that brand in semantic memory. This is where advertising’s influence should be stored.

But the divisions are not perfect. Some things slip from one bucket to the other. Many of our inherent evolutionary mechanisms were not built to handle some of the complexities of modern life. For instance, the emotional onslaught of modern advertising might slip over from semantic to episodic memory. There will also be impacts that reside at the implicit rather than the explicit level. Memory is not a neatly divided storage container. Rather, it’s like grabbing a bunch of ingredients out of various cupboards and throwing them together into a soup pot. It can be difficult knowing what came from where when it’s all mixed together.

This is what happens with false memories. Often, they’re external stories or information that we internalize, creating an imaginary happening that we mistakenly believe is an episode from our lives. Advertising has the power to plant images in our mind that get mixed up with our personal experiences, becoming part of our brand belief. These memories are all the more powerful because we swear they actually happened to us.

That Wascally Wabbit!

University of Washington researcher Elizabeth Loftus and her research partner Jacquie Pickrell have done hundreds of studies on the creation of false memories. In one, under the guise of evaluating a bogus advertising campaign, they showed participants a picture of Bugs Bunny in front of Disneyland, and then had them do other tasks. Later in the study, the participants were asked to remember a trip to Disneyland. Thirty percent of them remembered shaking Bugs Bunny’s hand when they visited the Magic Kingdom, which would be a neat trick, considering that Bugs is a Warner’s character and would not be welcome on Disney turf.

We all tend to elaborate on our personal experiences to make them more interesting. We “sharpen” our stories, downplaying the trivial and embellishing (and sometimes completely fabricating) the key points to impress others. When we do this, we will draw from any sources handy, including things we’ve seen or heard in the past that we’ve never personally experienced. To go back to last week’s Coke example, our fond memories of Coke might just as likely come from a Madison Avenue copywriter as from our own childhood. We idealize and color in the details so our conversations can be more interesting. It goes back to the human need to curry social favor by gossiping. When you have this natural human tendency fueled by billions of dollars of advertising, it’s often difficult to know where our lives end and our fantasies take over.

This mix of personal experience and implanted images explains part of where our brand beliefs come from. Next week, I’ll look at the power of word of mouth and the opinions of others.

Brand Labeling: Building Our Beliefs

First published August 14, 2008 in Mediapost’s Search Insider

Up to now in this series on search and branding, I’ve been looking exclusively at how and why we use search engines. But the idea of the series is to show how branding and search can work together. So in this column, I’d like to start from the opposite end of the spectrum: our brand relationships, from a memory retrieval perspective.

Storing Complex Concepts

In the computational theory of mind, the prevailing theory that seems to best explain how our minds work (although it’s not without its detractors), the elegance with which the brain processes complex patterns of information is remarkable. These are called constructs, and brands are no exception.

For any complex concept, the components of the concept are individual and scattered memory patterns, called engrams. Engrams are groups of activated neurons that fire together. But the more complex the concept, the greater the network of engrams. For a person we know well, like our mother, we could have a huge number of scattered components that make up our concept. Snatches of memories, what her voice sounds like, what she looks like, what her banana loaf tastes like. All these, and many more, individual memory components make up our concept of “mother.” And these fragments are stored in various parts of the brain. When we remember what our mother looks like, it’s an engram in our visual cortex that fires, the same part of the brain that fires when we’re actually looking at her. We’re actually picturing her in our mind. When we hear her voice, it comes from our auditory warehouse.

Our Neuronal Warehouse

The concept of a vast neuronal warehouse is actually a good analogy. When we call up our concept of “mother,” it’s assembled on the fly from the individual sections of the warehouse. The retrieval call goes out, depending on the need, to the various parts of the brain, and the required components are brought together in our working memory and assembled in the conscious part of our brain. Each memory is custom made from available parts. If we were looking at a model of the brain, we’d see maps of neurons “lighting up” across the cortex, almost like a lightning storm seen from above the clouds.

But with a construct as complex and extensive in scope as our mother, there needs to be a shorthand version. We can’t retrieve every single piece of “mother” every time we think of her.  So, the parts retrieved are restricted to the context we do the retrieval in. If we’re buying a dress for our mom, we retrieve components that include her body shape, her color preference and probably memories of other things she’s worn in the past. We don’t retrieve her banana loaf recipe because it’s not relevant.

