Improving search UX through investments in usability and contextual search results

Editor's Notes

• #2: Content Centered Design delivered to the Writers UA conference in 2007, where I discussed how Search could be improved by providing more context for search results. I recognized in Bing a lot of these elements and am excited by some of the other research being done related to search. The purpose of this talk is to explore how user experience design is improving search. Part of why I’m so excited about the investments in search UX that are being made is because so often we don’t know what it is we’re searching for, and when we’re asking the wrong questions, it doesn’t matter how relevant the results are. A good search UX enables serendipity—finding what we didn’t even know we needed. Like Mick Jagger says: You can’t always get what you want, but you just might find, you get what you need. Search is important. 50% of Web users use search every day I’m willing to bet that most of us in this room use search several times a day. And yet, search fails us more than ½ the time. There’s room for improvement.
• #3: There are over 1 trillion URLs, a fraction of which are useful to you or me, and in any event a trillion pages are meaningless unless we can exploit this information and derive insight. This is an exciting time to be a User Experience professional and search is opening new doors for our profession. Search engines were once the last place you would expect to find room for UX design—it was all command line. Now, people are recognizing the important of good UX design for search.
• #4: Here is Google in 2006. Nothing more than a command line. Now, people are recognizing the important of good UX design for search as an essential component. Relevancy is still most important component of a good search user experience, however. I will identify and explore some of the other ways to improve the search user experience. Developers of Bing, Google, and other search engines are recognizing that the Search Engine Results Page (SERP) can and must be much more than a list of results. Although I’ll be using Bing as my primary example, the point of this talk is not to praise Bing, but simply to use it as an example for how UX design can improve search. We will of course look at Google, but by now it is the gold standard by which search must be measured and is so familiar to most of us that I use Google primarily as a point of departure. While the focus will be on web search engines, as opposed to enterprise search or other specialized search applications, the principles of Search UX are universal. We’ll talk briefly about mobile. We can look forward to ubiquitous computing, and mobile devices are going to lead us there. What do I mean by search UX? Search UX is the combination of front and backend technologies that affect the user experience. So while I emphasize what the end user sees and experiences, I don’t mean to imply that search UX is just about moving pixels around on the user interface—sophisticated back-end technology (indexing, crawling, and link mapping) must be in place. I am saying that improving the back-end to achieve greater relevancy is only one part of the user experience equation.
• #5: Google is the State of the Art of Search Google Search Share is HUGE—it’s obviously meeting needs People like Google. According to the American Consumer Satisfaction Index this year Google (GOOG) received 86 points, out of a possible 100, besting Yahoo’s (YHOO) 77 points, Microsoft’s (MSFT) 75, Ask.com’s 74 and AOL’s 70. (Pre-Bing) Does this mean we should leave well-enough alone? No! There is vast room for improvement. Happiness with Google is relative to not finding anything at all, which as often as not is the case. I will add, that From September to October, Bing experienced the largest growth of query volume (of the top ten expanded search properties with an 8-percent increase in query volume to more than 1.2 billion searches.)
• #6: If you don’t know what you’re looking for, relevance is, irrelevant Search results fail to fully satisfy consumers on the first try 50% of the time Only one query in four is successful. Over 50% of the time searchers refine their query, go back to the SERP from a non-helpful result, or abandon search altogether. According to a MSFT study, 72% of people said current search results are too disorganized. Search is a dialogue – whether this dialogue is with an agent not only of superior knowledge and intelligence, but sensitive to our unique needs and habits, or with an idiot savant capable only of returning facts. Getting Search to emulate intelligent dialogue is part of the solution—but, we’re not there. Enabling serendipity (happy accident) means returning search results in such a way that you can find what you didn’t even know you were searching for. Getting what you needed if not what you originally wanted. I remember searching for books in the stacks at Suzzallo library—often the book I needed wasn’t the one I found in the card index, but the book next to it, or above it, or maybe even the next aisle over. This “happy accident” was made possible because related books were proximate to the book I thought I wanted.
• #7: Google.com from early 2006 and late 2009 iGoogle is also there, but it does little to improve the search experience. Google has been hard at work improving search, but the UI remains largely unchanged on both landing page and SERPs. This is intentional. Irene Au, the Director of User Experience at Google was quoted in Tech Radar as saying: “Google focuses on aggregating the data and presenting it to the users, so we don't want it to feel editorialized. We want it to feel machine driven, so people understand that it's not like we're presenting results with a certain point of view.” Right. If there’s anything I learned in my master’s program it is that all communication has an agenda, but that’s another story. While Bing’s most noticeable changes are user-interface related, Google’s new search architecture (code named ‘Caffeine’, and not yet released) improves the back-end: improved indexing speed, and crawling, and improved relevancy, but Google’s Caffeine is not expected to include any UI changes. The next version of Google will likely look much the same as it’s 2009 predecessor. One thing Caffeine will be offering is more real-time updates of search results as it will rapidly index new content.
• #8: Live.com from late 2006 Bing from late 2009 Microsoft has taken a different approach and has radically changed the UI of the search home page and the Search Engine Results Page. According to Stefan Weitz, director of Live Search, repeated searches is a key feature of Google: search, revise, repeat as necessary. Bing intends to help you find the answer the first time around.
