volume 05 - search 3rd edition

Upload: buddha-blessed

Post on 01-Jun-2018

467 views

Category:

Documents


21 download

TRANSCRIPT

  • 8/9/2019 Volume 05 - Search 3rd Edition

    1/153

      WWW.NNGROUP.COM 48105 WARM SPRINGS BLVD., FREMONT CA 94539–7498 USA

    Copyright © Nielsen Norman Group, All Rights Reserved.To buy a copy, download from:http://www.nngroup.com/reports/ecommerce-user-experience/ 

    E-Commerce User Experience

    Vol. 5: Search (Including Faceted Search) 

    Based on eyetracking, user testing, and diary studies

    3rd Edition

    By Amy Schade and Jakob Nielsen

    http://www.nngroup.com/reports/ecommerce-user-experience/http://www.nngroup.com/reports/ecommerce-user-experience/http://www.nngroup.com/reports/ecommerce-user-experience/

  • 8/9/2019 Volume 05 - Search 3rd Edition

    2/153

     

    2 [email protected] Copyright Notice 

    Copyright Notice

    Please do not post this document to the internet or to publiclyavailable file-sharing services.

    This report required hundreds of hours of planning, recruiting, testing, analyzing,

    writing and production. We sell these reports to fund independent, unbiased usability

    research; we do not have investors, government funding or research grants that pay

    for this work.

    We kindly request that you not post this document to the Internet or to publicly

    available file-sharing services. Even when people post documents with a private URL

    to share only with a few colleagues or clients, search engines often index the copy

    anyway. Indexing means that thousands of people will find the secret copy through

    searches.

    If someone gave you a copy of this report, you can easily remedy the situation by

    going to http://www.nngroup.com/reports  and paying for a license.We charge lower fees for our research reports than most other analyst firms do,

    because we want usability research to be widely available and used.

    Thank you!

    Report Authors: Amy Schade, Jakob Nielsen

  • 8/9/2019 Volume 05 - Search 3rd Edition

    3/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 3

    Table of Contents

    Summary of Research Studies and the E-Commerce Report Series ............... 4 

    E-Commerce Report Series .......................................................................................................... 4 

    Research Studies ........................................................................................................................ 4 

    Study One .............................................................................................................................. 5 

    Study Two .............................................................................................................................. 5 

    About the Third Edition ............................................................................................................... 6 

    Search Behavior ............................................................................................ 7 

    Why Users Search ...................................................................................................................... 7 

    What Users Search For .............................................................................................................. 10 

    Query Length ........................................................................................................................... 13 

    Search Success ........................................................................................................................ 18 

    Elements of Successful Search .................................................................... 20 

    Supporting Search .................................................................................................................... 20 

    Knowing How to Search............................................................................................................. 20 

    Presenting Search Results ......................................................................................................... 21 

    Narrowing the Choices .............................................................................................................. 22 

    Guidelines List ............................................................................................. 24 

    Search Visibility .......................................................................................... 26 

    Supporting Users’ Searches ........................................................................ 38 

    Advanced and Scoped Searches .................................................................. 60 

    Presentation of Search Results ................................................................... 65 

    Sorting Results .......................................................................................... 102 

    Narrowing Product Choices: Guided Navigation or Faceted Search ........... 105 

    About the Authors ..................................................................................... 147 

    Acknowledgments ..................................................................................... 148 

  • 8/9/2019 Volume 05 - Search 3rd Edition

    4/153

     

    4 [email protected] Summary of Research Studies and the E-Commerce ReportSeries

    Summary of Research Studies and the E-CommerceReport Series

    E-COMMERCE REPORT SERIES

    This report is one of 13 reports about the E-Commerce user experience. Ten of thereports in the series were generated from the findings of two rounds of e-commerceresearch studies. The first editions of these reports were published as a book, witheach chapter also available as a downloadable report. The second and third editionsinclude an additional report, based on the same series of studies, about customerservice.

    This series also includes three volumes which are a result of additional researchstudies, separate from the main e-commerce research. These reports are included inthe series due to their direct relationship to the e-commerce user experience andcover the topics of wishlists and gift certificates, store locators, and confirmation andtransactional email messages. Each of these reports includes a section aboutmethodology, covering the details of each research project.

    The entire E-Commerce User Experience series is available for download athttp://www.nngroup.com/reports/ecommerce-user-experience/ and includes thefollowing titles:

    1.  General User Behavior & Executive Summary2.  Homepages and Category Pages (including Product Listing Pages and

    Product Comparisons)3.  Product Pages (including Reviews)4.  Shopping Cart, Checkout & Registration5.  Search (including Faceted Search) [this report]

    6.  Customer Service7.  Selling Strategies8.  Wishlists, Gift Certificates and Gift Giving9.

     

    Trust and Credibility10. International Users11. Store Finders and Locators12. Transactional Email and Confirmation Messages13. Methodology

    RESEARCH STUDIES

    The information in these reports is a result of two separate rounds of e-commercestudies conducted by Nielsen Norman Group as well as a round of design reviews ofa set of e-commerce websites. The studies took place in the United States, UnitedKingdom, Denmark and China (Hong Kong), and involved user testing, a diary-basedlongitudinal study and an eye tracking component.

    The Methodology report in the E-Commerce Report Series includes the full details ofeach study, the list of sites tested, and information about participants.