Executive Summaries of Memories

But there’s also a labeling process that goes on. For complex constructs, like our mother or a familiar brand, we need a quick and accessible “label” that sums up our feelings about the entire construct. This is the top of mind impression of the construct, the first thing that comes to mind. It helps us keep the world straight by providing a shorthand reference for the many, many constructs stored in our memory warehouse. These labels have to be simple. In the case of people, the summing up usually determines whether we like or dislike the person. It’s a heuristic shortcut that is built up from the sum of our experience and exposure which determines whether we’re willing to invest more time in the person. The same is often true of brands.

The power of these labels for brands is absolutely essential, because they determine our attitudes to everything that makes up the construct. The brand label, or belief, is a gut feeling that impacts every feeling or attitude towards the brand.

Top-of-Mind Brand Beliefs

Often when I’m speaking, I’ll do a little exercise where I’ll show well-known brand labels and ask people to write down the first thing that comes to mind when they see it. What I’m capturing is the brand label, the top-of-mind belief about the brand. Apple generally brings out labels like “cool,” “cutting edge” or “design.” Starbucks is labeled “indulgence,” “great smell,” “delicious” or, less positively, “overpriced.” The entire scope of our experience with the brand is labeled with a few words. Obviously, our entire concept of Starbucks is usually much greater than just the way it smells or tastes, but for the people that have assigned it this label, that’s the best overall descriptor and the easiest access point. The rest of the details that make up our concept of Starbucks can be unpacked at will, but for these people, they’re all packed in a box that is labeled with “great smell” or “delicious.” If the label is “overpriced,” this may be a box we seldom unpack.

Next week, we’ll continue to look at how we store our concepts of brands, what can make up our brand constructs and the role emotion plays.

Needs, Beliefs and Search

First published August 7, 2008 in Mediapost’s Search Insider

In the last few weeks, I’ve looked at how we gather information, depending on how complete the information is we already have. But it’s not just information that colors the search interaction. Like all human interactions, we are governed by our desires, our objectives and our beliefs, and this is certainly true in search.

Computing Concepts

Steven Pinker is one of the foremost proponents of a computational theory of mind. Following in the footsteps of Alan Newell, Alan Turing, Herbert Simon   and  Marvin Minsky,  Pinker argues that our “minds” lie within the patterns of information processing and functionality founds in the specialized modules of our brains. Like a software program being executed step by step, our minds break down the incredibly complex concepts we are faced with each day and feed them through these patterns. We create objectives that get us closer to our desires, and in order to get there in the most efficient way possible, we depend on a vast library of heuristic shortcuts that include our beliefs. We don’t think everything to death. We make quick decisions and create short cuts based on existing beliefs. Simon called this  bounded rationality.

Irrational Short Cuts

The challenge with these short cuts, as  Amos Tversky,  Daniel Kahneman, and more recently,  Dan Ariely, have discovered, is that they’re often quite irrational. Our beliefs are often driven by inherent patterns that have evolved over thousands of years. While they may be triggered by information at hand, the beliefs lie within, formed from a strange brew of inherent drivers, cultural influences and personal experience. In this brew, it’s almost impossible to see where one belief shaper begins and another leaves off. Our beliefs are largely formed in our vast mental sub-cortical and subterranean basement, below the hard white light of rational thought. But, once formed, beliefs are incredibly stubborn. Because we rely on beliefs to save our cognitive horsepower, we have an evolutionary interest in keeping them rigidly in place. Heuristic shortcuts don’t work very well if they’re based on ever-changing rules.

And there you have the crux of marketing. Every time we’re presented with a symbol that represents a concept, whether it be a word, a picture, a sound or a logo, it unlocks a mental concept complete with corresponding beliefs. Unless it’s a brand we’ve never heard of before (in itself a significant marketing challenge), that brand comes with corresponding belief luggage, some of it undoubtedly highly irrational. We are built to quickly categorize every new presentation of information into existing belief filing cabinets or “schemas.” The contents of those filing cabinets are difficult to explore, because they exist at a subconscious level. Consultants such as  Gerald Zaltman and  Clotaire Rappaille have carved out lucrative careers by creating methods to unlock the subconscious codes that lie behind brands.