• #9: Relevance remains most important—no matter how good the UI, if you can’t ultimately find something that meets your needs, your search has failed Disambiguation – do you mean apple the fruit or Apple the computer? From the minimal snippet to a full document or image preview Search results must be contextual—who you are, where you are, and what you’re doing should influence search results Instant results provide you with a one click answer such as an air fare or the current weather Results should be grouped appropriately Scanability enables users to quickly grok large quantities of information—scanability can be text based or visual Tasks and scenarios provide search results based on a best guess of your intended activity. If you are entering “Seattle to Las Vegas” chances are you’re looking to book travel. A picture is worth a thousand words
• #10: Search on “coffee” New UI components of Bing Bing provides explicit organization… Google’s organization is more subtle.
• #11: Google categorization on the page (searched for coffee). Google is doing many of the things that Bing is doing; it’s just not as obvious. This reflects Google’s desire to be streamlined.
• #12: Here’s an example of a task-based search. Note how relevant content is surface on the SERP—no additional clicking is required. Bing has focused its design efforts on improving searches for travel, shopping, and local, i.e. heavily monetized activities—that sweet spot between what people want and what advertisers will pay for.
• #13: Rich feature set after one click gets you instant results surfaced on the SERP. Here we see categorization and instant results. Note that the UI changes considerably from a standard SERP list of results.
• #14: Google also has instant results, but its UI retains the dominant style of a SERP. As mentioned earlier, Google doesn’t want it’s results to appear editorialized, but they are.
• #15: Bing visual search galleries Digital cameras. Bing is investing effort into enhancing the user experience for key categories of search results. Facilitating shopping is definitely a motivator here. A Microsoft study revealed that users could process results with images 20% faster than text-only results.
• #16: Bing video preview Previewing in context – fewer clicks, faster identification
• #17: Google image swirl Visualizing relationships and related content Strength of relationships Works just with images for now
• #18: Mobile search is driven more directly by mobile-specific scenarios. The Result set is paired down considerably to allow for the mobile phone form factor When searching for a movie on your phone, you’re probably not looking for inside gossip, box office rank, or production notes, you’re looking for movie times and locations. Note that your location is set explicitly.
• #20: Pivothttp://www.getpivot.com/ Brand new experimental technology (announced yesterday--November 18th--at the Microsoft Professional Developers Conference – PDC) the goal of the Pivot project is to enable users to view the Web as a “web” rather than a series of isolated pages. Pivot allows users to visualize hidden patterns so they can “discover new insights while interacting with thousands of things at once,” according to the Web site.
• #21: Pivot enhances many of the components of search user experience I’ve identified, but it doesn’t go far enough to address the need for context. In my spare time, I’ve been thinking about a new mobile app called “WTF.” This is a search app reduced to pure context. WTF has just one function, simply select WTF and you will get the most appropriate response based on who you are, where you are, what you're doing. Your phone already knows who you are, where you are, where your friends are, what you've been searching for, who you've been talking to, etc., so we ought to be able to leverage this contextual data to provide a rich search experience. Traffic at a dead standstill? Select WTF and you will be told "Blue Angels are in town." Waiting for a friend at the restaurant? Just press WTF to learn, "Steve's been in the bathroom on the 3rd floor for the past 20 minutes. Maybe you should ask if he's ok?" Girlfriend not texting you back? Press WTF to learn: "She's with Bill--maybe it's time to move on."Thing is with smart, GPS enabled phones, there's no reason an app couldn't infer all of this information today. But why stop with “WTF”? Computing power is reported to increase by as much as 100 times with multi-core computing. This means that in 3 years, the computer in my car is likely to be many times more powerful than my present day “top of the line” notebook. Not only will computing become increasingly ubiquitous, but we will begin to see anticipatory computing. Much like IBM’s Deep Blue chess computer which thinks not only of the next possible move, but all possible moves, our computers will soon have the spare cycles to anticipate our needs based on past behavior. This might mean loading Excel into memory seconds before I actually open the application, or it might mean coming home to find that a box of socks and underwear has been delivered because my computer knew I could use these things and so ordered them for me.
• #22: Categorization remains arbitrary, determined by taxonomy professionals in dark caves. Abstract concepts slot into multiple categories—biological binomial nomenclature it isn’t. We’re not yet realizing the promise of the semantic web. Context depends upon access to your personal information: identity, GPS coordinates, social network, and so on. Like it or not, much of this personal information is already “out-there,” but it remains to be seen how individuals will be able to manage how this information is accessed by various applications. Will the information search uses to determine context be transparent and available to the user? This is not only a privacy issue, but what if search is making inaccurate assumptions about the user? Getting context right is going to remain a challenge—a native of the UK is likely not to be happy when results for the National Football League are returned when they are searching for “football”, and this true regardless of what country they happen to be in when the initiate a search. Multi-modal interaction is full of potential for search. But, we’re a long ways from perfect speech recognition. We are already seeing apps which allow you to search based on a picture taken on your phone, but getting the app to recognize any object photographed, and then getting the intent of the search correct is hugely problematic. Microsoft is colossal, getting marketing, R&D, mobile computing, search, and other organizations to work well together to make search work well, and consistently, across platforms and applications is difficult. Google has an edge here as search is their bread and butter, but they too are a massive company. Relevance is still king—a pretty UI is no substitute for relevant results, and there’s still room for improvement