    The Wishlists and Gift Certificates, Transactional Email and Confirmation Messagesand Locator Usability reports are based on additional research studies. Each of thesethree reports includes its own methodology section.

    http://www.nngroup.com/reports/ecommerce-user-experience/http://www.nngroup.com/reports/ecommerce-user-experience/http://www.nngroup.com/reports/ecommerce-user-experience/

  • 8/9/2019 Volume 05 - Search 3rd Edition

    5/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 5

    Study One

    The first research study was conducted by a team of five usability experts. Theyconducted usability tests of 20 business-to-consumer e-commerce websites. A totalof 64 people participated: 39 from the United States and 25 from Denmark. Nineteenof the twenty sites tested were American websites, which were tested by users in

    both countries.Users ranged in age from twenty to sixty. All users had previously shopped onlineand most had made purchases; however, we screened out people who had extensivetechnical knowledge of the Web.

    Usability testing sessions lasted two hours, and users typically tested three of the 20selected sites in that time. Each site was tested by a minimum of nine users: sixfrom the U.S. and three from Denmark. Sites were selected in seven differentindustries, such as clothing and toys, so that within each industry we had two orthree sites for comparison.

    Tasks were modeled on common goals of online shoppers. Most tasks asked users tofind a specific item or were open-ended, allowing freedom to shop according to theirown preferences. In most cases, we stopped users before they entered a credit cardnumber, so they did not complete the purchase. We also included tasks involvingcustomer service information.

    For each test session, a facilitator sat next to the user, providing instructions to theuser and taking notes. Users were asked to think aloud as they worked, describingtheir decision processes and any positive or negative reactions to the sites.

    Study Two

    The second study included a diary-based longitudinal study and user testing,including an eyetracking component.

    Research began with the diary-based study. Ten participants from around the UnitedStates were asked to record information in a notebook about their online shoppingexperiences for a period of six weeks during the winter holiday shopping season. Thegoal was to understand how users shopped online.

    Users answered questions including the goal of visiting the site, why they visited thatparticular site, and if they achieved their goal. Users were also asked about whatthey liked and disliked about the site. Information from this round of research wasused in part to develop tasks for the user testing portions of the study.

    The study also included user testing with participants in London, United Kingdom;Hong Kong, China; Munster, Indiana; Kennesaw, Georgia; and New York, New York.The New York City component, which was the largest, included eyetracking.Eyetracking allowed the facilitator to observe and record where the user was lookingon the screen.

    Ninety-eight users participated in user testing. Participants included an almost evensplit between men and women who ranged in age from 18 to 64. All participants hadpurchased online previously, with varied amounts of online shopping experience. Theleast experienced user had purchased online once in the past year and 10participants had made more than 30 online purchases in the past year. Users wererecruited across a range of household income levels and general online experience.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    6/153

     

    6 [email protected] Summary of Research Studies and the E-Commerce ReportSeries

    More than 100 sites were included in the user testing component of the study. Sitesselected for testing included sites big and small, from various industries with variedproduct offerings and different design approaches. In addition, users completed taskson sites they had previously visited. Participants provided a list of sites during therecruiting process and were asked to visit one of them during the study. Thisexpanded the number as well as the types of sites tested.

    Tasks were modeled after those in the first study, including directed tasks askingusers to find specific items, open-ended tasks allowing for site exploration, andcustomer service related tasks. Users proceeded as far as possible through thepurchase process with fake user information.

    Users completed three additional types of tasks in the second study. Users visitedsites they had visited before, which allowed us to observe users returning to a site asa repeat visitor. Users also completed open-ended tasks where they were given agoal of something to purchase, but were not directed to any particular website tomake the purchase.

    The New York component of the study also included a task where users completed apurchase. Users selected one of five sites on which to shop and were given a budget.

    They could purchase any item or items they wanted from the site within theirbudget, send the purchase to themselves, and be reimbursed for the purchase price.

    The same facilitator ran all sessions in the second study, except for the Georgiatests. In all sessions, the facilitator sat next to the user, providing instructions,observing and taking notes. Users thought aloud as they worked.

    ABOUT THE THIRD EDITION

    The third edition includes guidelines derived from both research studies as well asrevisions, clarifications and further examples of guidelines from the previous editionsof this report. For the third edition of this report, we used the existing guidelinesfrom our research studies to complete heuristic reviews on additional e-commercesites and updated examples as appropriate.

    Some screenshots remain from earlier rounds of research. Older screenshots areretained when necessary to reflect the appearance of the site at the time it wastested. Some sites’ designs may have changed since the sites were tested.

    For instructional purposes, all examples are valuable. They reflect actual designs andreal user behaviors, which in turn create best practices that stand the test of time.Lessons learned from these designs are valid, even when designs have changed.Including examples helps illustrate good and bad usability examples, which can helpdesigners learn from previous mistakes and successes.

    Participants' personal information has been blurred on screenshots.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    7/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 7

    Search Behavior

    WHY USERS SEARCH

    In our first study, 81% (52 people) of users searched on one or more sites. In our

    second study, we tracked search behavior in the eyetracking portion of the study,and 83% of users (54 of 67 people) searched on one or more site.

    Users had many different reasons to search on sites. The ten main reasons forsearching were:

    •  To get a feeling for the breadth or depth of the site’s product

    selection

    Some users searched to get a feel for the product selection on a site –typically on a site they had not visited before. For instance, a user searchedfor “shoes” on Boden’s site to see if the site offered any. Another searched onCDBaby for a particular artist to try to determine what type of music the siteoffered.

    •  To quickly locate a product category, instead of using site navigation

    Search was sometimes an alternative for navigation, with users immediately

    searching for a product category rather than using site navigation. A user on

    CompUSA searched for “printers,” one on Links of London searched for

     “cufflinks,” and another on Park N Shop searched for “detergent.” The Park N

    Shop user said, “The word ‘detergent’ just gave me the choices I needed. No

    other useless information.”

    Search can be particularly helpful on sites with large or diverse inventory. A

    user on Walmart.com said, “With Walmart, you know there’s so much of

    everything that I can’t just look at the homepage. I just automatically type

    something in and be as specific as possible with what I want to purchase.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    8/153

     

    8 [email protected] Search Behavior

    Search helped a user navigate the large product selection on Walmart.com.