Search and Our Subconscious Baggage

So, when we interact with a search engine, it’s important to understand that this is not entering new information onto a blank canvas. Each word (or now, image) on a search page has the potential to trigger an existing concept. This is especially true for the appearance of brands on a page. Brands are neat little labels that can sum up huge bundles of beliefs.

It’s actually amazing to consider how quickly we can filter through the degree of information presented on a search page. We quickly slice away the irrelevant and the items that don’t fit within our existing belief schemas.

It’s not just the information on the page that we have to filter through. It’s all the corresponding baggage that it unlocks within us. Somehow, through the power of our subconscious mind, we can scan 4 or 5 listings, let the words we scan trigger corresponding concepts in our minds, quickly evaluate which listing is most likely to get us closer to our objective (based on beliefs, aligned with our desire) and click, all within a few seconds.

This simple act of using a search engine is actually a very impressive and intricate cognitive ballet using the power of our conscious and subconscious minds.

Search Behavior: I Don’t Know What I Want, or Where to Find It

First published July 31, 2008 in Mediapost’s Search Insider

In my last two columns, I looked at how search plays a part when we’re in two information gathering states: I know what I’m looking for and where to find it; and, I know what I’m looking for but not where to find it. Today, I’ll look at what happens when we don’t know what we’re looking for or where to find it.

In the first two states, our intent is pretty well defined. We’re looking for a piece of a puzzle and we know the shape of that piece when we see it. In information-foraging terms, we’ve already defined our diet. It’s just a question of which patch we look in. When we extend that to search engine usage, we have already defined our query, and it’s just a question of how we interact with the results page. In both these states, search engines work pretty well.

The Missing Puzzle Piece

But what if we have no idea what the puzzle piece looks like. We don’t know the shape, we haven’t assembled the adjacent pieces and we only have some vague idea what the finished picture should look like. This is the ultimate challenge for online search, and one that all search engines have largely failed to meet until this point. This is where we need a guide and advisor, a connector between ourselves and the universe of potential knowledge available. Because our knowledge is imperfect, we need a sage whose knowledge is perfect — or, at least, much less imperfect than our own.

Of Disambiguation, Discovery and Berrypicking

This is where three concepts play an important role: the need to disambiguate, the thrill of discovery, and a revisit of Marcia Bates’ concept of berrypicking. Let’s begin with disambiguation.

When we have no idea what we’re looking for, we don’t know how to define it. We don’t know the right query to present to the search engine. The more imperfect our knowledge, the more ambiguous our query. This is where search needs better knowledge of who we are. It needs to know — through implicit signals such as our areas of interest, our past history and our social connections –what it is we might be searching for. If a search engine is successful in lending more definition to our query, it stands a chance of connecting us to the right information.

The second piece is discovery. If a search engine is successful in introducing potentially relevant information to us, our interaction is quite a bit different than it is in the first two information gathering states. We spend more time in our interaction and “graze” the page more. We’re also open to more types of content. In the first state (know what we want and where to find it) we’re just looking for the fastest navigation route. In the second state (know what we want but not where to find it) we’re looking for confirmation of information scent to judge the quality of the patch. But in the third state, we could be enticed by a website, an image, a news story, a video or a product listing. We’re pretty much open to discovering anything.

And this brings us back to Bates’ theory of berrypicking. Because we have no preset criteria for the type of information we’re looking for, we can change direction on the turn of a dime. In our pursuit, we fill in the definition of our prey as we go. We follow new leads, change our information-gathering strategies and sometimes completely change direction. Our interactions with search turn into a serendipitous journey of discovery. It is in the third state where our patience is generally higher and our scanning pattern the broadest. Any cues on the page that trigger potential areas of interest for us, including brands or cultural references, could catch our attention and lead us down a new path.

