"How can a part know the whole?"
In anatomy, we divide to understand. We dissect the whole to study its parts. We identify internal organs and map their relationships. As a major branch of biology, anatomy reflects both the power and the limits of specialization. For we must not allow our focus on form and structure to distract us from function or blind us to context. Anatomy can't tell us how the mind works. It can't reveal the sublime experience of vision. And it certainly can't predict the behavior of an ant colony or a stock market. These complex adaptive systems exhibit macroscopic properties of self-organization and emergence. Not only is the whole greater than the sum of its parts, but it's also different. It's a territory off the map. And yet, our simple models have value, for they offer us a very good place to start.
Our map to search features five elements: users, creators, content, engine, and interface. Like any map, it hides more than it shows. It's deceptive by design. It shifts attention from software and hardware to the elements of user experience. Our plan is to study each element without losing sight of the whole. We must know enough about the technology to understand what's difficult and what's possible. But we need not become intimately familiar with load balancing, pattern matching, and latent semantic indexing. That's why we have specialists. Instead, we'll study the components and context in sufficient detail to inform strategy and design. We'll survey the terrain in search of the big picture. So, let's start at the very beginning with the users for whom we design.
There's a lot we can learn about our users. Demographics cover income, age, and gender. Psychographics reveal values, attitudes, and lifestyle. Technographics segment user populations by their adoption of tools and software. Some of this data is useful for designing better search systems. Much of it is not. It's all too easy to stuff a treasure chest with worthless facts and figures. Organizations do it all the time. The key to profitable user research is knowledge. We must know enough to ask the right questions. We must understand the basics of user psychology and behavior as they relate to the type of system we plan to build. We require a conceptual framework that lets us focus on the pivotal questions that make the difference in design. We need a scalpel, not a hatchet.
For instance, we know the paradox of the active user is a constant in search. Most people refuse to read the manual, personalize the system, or prepare a strategy before they begin, despite evidence that such initial "presearch" improves overall efficiency. We enter two or three keywords and we GO. We're seduced by the illusion of speed. It's only when we find we're lost that we check a map or ask for help. Of course, some users love manuals, while others must study them in training. But for most users in most contexts, this paradox is active. Knowing this helps us to excise unfit questions and designs.
Another timeless topic is the question of precision versus recall. Do our users care more about finding only the relevant results or all the relevant results? In search design, the two are inversely related. Like kids on a seesaw, when one goes up, the other comes down. High precision generally means we miss some of the good stuff, while high recall forces us to sift through the good, the bad, and the ugly, except when a better algorithm or a richer interaction model lets us "bend the board" by amplifying the signal without adding noise. Either way, it's worth asking, especially since the answer may signal a pivot point where user and business goals diverge. In e-commerce, for example, a user may want to find a specific product as quickly as possible, whereas a vendor may allow for more noise, hoping that cross-selling will spur an impulse buy. It's critical to identify and manage this predictable source of tension in the user experience.
We should also consider expertise with respect to search in general and the domain in question. Let's say we're building a search application for health and medicine. Will most users (or our most important users) be familiar with medical terminology? And how about their digital literacy? Are they fluent or fumbling? It's a common mistake to conflate these two types of mastery. People assume that good doctors are good searchers, but that's not true at all. Their magic has limits. In most domains, there are wizards and muggles who can and can't search, and each group needs different support. Knowing the relative strengths and weaknesses of our target audiences is a key to good design.
Type of search is another key variable. There's a big difference between the simple lookup of known-item search and the dynamic learning of exploratory search. Google's got lookup down. Fast, simple, relevant. If you know what you want, Google will find it in less than a second. It's so fast, we use it for navigation, running queries even when we know the URL. But if you're unsure what you need, Amazon offers a better model. Faceted navigation plus tools for recommendation help us learn. Search becomes an iterative, interactive experience where what we find changes what we seek. While each type begins in a box, the types diverge by process and goal. Many systems must support both. To design well, we must think carefully about how and where to strike the right balance.
Of course, we must also consider platform, purpose, and context of use. Are we designing for desktops, laptops, televisions, kiosks, or mobiles in motion? The iPhone's small screen and soft keyboard place new limits on search, especially when it's jiggling in your palm in the back of a taxi cab in downtown Berlin at midnight. On the other hand, its multisensory I/O tears down old walls. When we integrate a microphone, speaker, GPS, accelerometer, magnetometer, and a multitouch interface, we redefine what's feasible in search. The whole may not recognize the sum of its parts. Design must respond to context. That's why it's good to ask where your users will be when they need you.
Finally, we should seek to balance the qualities of the user experience. In mobile, search must be useful and usable. Simple, fast, and relevant wins the day. But in music, search begets desire. Cover Flow makes it OK to look, while Pandora tempts us to buy. In government, accessible and credible are tops, but in business, search must be found, and real results add value to the bottom line. In each context, we must identify which qualities our users and organizations value, and then design with these priorities in mind.