    •  After unsuccessfully using the site navigation

    While sometimes searching for a product category was simply a shortcut toavoid using navigation, in other cases, product category searches stemmedfrom users being unable to find products via the site navigation. Searchbecame a replacement for failed navigation. For instance, a user on Sasa.comcould not determine which category would include suncare products, sosearched for “sunbathing” after pursuing several different navigational paths

    to such products with no luck.When searches resulted from failed attempts at navigation, users oftensearched for specific phrases, including products and criteria, rather thanbroad product categories. A user who could not navigate to frozen spinachunder the Sainsbury’s category of Green Vegetables turned to search to lookfor “leaf spinach, frozen.” A user searched for “cowboy pyjamas” when shecouldn’t determine how to find sleepwear on Boden’s site.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    9/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 9

    A user didn’t know where to look for shortbread on Harrods’ site, so turned tosearch.

    •  To locate a specific product

    If users knew the name of the product they were looking for, search was(hopefully) an easy solution. Users armed with the title of a book or name ofa movie could easily enter the information directly in search. A user onAmazon searched for “Harry Potter” to find the latest book in the series. Auser on FYE.com searched for the movie Heartburn, which she had previouslyhad trouble locating elsewhere.

    •  To locate products with a specific attribute

    When users had particular needs, searches became more detailed. Userssearched on product criteria rather than categories. On NFL Shop, a usersearched for “Peyton Manning” to locate a football jersey for that player. Auser on Links of London searched for “onyx” when looking for cufflinks madeof that material.

    •  When product descriptions did not answer shoppers’ questions

    Users also turned to search when product descriptions did not answer theirquestions. On Comet’s site, a user could not determine if the products he wasviewing met his criteria, so tried to narrow down his choices by searching for

     “combination microwave.”•  When faceted search or guided navigation did not return the desired

    product results

    Users attempted search when faceted search did not return an expected

    products. On Nike.com, a user couldn’t find her desired hat using the site’s

    offered facets, so instead searched for “grey hat.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    10/153

     

    10 [email protected] Search Behavior

    •  To narrow product options

    Some users navigated to categories and then entered product characteristics,

    trying to narrow results. A user looking for specific earrings navigated to

    earrings, then entered "pink gold" when faced with too many options.

    •  To return to an item previously located

    If a user wanted to view an item they'd previously located, he sometimes

    searched for the name or brand of the item to relocate it quickly. For

    instance, a user searched for "carmelo anthony" when looking at athletic

     jerseys, then switched to "michael jordan" and then searched again for

    "carmelo anthony" when he decided that's the item he wanted to buy.

    •  When looking for a piece of information, rather than a product, on the

    site

    The final reason users searched on sites was to look for non-product

    information. Users tried the search box to find information about return

    policies, shipping timeframes, assembly instructions and price matching. A

    user on the Metropolitan Opera’s site searched for “subtitles” to see if

    translations were offered at performances, and a user on Cinema.co.hksearched for “ticket” to find out if refunds were offered.

    WHAT USERS SEARCH FOR

    Our users searched for:

    •  Product categories (shredders, earrings, digital cameras)

    •  Product numbers (Sharp LC-32D40U, TC320X60, Canon BCI-6M)

    •  Product names (Richest Man in Babylon, The Family Stone, A Heartbreaking

    Work of Staggering Genius, Gio Café Table)

    •  Product characteristics (pink gold, peanut, smartwool)

    •  Product categories and criteria (espresso dresser drawer, air conditioner

    energy efficient, women ski pant, least expensive mouse)

    •  Brands (Yellowtail, Steve Madden, Eukanuba)

    •  Customer service terms (price guarantee, return policy, delivery)

    To serve its users, a site must accommodate all types of searches. Ensure that the

    site search adequately returns results for categories, product names, and product

    criteria. Regularly check search logs to learn about the language shoppers’ use when

    searching for products or content on the site. Review successful and failed searches

    to determine how to tweak the search engine to return better results.

    We looked at web queries based on if users were searching within a particular site or

    if they were conducting a web-wide search to locate a site offering a product.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    11/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 11

    When users searched within a site, the most common type of query at 34% was a

    combination of category and at least one product criterion. The second most popular

    search type, accounting for 22% of the queries, was a criterion alone, with 20% of

    queries for a category alone.

    Of 332 queries conducted within a site, 4 submitted the search box filler text, 10included specific product numbers, 23 were related to customer service requests,43 included a specific product name, 66 were for a product category with noparticular criteria, and 116 included both a category and at least one qualifyingcriterion.

    In web-wide searches, more than half of the queries (57%) combined a product

    category as well as a criterion. This makes sense, as users searching the entirety of

    the web want to be more specific in their search criteria. 14% of queries were

    intended to find a specific website or the site for a specific company, and another

    14% of queries were for product criteria alone.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    12/153

     

    12 [email protected] Search Behavior

    Of the 79 web-wide queries, 1 was for a product name, 1 was for customer serviceinformation, 4 were categories, 6 were product names, 11 were for product criteria,11 were to find a company’s website, and 45 included both a product category andat least one qualifying criterion.

    The phrasing of our tasks influenced users’ exact search queries, as users oftenpicked words directly from the task to use in the search box. However, users didn’tconsistently pick out the same combinations of words for their searches. Though allwere working from the same written task, some users naturally searched for aproduct name, others for a criterion or characteristic, and others for a combination ofcriteria and category. For example, in a task asking users to find a box that canorganize medication, users searched for “medication organizer,” “medicine,” and “medication box.”