Search Pursues Discovery

It’s this type of search that Ask’s 3D interface or Google’s Universal results set was built for. It’s also the thorny problem of disambiguation that has spawned a number of approaches, from Google’s exploration of personalization to the human assisted approaches of ChaCha and Mahalo . Even Yahoo’s Answers is a discovery tool, using the more natural approach of question and answer to lend some definition to our information quest. But even though we are defining our criteria as we go, we still seek to conserve cognitive energy. We have a little more patience in our seeking of information scent, but just a little. We still spend seconds rather than minutes looking for it, and because search is still trying to get discovery right, our sense of frustration can mount rapidly. We’re still a long way from finding a universally satisfying online source for discovery.

Search Behavior: I Know What I Want, But Not Where to Find It

First published July 24, 2008 in Mediapost’s Search Insider

Last week I looked at search behaviors when we knew what we were looking for and where to find it (http://www.mediapost.com/blogs/search_insider/?p=832). This week we’ll look at what happens when you know what you’re looking for — but you’re not sure where to find it.

Judging a Patch By its Scent

In the first instance, when you know what you’re looking for and where to find it, you have defined your patch and you have a pretty good idea what route to take to find your specific piece of information. In the second instance, you don’t know in which patch you’ll find the piece of information. This is where classic way-finding behaviors and information scent can play a critical role in seeking information.

When you’re not sure which information patch contains the right information, you have to judge each patch by its relative “scent.” This pretty much guarantees you’ll visit more than one patch, which for our purposes translates to Web sites. You’ll try to do a preliminary assessment of scent based on what you see on the results page, but you’ll reserve most of your judgment for when you click through to the site.

Looking for Greener Grass

One of the interesting aspects of optimal foraging for food is that there are costs to move from patch to patch. You have to literally assess whether the grass is truly greener on the other side of the fence, or whether it would just be a senseless waste of effort. Most animals have a highly developed heuristic instinct about when the time is right to move on to the next patch. Biologist Eric Charnov, who reached out to me (I’m still following up with Eric to get a follow-up interview for a future column) after my original information-foraging column, called it the  Marginal Value Theorem. In a nutshell, Charnov’s Theorem says that we decide how long to stay in a patch based on how rich the current patch is and how distant the next patch is.

One of the challenges of the Marginal Value Theorem is that we often have no way of knowing what the “richness” of the next patch might be until we commit to expending the energy to go see it. We risk the effort based on our assessment of the current patch and the hope that better patches lie ahead. And the risk lies in the fact that it takes energy to move from patch to patch. The degree of risk lies in the distance to the next patch, our expectation of the richness of that patch and the value of the patch we’re currently in.

Patch Hopping with Search

But online, the Internet is non-dimensional in the traditional sense. There is no distance; the only dimension is time. How much time are we willing to expend to find the next patch? And search gives us the ability to greatly reduce the time needed to navigate from patch to patch. We structure queries to define the “diet” we hope to find in each patch. We then can click through to see if the scent matches our definition of diet.

Remember, time is the resource we hope to conserve, so these explorations from the search page are very quick. We can visit a number of patches in seconds. We define the diet (what we’re looking for) and start down the page visiting the most promising patches. Based on user research we’ve done at Enquiro, searchers typically take 10 to 12 seconds for the first click from the search results page, and spend about 15 seconds assessing the scent on the pages they visit.

Because we are programmed to save effort, if we visit a few patches and come up short, we’ll use a new query to define a new collection of patches. Because we have no defined notion of which patch will be the right one, we have to use shortcuts to judge each patch quickly and efficiently. We have little patience for unpromising patches.

Of course, our level of patience is also determined by how rare the prey is we’re pursuing. If we believe it should be rather plentiful, we also believe the scent should be easy to pick up. But if our prey is elusive, we’ll be more patient in our quest to pick up its scent. Those are the searches that will drive us to the second or third page of results.

We Don’t Consume Information

If we find a rich patch, we file it away for future consideration. This is another area where information foraging diverges from biological foraging. Looking for food is a zero-sum game. If we don’t eat the food we find, someone else will. So when we find a rich patch, we stay put until we eat as much as we can (or until a richer patch beckons).

But online, information is not really consumed. Even if we use it, it’s still there for the next visitor. There’s no risk to move on and find other information patches. This is where traditional way-finding strategies come in. As we explore for information, we define the landscape based on the richest information patches. These become landmarks which we return to again and again. So we quickly use search to define the best patches and tag them for future reference. Then, we return to them at our leisure, knowing the information will still be there, waiting for us.

Next week, we’ll looking at the third state of information-seeking — where we don’t know what we’re looking for or where to find it — and how this impacts our search behavior.

 

Search Behavior: I Know Just What I’m Looking For

First published July 17, 2008 in Mediapost’s Search Insider

We seek information to fill gaps in our existing knowledge. The extent of that current knowledge and how we’ve structured it will play a large part in determining intent. It will help shape our knowledge requirements, our strategies for retrieval and how we will evaluate information scent. As stated in my previous two columns, we’re generally in one of three states when we turn online for information; we know what we’re looking for and where to find it; we know what we’re looking for, but not where to find it; or, we don’t know what we’re looking for or where to find it. Today, I want to explore intent in the first of these states:

We Know What We’re Looking For — and Where to Find It

In the first case, we have a solid idea of the information we’re looking for. Our mental representation has a defined structure and we have a good idea of what the missing piece looks like. For example, we’re looking for a phone number, an address or another missing detail. Because the structure of the information in our minds is almost complete, we have a similarly clear cut idea of where we’re most likely to find the information. We know the right “patch” to look in and where to find the information in the “patch.” In this case, we’re looking for the simplest route from point A to point B.

Googling Google on Google

One of the ongoing anomalies in search is the number of people who go to their favorite engine to search for proper domain names. Some of the most popular queries on every engine are the URLs of their competitors. People search for Yahoo.com on Live Search, or Google.com on Yahoo. People even search for Google on Google. In looking at the query logs, the only explanation seems to be mass stupidity. But in actual fact, this is foraging playing itself out. We habitually use engines to navigate the Web, so even when we know the Web site, why should our behavior be any different? (This still doesn’t explain the searching for Google on Google. Perhaps stupidity is the right answer here.)

Let’s say you’re looking for the address of the head office for a corporation. You know it will be on their site somewhere, and you have a pretty good idea it will be somewhere within the “about us” section. Rather than go directly to the site and navigate through it, you choose to search for “Company X head office address.” Or, even more likely, you just search for “Company X,” knowing that the official site will come up high in the results.

Pre-Mapping the Search Results Page

In this case, before the results page even loads, you know exactly what you’re looking for and where you’re likely to find it. If you’re searching on Google, it’s likely that you’ll get an extended result in the number one organic spot with Site Links to key parts of the site. This is a great match for your expected information scent. Previous to this introduction by Google, we saw that for navigational searches where we knew the destination we were looking for, there was a higher degree of scanning of the site URL at the bottom of the result listing. Normally there’s not much interaction with this part of the listing.

In this first category, we look at search as a tool, the quickest possible route to the information we know exists. We will quickly zero in on the only relevant information on the results page, the listing for the site we’re looking for. Now, the question for marketers is, what happens when there’s both an organic and sponsored listing for the same site on the same page? Will one cannibalize the other? While we’ve never tested for this specific intent, I can speculate based on what we’ve seen in other research.

Habitual Scanning Behaviors

In one study, we split our group in half, giving one a purchase-type task and the other an information-gathering task. In both cases, we looked at scan patterns in the top sponsored and organic results. We expected to see our information-gathering group relocate their scanning down to the organic results. But this didn’t happen. What we realized is that we scan the search page out of habit. We’re not rationally optimizing our scan path based on intent. We’re following the same pattern we always do, the top to bottom, left to right, F-shaped pattern that’s common across all users. That behavior is conditioned and engrained. It’s been etched at the sub-cortical level of our brain in our basal ganglia and executes subconsciously (see Ann Graybiel’s work  on this for more). But what does change is how we respond to the information scent cues on the page.