    Recognizing the type of search a user is conducting can help direct him to the rightarea of the site. For instance, on CompUSA.com, a search for “printers” nicely tookusers to a page specifically designed for printer searches, allowing users to select acategory of printer or brand.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    13/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 13

    A search for “printers” on CompUSA brought users to this page. The site explains, “We’ve brought you to our ‘Printers’ page based on your search.”

    QUERY LENGTH

    Users conducted 411 searches in the eyetracking portion of the second study. Ofthose, 332 searches were conducted on e-commerce sites, and 79 searches were

    web-wide searches on search engines such as Google or Yahoo.In four instances, users ran searches on the filler text that appeared in the searchbox on The Container Store’s site. We removed these search queries from theanalysis of query length, leaving us with 407 overall searches.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    14/153

     

    14 [email protected] Search Behavior

    The average number of characters per query across all searches was 15.5. Withinsite searches were almost the same at 14.5 characters. Web-wide searches werelonger, at an average of 20.5 characters.

    The average number of words per query was 2.5 for all searches, 2.3 for within sitesearches, and 3.4 for web-wide searches.

    Since 57% of web-wide searches included a combination of a product category andat least one product criterion, the longer length of users' queries and number ofwords in those queries makes sense.

    Search Type: Study 2 Average number of

    characters

    Average number of

    words

    All searches (407) 15.5 2.5

    Within site searches (328) 14.5 2.3

    Web-wide searches (79) 20.5 3.4

    Across all searches, the average query length was 15.5 characters. Within a site,the average length was 14.5 characters. Web-wide queries averaged 20.5characters in length.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    15/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 15

    Across all searches, the average number of words per query was 2.5. Within a site,the average number was 2.3. Web-wide queries averaged 3.4 words.

    Two word queries were most common for within site searches, with single wordqueries the second most popular. In Web-wide searches, however, two- and three-word queries were most common. Again, this makes sense considering many web-wide queries were more specific and often included a combination of productcategory and product criteria.

    Within a website, two word queries were most common, accounting for 135searches (41%). One word queries accounted for 95 searches (29%), three wordsfor 51 (16%), four for 17 (5%) and 30 searches used five or more words (9%).

  • 8/9/2019 Volume 05 - Search 3rd Edition

    16/153

     

    16 [email protected] Search Behavior

    In Web-wide searches, three word queries were most popular, accounting for 32%(25) of searches. Two-word queries followed closely at 28% (22 searches). Single-word queries accounted for 6% (5 searches), four-word for 15% (12 searches) andfive or more words for 19% (15 searches).

    In our first study, 54% of search queries consisted of more than one word. In oursecond study, 75% of all searches included more than one word. Search querieshave gotten longer: users are more specific in their searches.

    In our first study, only 54% of queries included more than one word. In our secondstudy, 75% of all queries included more than one word.

    The biggest difference between within site searches and Web-wide searches in oursecond study was the number with more than one word. Ninety-four percent of Web-wide searches included more than one word, versus 71% of within site searches.Again, this indicates that users are more specific on the Web as a whole than theyare on a specific website.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    17/153

  • 8/9/2019 Volume 05 - Search 3rd Edition

    18/153

     

    18 [email protected] Search Behavior

    SEARCH SUCCESS

    In our first study, users' searches were successful 64% of the time. In our secondstudy, within site searches were successful 74% of the time.

    Search success did not necessarily dictate success or failure in completing a task.Some successful searches led to failed tasks, and for some failed searches, users

    found other ways to successfully complete tasks.

    The bar chart shows, by percentage, how many queries our users had to enterbefore they got meaningful search results. It also shows, by percentage, how manyqueries users entered before they gave up.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    19/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 19

    The probability for success declined as one needed more queries in the second study.

    •  First query: 64% success rate

    •  Second query: 38% success rate

    The 38% success rate is the success rate for users who failed in the first query, butcontinued to do a second (and perhaps third or fourth) query. It is a result of 8% outof the 21% who continued after the first query (8%/21%). The 21% is the result ofthe 10% who had success in the 2nd, 3rd or 4th query or later plus 11% who gaveup after 2, 3 or 4 queries.

    The following list shows which query led to success when a user was successful.

    •  First query 86%

    •  Second query 11%

    •  Third query 0%

    •  Fourth or following query 3%

    When users entered multiple queries, they were usually a variation of the same

    word(s). For example, users searched for “coffee pot” and “coffee maker” or “returns” and “exchanges.” We think this lack of variation partly explains the drop insuccess rates with subsequent queries.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    20/153

     

    20 [email protected] Elements of Successful Search

    Elements of Successful Search

    SUPPORTING SEARCH

    There are many elements that need to work well to provide a successful search

    experience for the user.•  Locating the search: A clearly visible search box appears on each page

    •  Entering the search: The search box must accommodate enough characters tosupport users’ searches and should default to the broadest search scope

    •  Interpreting the query: Searches need to support queries for products,information, and product characteristics and need to support the languageand terminology shoppers use to describe products or services

    •  Returning results: Users need to know if the search was successful or if noresults were found. They must understand the results, know why they werereturned, and be able to quickly assess if they need to refine the search or ifthey’ve found the right items.

    • 

    Narrowing options: More and more sites offer users options to narrow theirsearch results to meet their specific needs. Such options need to workperfectly to gain the user's trust and support the sale.

    Users were frustrated when they searched for items they knew the site carried, butthe search results did not include the item. This made users doubt the site wasworking properly.

    KNOWING HOW TO SEARCH

    Search should be simple. Users shouldn't have to figure out where to locate yoursearch box, or how to submit a query. Present an open text box at the top of thepage, followed by a clear Search button. Default to the most lenient search scope.

    Users have an expectation that they can simply enter a term, any term, and getrelevant and meaningful results from your site. Don't overcomplicate the search boxwith filler text, extensive options, instructions, or confusing buttons. Keep it simple.