Although scanning followed the same pattern for both groups (in fact, the interaction was even higher with sponsored listings for the information-gathering group, likely because they weren’t exactly sure what they were looking for and so were in a more deliberate mode) the click patterns were significantly different. The official site that marked the successful destination in the scenario was in both the top organic and sponsored location. In the commercial task, the clicks split almost 50/50, with half happening in the sponsored listings, and half in the organic listings. But in the information-gathering group, all the clicks happened in the organic listing. Based on our preconceived idea of the information we were seeking, that particular “patch” seemed more promising.

Next column, we’ll look at intent and how it impacts search behavior when we know what we’re looking for, but not where to find it.

 

Foraging for Information with Search

First published July 10, 2008 in Mediapost’s Search Insider

In my last column, we looked at berrypicking as an analogy for gathering information. The theory was put forward in 1989 by Marcia Bates. Then, in 1995, two researchers found even more inherent behaviors demonstrated in the way we seek information. It turns out that we may literally hunt for information.

The Genetic Case for Searching

We didn’t come equipped with an inherent strategy to pull information from a Web search results page. There is no genetic coding specific to Google. But as two researchers at Xerox’s PARC research facility, Peter Pirolli and Stuart Card, really started to explore how humans navigated online environments looking for information in 1995, they found something fascinating. They found that the way we seek information online is very similar to an activity that is as old as evolution itself: the hunt for food. Pirolli and Card called this the information foraging theory.

The basic principle behind information foraging is not so much about gathering the maximum amount of information, but rather in maximizing our time and efforts in pursuit of the right information. This goes to the human knack for conserving our resources in pursuit of our objectives.

The Easiest Route to Information

We must remember that any interaction with a search engine is part of a much broader range of activity that will hopefully result in achieving a large objective that is aligned with a human drive: learning, bonding, acquiring or defending (Nohria/Lawrence ). We take these macro objectives and break them up into distinct tasks and allocate resources against those tasks based on the expected usefulness of the outcome. This is where the food-gathering analogy provides some useful perspective.

We eat food to survive. Food is the fuel that powers our activities. In the stripped-down logic of evolutionary survival, it doesn’t make sense to expend more energy in the pursuit of food than the food itself contains. We would starve and die. So we have become remarkably effective at finding food in the easiest way possible. The big objective of the pursuit of food and survival is broken down into discrete tasks or actions, and we instinctively determine how much time and effort to spend on each of these tasks or actions depending on how much closer it will get us to the objective: our next meal. There is a cascading series of risk/reward decisions and mental trade offs happening below the level of our rational awareness. Our evolutionary programs play themselves out subconsciously.

Born Foragers

While seeking information is a more abstract concept than finding food, Pirolli and Card argue that the same inherent skills are used, including the same trade-off decisions. In evolutionary terms, our information-seeking skills are an adaptation of our food-gathering skills. Each time we seek information, we “hunt” for it and make decisions about how much cognitive energy we want to expend in the pursuit and the optimum strategy for gathering the information. We forage for information.

This explains much of the typical behavior we see with online properties, especially search. We quickly seek and filter through information, using our heuristic guidelines and trade-offs.

And when we look at our use of search engines, there are two important concepts put forward by Pirolli and Card that must be considered: the importance of information patches and diets.

The Right Patch and The Right Diet

As we seek information, the same as seeking food, we will spend our time where the promise of successful pursuit is the greatest, based on clues or telltale hints we encounter. We look for the best information “patch,” which is determined by information “scent,” the smell of informational relevance. The greater the scent, the greater the promise of an abundant information patch.

Search engines give us the ability to create our own patches, somewhat like a spider spinning a web to catch prey. We see what we catch based on the scent, and if we don’t like what we see, we quickly spin another web with another query. There is almost no effort expended in the process, so we have little patience if we’re not presented with adequate scent. This is why so much time is spent scanning the top of the page. I call it the area of greatest promise, that tiny yet critical patch of real estate in the extreme upper left corner of the search page, where we expect the strongest scent, figuratively. We judge the value of the whole patch based on what we see in the first few words in the first few listings on the page. If we don’t find strong scent, we start questioning the value of the patch.

But we also have to make a determination of which information we include in our diet. Remember, it makes no evolutionary sense (assuming we’re using the same mechanisms we use for foraging food) to expend energy pursing food that doesn’t return an equal or greater return on our investment. So we will quickly filter out low-quality information. In fact, if we think a patch contains only low-quality information, we’ll exclude it from our diet.