    Keeping search simple also means making the scope of the search apparent. Usersare most successful when search is automatically scoped to the broadest category,such as All Products. When sites scope searches to the currently selected section,users often miss that information. If search boxes and search results don't make thescope immediately clear, users can be left thinking a site does not carry the productsthey want.

    A user on Circuit City searched for a movie within the TV category. The site returnedno results, but worse, there was minimal indication on the page that she had run ascoped search. She assumed the site did not carry the product and moved on to acompeting site.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    21/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 21

    A search scoped to the wrong category led a shopper to believe Circuit City did notcarry the item she was looking for, when it did.

    PRESENTING SEARCH RESULTS

    Search results need to clearly display products (or site content) that relates to theuser’s query. Some searches in our studies failed to deliver successful results for thefollowing reasons:

    • 

    Contained insufficient information for the user to select the right product•  Presented a list that appeared to be in random order to the user

    •  Contained too many items or had too many results on the first page

    •  Contained only one item

    •  Misled users into thinking fewer items were available by poor presentationof “featured” results

    •  Had incomprehensible or inadequate page-to-page navigation within theresults

    •  Presented irrelevant, poorly placed, or hard to use faceted search options

    •  Included items that appeared to be — or actually were — unrelated to theuser’s search criteria

    Successful search results are a matter of the quality of the results returned as wellas how those results are presented.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    22/153

     

    22 [email protected] Elements of Successful Search

    NARROWING THE CHOICES

    More and more sites offer both sorting and filtering options to users to narrow theirselection based on their own shopping criteria. In fact, these tools are so commonacross e-commerce sites that many users expect such options to be available.

    Price is the sorting option most commonly offered by sites, with some also offering

    brand, popularity, customer ratings, or time on the site (newest first). Such optionsgive users a quick and easy way to see all offerings, sorted by whatever is mostimportant to them.

    Users found sorting by price particularly helpful when trying to qualify for offerswhich required minimum purchase totals. Several sites offered free shipping with aspecified minimum purchase. Users found sorting by price helpful when trying to adda low priced item in order to qualify them for the offer.

    Faceted search, a type of filtering also known as guided navigation, helps users onlyview those products that meet their needs or specifications. Users can easily find asmall black cardigan sweater on a site with hundreds of sweaters, for instance, byspecifying size, color and style. Rather than navigating through page after page ofoptions, hoping to find a product in the right style, size, or color, users can movedirectly to what they need and see those items listed side by side. Faceted search,when done well, allows users to see and focus on only those items that meet theirspecific needs.

    Such tools are a huge help for shoppers who come to a site with a specific goal inmind. They also help users who don’t yet know what they want by identifyingimportant characteristics to consider about the product.

    A user on Yoox.com looked at sneakers and was put off by the more than 20 pagesof results. He then saw he could select his size, and ended up with 6 pages ofresults. He said, “And this is good because I won’t be disappointed later if they didn’thave my size. I know they have it in my size if it’s here.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    23/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 23

    Faceted search allowed users to see only those items that met their needs.

    However, such tools are often not well-executed. Some sites try to use the sameproduct categories across different product types, resulting in poor choices for allproducts. Others don’t categorize products appropriately, showing users incorrectinventory. Others make it difficult to see what options have already been selected, orto remove those options. Such mistakes can make a user lose trust in a site and goelsewhere.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    24/153

     

    24 [email protected] Guidelines List

    Guidelines List

    Search Visibility .......................................................................................... 26 

    1.  Use an open text field instead of a link to a search page..................................................... 26 

    2. 

    Make the search box easily identifiable, with an open text field followed by a Search button. .. 27 

    3.  Do not put filler text in the search field. ........................................................................... 28  

    4.  Put the search box on every page. ................................................................................... 33 

    5.  Ensure search fields are at least 30 characters long, allowing enough space in search boxes

    for users to enter and see their queries. ........................................................................... 34 

    6.  Let users submit a search using the Enter  key. .................................................................. 35 

    7.  Present only one search box for users. ............................................................................. 35 

    8.  Make sure navigational menus do not cover the search box. ............................................... 36  

    Supporting Users’ Searches ........................................................................ 38 

    9.  Regularly review search logs for search trends and users’ terminology. ................................ 38 

    10. 

    Adjust the search engine based on search logs. ................................................................. 38 

    11.  Adjust content or offerings based on search logs. .............................................................. 38 

    12.  Consider special treatment for frequent queries, such as categories or brands. ..................... 39 

    13.  Support many types of searches, including non-product searches. ....................................... 41 

    14.  Indicate what users can search for, when appropriate. ....................................................... 43 

    15.  Consider displaying suggested search terms as the user types his query. (Ensure suggested

    terms return relevant results.) ........................................................................................ 47 

    16.  Carefully categorize products to ensure solid search results. ............................................... 49 

    17.  Accept synonyms commonly used by customers. ............................................................... 52 

    18.  Accommodate misspellings and variant forms (spaces, plurals, etc). .................................... 53 

    19.  Accommodate searches with multiple words, including characteristics such as color, size and

    brand. .......................................................................................................................... 55 

    20.  Recognize search operators. ........................................................................................... 59 

    Advanced and Scoped Searches .................................................................. 60 

    21.  Consider the need for advanced search. ........................................................................... 60 

    22.  Provide a clear link to advanced search – and back. ........................................................... 60 

    23.  Explain the scope of the search. ...................................................................................... 60 

    24.  Default searches to the most lenient category. .................................................................. 61 

    25.  Simplify options in scoped search. ................................................................................... 62 

    26.  Consider the use of facets rather than scoped search. ........................................................ 63 

    Presentation of Search Results ................................................................... 65 