Search has been remarkably successful in becoming the preferred “patch” for a diverse set of information needs, but it still comes up short in one particular category. It doesn’t do very well at helping us find information when we don’t have a clear idea of what we’re looking for. Search is still rather ineffective as a “discovery” engine. But despite its limitations in this area, we have still been increasingly conditioned to turn to search when we forage for information because of its remarkable efficiency.

Berrypicking Your Way through the Web

First published July 3, 2008 in Mediapost’s Search Insider

In the last three columns, I explored the fundamentals of humans seeking information. To refresh your memory, the purpose of this series is to explore the importance of branding on the search page. Taking several steps back to begin my run at this topic (and risking a series that is “long and extremely wind baggy,” in the words of one reader), we’re now starting to get at some of the important concepts to understand how we interact with a search results page.

Bates and Berrypicking

In 1989, Marcia Bates took a fresh look   at the classic model of information retrieval that had dominated for the previous 25 years. The model was a fairly straight equation, with on one side a collection of documents and how the contents of those documents were represented, and the seeker’s information need and the query they constructed to express that need on the other side. In the middle was the desired outcome, the match of query and document representation. Bates found this admirably simple equation didn’t hold up too well in real-world search situations, especially given the advances in information technology.

The problem with the classic model was that it assumed that the successful search for information was a relatively static event, where one search and retrieval strategy took you eventually to the desired information. Even if you took into account feedback and iterative query refinement, it still looked at the process as a continual and linear one, with incremental progress towards a constant goal. In looking at actual behavior, Bates found that the process was more complex. As we pursued the information we thought we wanted, she found the path was less a straight line and more a looping and meandering path. In fact, it reminded her of picking huckleberries, hence the title of her theory, berrypicking.

Meandering through the Web

Bates found that as we start down the path to the information we seek, we pick up bits of information a little at a time, like picking berries. What’s more, as we pick the berries, we may head off in different directions depending on the information we gather. We follow the berries to more promising clusters of berries, or “patches.” We don’t just refine our queries, we change search and retrieval strategies, the places where we think we’ll find information (our “patches”) and, in the more extreme cases, our ultimate destination. Search is an evolving behavior, not a linear one.

Bates looked at 6 different strategies that academics use to search for information: footnote chasing (backwards chaining from articles of reference, tracking back footnotes); citation searching (forward chaining, using a citation index to jump forward); journal run (using authoritative journals on a subject and going through the entire run); area scanning (using the physical location of a subject in a library on the assumption that relevant materials will be in the same location); abstracting and indexing searches (using organized bibliographies and indexes, usually arranged by subject area); and author searching. At the time of her paper (1989), Web search was still unknown. The first search engine (Archie) would be created in 1990 at McGill University. But as you look at the six methods outlined, it’s clear that Web search lets you do any and all of them.

The Web: The Ultimate Berry Patch

Bates theorized that berrypicking would play out in different environments and you would change strategies as you went from environment to environment. The timeline could be days, weeks or even months. But with the Web and search, you could go from strategy to strategy in seconds, berrypicking your way through the Web. What’s more, you could be diverted from your original path through a serendipitous display of information that catches your attention. For example, you could be footnote chasing (i.e., the source link for a snippet of a review on one page) which leads you to launch a search for other reviews. There you see results for a magazine dedicated to the topic (an example of journal run), a link for other consumer reviews (citation searching) and a book title written by an authority on the subject (author searching). You’ve just used all six of the strategies outlined by Bates in one session, as you used abstracting and indexing (this is ultimately what a search engine is) and area scanning (in this case the physical collocation is defined by the search page real estate).

When it comes to the impact of branding on search, it’s important to understand Bate’s berrypicking model. Any search result could represent a “berry” that could lead us to an entirely new patch. Our search path could evolve in a totally new direction based on what we pick from a page. We can be introduced to brands or have them reinforced as we berrypick our way along.

In the next column, we’ll continue to redefine information retrieval by looking at the Information Foraging theory.