    27.  Don’t put any steps between the user’s search and the search results. ................................. 65 

    28.  List the number of results found. ..................................................................................... 66 

    29.  Be cautious if promoting specific products in the search results. .......................................... 66 

    30.  Repeat the user’s search query at the top of the page. ....................................................... 67 

    31.  Allow users to View All  or view more results per page. ....................................................... 69 

  • 8/9/2019 Volume 05 - Search 3rd Edition

    25/153

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 25

    32. Provide a descriptive product name, small product image, and price in search results. ........... 70

    33. Include relevant product details or a brief product description in search results. .................... 73

    34. Include inventory information on search results pages. ...................................................... 77

    35. List all results and let users filter them as needed, rather than listing results separately by

    type. ............................................................................................................................ 79

    36. Design search results pages to maximize the number of results a user can see andcompare, without cluttering the screen. ........................................................................... 83

    37. Clearly list content matches as well as product matches. .................................................... 86

    38. Indicate if only one result is found. .................................................................................. 89

    39. State when no results are found. ..................................................................................... 90

    40. Write constructive and comprehensible messages when no results are found. ....................... 94

    41. Allow users to search again from the No Results page. ....................................................... 95

    42. Clarify if there are no exact matches but the site is making a suggestion or best guess. Be

    cautious in doing so. ...................................................................................................... 96

    43. Indicate in search results if additional products may be available in stores. ........................... 99

    Sorting Results .......................................................................................... 102 

    44. List results in by relevance. .......................................................................................... 102

    45. Offer appropriate sorting options, such as by price, customer rating, or popularity. .............. 103

    46. For sorting by price, present an option both to sort from high to low prices and from low to

    high. .......................................................................................................................... 103

    Narrowing Product Choices: Guided Navigation or Faceted Search ........... 105 

    47. Use facets to help users narrow product choices if search results page or category pages

    frequently offer more than 20 options. ........................................................................... 108

    48. Offer both sorting and filtering options, as appropriate. .................................................... 114

    49. Pick attributes your customers understand. .................................................................... 115

    50. Provide the most useful differentiating attributes specific to product type. .......................... 118

    51. Ensure products are listed in all appropriate categories. (Filtering must work perfectly.) ....... 123

    52. Clearly present filtering options. Do not scatter them around the page. .............................. 125

    53. Display the most commonly used product attributes first. ................................................. 128

    54. If there are a large number of options in a category, such as brand, consider making only

    the most common visible by default and hiding the rest behind a link. ............................... 131

    55. Offer further refinement of categories as needed. ............................................................ 132

    56. List the number of products available with each attribute. ................................................ 132

    57. Let users select more than one attribute in a category as appropriate. ............................... 134

    58. Allow users to select criteria in more than one category.................................................... 135

    59. Show users what attributes they’ve already selected. ....................................................... 137

    60. Let users clear previous selections. ................................................................................ 141

    61. Don’t show filtering options when only one item is listed. ................................................. 144

    62. When displaying products that meet users’ criteria, use images that reflect the users’ 

    selections. .................................................................................................................. 146

  • 8/9/2019 Volume 05 - Search 3rd Edition

    26/153

     

    26 [email protected] Search Visibility

    Search Visibility

    1. 

    Use an open text field instead of a link to a search page.

    Users look for open text fields when they want to search. They do not look for a link.

    If search is hidden behind a link, some users assume the search is bad and thecompany is trying to hide it.

    In the first study, Nordstrom’s site did not have an open text field for search, butinstead had a link at the top of the page. On Nordstrom, several users looked in vainfor the search function. Expecting a text box, they did not notice the link at the topof the page.

    One user, who eventually did find the search link on Nordstrom, complained, “Whywould you have to work so hard to get to where you’d want to buy something? I’mlooking for a watch, and it took me 10 minutes to find the search box. And it was inthe smallest of print.”

    Search buttons were harder for users to find and understand than search boxes.

    NASCAR's shopping site had an easy-to-locate search box in the upper right-handcorner of the page, accompanied by a clear Search button. This was well done.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    27/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 27

    NASCAR's shopping site offered a clear open text field followed by a nicely labeledSearch button.

    2.  Make the search box easily identifiable, with an open text

    field followed by a Sea r ch  button.

    The search box needs no label. A clear Search button next to the field will identifythe search for the user. The search button both identifies the search as well as tellingusers how to execute the search.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    28/153

     

    28 [email protected] Search Visibility

    The search box was easy to locate on Greenhome.com.

    3.  Do not put filler text in the search field.

    Some sites place filler text in the search field. This can be problematic for a fewreasons. First, it makes it more difficult to locate the search. Users look for an opentext field, so filler text can make the box less noticeable.

    Second, it can cause problems when users enter a search term. Any filler text needsto immediately disappear when the user clicks in the box. Otherwise, users may

    accidentally type the search query next to, rather than in place of, the filler text.Resulting search queries may have “enter product name” or other filler text in themiddle of them.

    Third, if text is there to help users know what they can enter in the search box, it isa bad idea to place that text in the search box itself. Once users click, that text willbe gone.

    Users had a difficult time understanding the search box on The Container Store’s sitebecause of its filler text. The search box had the filler text, “enter product or item#.” The text itself confused one user, because it made her think she needed to knowa product number: "But I don't have a product number."

    However, the more significant problem with the filler text was that it made two users

    think the Search button was a link to get to the search. They did not recognize thesearch field as a place to enter a query. The filler text completely filled the box andwas the same color as the box outline. These users both clicked Search withoutentering a search term.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    29/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 29

    To make matters worse, the site ran a search on the filler text and returned 100results. The users who were already confused because search wasn't acting in theway they expected, now received results even when they did not truly run a search.

    The filler text made the search field on The Container Store’s site unrecognizable totwo users. They clicked the Search button without entering a search term, thinkingthey would see a Search page.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    30/153

     

    30 [email protected] Search Visibility

    The Container Store’s search engine returned 100 results for the filler text in thesearch box, “enter product or item #.”

    The users who clicked Search without entering a query had no idea what happened.One returned to navigation because she didn’t understand the results.

    The other user tried to use the site search again, but again did not enter a searchterm. When she saw the results, she scrolled down and saw a New  Search button atthe bottom of the page. However, this search box had filler text as well. She clicked

    the button, thinking this would start a new search, but again ran the same search forthe same filler text, bringing back the same results.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    31/153

  • 8/9/2019 Volume 05 - Search 3rd Edition

    32/153

     

    32 [email protected] Search Visibility

    A user clicked Go on Staples.com without entering a search query and received apage stating there were no results found for “type search here.”

    Oriental Trading Company used the filler text “search item# or keyword.” If a userhit the Go button without changing the text, the site returned an error which said, “Sorry, you must enter an item number or keyword to search.” The page providedanother search box and some search tips, including information about how to searchfor an item number from a catalog. This was helpful.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    33/153

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 33

    If a user tried to run a search on an empty search field, Oriental Trading’s errormessage told the user to enter an item number or keyword and provided somesearch tips.

    Some sites simply did not submit the search if the search box was empty, leavingthe user on the same page. This feedback is inadequate, as users can think thesite is broken or not functioning correctly. Costco presented an error message onthe page, below the search box. However, this message only appeared for a fewseconds, making it easy for users to miss. It would have been better to keep themessage on the page.

    Costco's error message appeared too briefly to be helpful.

    4. Put the search box on every page.

    People want easy access to search. Placing the search box in a consistent place oneach page allows users to start a search regardless of their location on the site.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    34/153

     

    34 [email protected] Search Visibility

    On The Scholastic Store’s site, each page offered a consistently placed search box.

    5.  Ensure search fields are at least 30 characters long,

    allowing enough space in search boxes for users to enter

    and see their queries.

    We recommend search boxes 30 characters in length. This will accommodate mostusers' queries. In our second study, the average search query on an e-commercesite was 14.5 characters, with a minimum of 3 and a maximum of 50 characters. Themedian query length was 13 characters.

    Look at your search logs to see how long users’ queries are. It is easier for users to

    enter a query when the box accommodates their full entry.

    The ideal length may vary depending on the site and its inventory. For instance, asite selling books or movies may need to accommodate longer titles. A site withspecialized equipment may need to accommodate searches for very specific orlengthy sets of product criteria.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    35/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 35

    The search field on Urban Outfitter’s site only accommodated 14 characters.

    6.  Let users submit a search using the En t e r  key.

    When users enter a search term in the search box, they expect to be able to use thekeyboard to submit the query. On Lake Champlain Chocolate’s site, users had to click

    on Go to invoke the search. Using the Enter  key on the keyboard did not work. Thisrepeatedly frustrated one user and slowed several others down.

    In our first study, it was common that hitting the Enter  key simply reloaded thepage, rather than submitting the search. This gave users the impression that thesearch had returned no results.

    7. 

    Present only one search box for users.

    Some sites offered multiple search boxes for different types of searches. Users don’twant to have to decide between search boxes, they simply want their searches towork. Don’t make users do extra work. Do the work on the backend so the singleopen search field can accommodate any type of search the site supports.

    Walmart’s site had boxes such as Search our Site and Find…in Baby Shop on somepages. Several users did not understand the difference between the boxes. (Formore on scoped searches, see page 60.)

  • 8/9/2019 Volume 05 - Search 3rd Edition

    36/153

     

    36 [email protected] Search Visibility

    8. 

    Make sure navigational menus do not cover the search box.

    On some sites, dropdown navigational menus were so sensitive to mouse movementthat users repeatedly triggered them accidentally as they were trying to clickelsewhere on the page. In many cases, this resulted in the search box being coveredby navigation, making it difficult for users to enter a search term.

    This happened to a user on BestBuy.com. When he tried to move his mouse to thesearch box, he accidentally triggered the navigational menus by moving his mouseslightly into the navigational target areas in the main horizontal navigation. He hadto try to move his mouse to the open text field multiple times before he did sowithout triggering the sensitive menus.

    If such menus appear above the search box, ensure they are not easily triggered byusers attempting to search. Delay the response slightly or increase the spacebetween the menus and the open search field.

    A user had trouble entering a search term on BestBuy.com when the navigationalmenus repeatedly covered the open search field. The second half of the search boxcan be seen to the right of the navigational menu in the screenshot above.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    37/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 37

    A later redesign moved the search box above the site navigation.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    38/153

     

    38 [email protected] Supporting Users’ Searches

    Supporting Users’ Searches

    9. 

    Regularly review search logs for search trends and users’

    terminology.

    The search box is the way users can “talk” to the site. Search logs can reveal a lot

    about how users think about your merchandise, shopping trends, problematicnavigational areas, and offerings that are missing from your site.

    Review search logs to see what terminology users have for your products and fortheir needs. Make sure that your search engine accommodates the searches usersare conducting.

    Look for patterns in how users search. A sudden spike in searches around aparticular item or term may indicate an article, blog, or TV mention of an item thathas suddenly raised its popularity.

    Repeated searches for certain types of merchandise can indicate that users don’tknow how to otherwise locate those items. While some users prefer to search, othersturn to search as a last resort when navigation fails them. A high number of searches

    may indicate users cannot otherwise locate the items. A high number of searches for “picture frames” may indicate the Housewares category label isn’t clear, for instance.

    Also, consider the search engine as informal market research. If users repeatedlysearch for “tablecloth” but your site only offers napkins, consider if it makes sense tooffer those related items.

    Search logs can also clue you in to common misspellings. If users regularly misspellbrand or product names, ensure those misspellings take them to appropriate results.

    10.  Adjust the search engine based on search logs.

    After analyzing search logs, make any necessary adjustments to the search engine

    to accommodate users. For instance:

    •  Decide whether there are any frequent queries that need special treatment

    •  Configure the search engine to recognize common synonyms

    •  Accommodate misspellings

    Determine if the search engine is returning appropriate results to users. Seeingrepeated searches for similar terms, such as “sock” followed by “socks” and “stockings” may be an indication that no results are being returned. Make sure yourtop search queries are returning relevant results to users, and those results arebeing presented at the top of the search results list.

    11. 

    Adjust content or offerings based on search logs.

    Consider what the search engine can tell you about the site and its content.Keywords and topics that appear in search logs may make for good FAQs or Help content. Products searched for, but not offered, on the site might make goodinventory additions.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    39/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 39

    12. 

    Consider special treatment for frequent queries, such as

    categories or brands.

    In our first study, eToys delivered custom results pages for its most popular toys,but SmarterKids returned standard results for all toy queries.

    When our users typed the name of a popular toy, such as Thomas the Tank Engine,Barney or Hot Wheels into the eToys search box, the results were specially designedcategory pages rather than a list of links. This worked well for most of our users andthey thought it was helpful. However, some users were confused because the pagelayout differed from the search results page they expected.

    Like eToys, Office Depot sometimes took users to a category page instead of asearch results page. For instance, if users searched for “shredders” they received ashredder-focused page with links to types of shredders and options for narrowingresults. This was helpful, in that the page looked similar enough to search resultsthat our users were not confused, and the page offered helpful links. However, userswere confused by the jumbled layout on the page, not knowing if they weresupposed to select from the Heavy- to Light-Duty  links at the top, look at featured

    items, or narrow their options by criteria on the side of the page.

    While it was helpful that Office Depot took users who searched for “shredders” to aproduct-focused page, users weren’t sure what to do next on the page becausethere were so many options.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    40/153

    40 [email protected] Supporting Users’ Searches

    If frequent search queries take users to a specific page, ensure that page clearlylists product options. Costco.com did a nice job taking users who searched for “televisions” to a category page with promoted items at the top, and category andbrand options below. The page also had categories listed on the side of the page.This was more successful than the Office Depot example because the categorieswere clear.

    Costco’s site took users to a category page for Televisions when they searched for “television” or “TV.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    41/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 41

    13. 

    Support many types of searches, including non-product

    searches.

    Users searched for all of the following:

    •  Product number (BC3104, s770)

    •  Customer service terms (returns, delivery)

    •  Product names (iPod shuffle, Richest Man in Babylon)

    •  Category names (air conditioner, digital camera)

    •  Criteria, including brand names (ballet, Uniden)

    •  Categories and criteria (Canon printer, peanut candy bar)

    For instance, a user on Boden’s UK site was trying to get a sense of what the sitesold, so searched for “shoes.” A user on Nike.com wanted a hat with specific colorson it, so searched for “grey white purple hat.” (See page 11 for a breakdown ofsearches by type.)

    Users also searched for other site content, like a company address. They alsomisspelled search terms.

    A robust search will accommodate each of these types of searches. Some sitessimply accommodate searches for specific, properly spelled product names. Manysites neglect to support searches for site content, including crucial customer serviceinformation.

    When considering what terms the search needs to support, review search logs asmentioned in the guideline above, but also think of other sources. Talk to anyone inthe company who regularly has contact with customers, such as a call center forcatalog orders or a support desk for the website, or even store clerks. Find out whatcustomers ask for and how they phrase their questions.

    Knowing customers’ terminology is essential. One user on FineStationery.comwanted to buy, “You know, those stickers you put on your mail going out, in thecorner? That’s what I want.” He looked in the navigation under Corporate andStationery , but then turned to search, entering the query “sticker.”

    The search didn’t return any results for return address labels, which is what he waslooking for.

    On Bluefly.com, a user saw the option to navigate to Lightweight & Rainwear  optionswithin coats. When she later tried to search for “rainwear” she received no matches,though it was a category name on the site. She later searched for “lightweight” andsaw products with the term in their description or name, but no way to navigate tothe category.

    Product criteria should be searchable as well as product names. A user onFlight001.com was looking for a guide book for England. When he searched for “England,” he only received one result: a travel adapter. The site offered a travelguide for London, but it did not show up. He said, “They have an adapter. No booksor guides for England? But there’s a London city guide right there. Why didn’t thesearch for England bring that up?”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    42/153

     

    42 [email protected] Supporting Users’ Searches

    When a user searched for "England" on Flight001, no guide books displayed, eventhough the user had previously seen a city guide to London.

    It is also essential to support searches for customer service information. Many usersturned to search when trying to locate policy or return information. For instance, a

    user on Anthropologie's site was buying a gift for a friend and wanted to make surebefore buying it that his friend could return it. He searched the site for "returnpolicy" and received product results, but no information about the site's returnpolicy.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    43/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 43

    A search for "return policy" on Anthropologie.com returned only product results.

    By contrast, a search for “returns” on Build.com took users to the full return policy.

    14.  Indicate what users can search for, when appropriate.

    Searches should accommodate any term users enter. If the search accommodatessomething that a user may not expect, or if the search works in conjunction withproduct numbers in a mailed catalog, for instance, it can be worth telling users theycan enter such queries.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    44/153

     

    44 [email protected] Supporting Users’ Searches

    For example, the Boden site offered a Product  Code search field, followed by aQuickshop button. This allowed users to enter product codes from the print catalogto find desired items quickly. The field was separate from the main search, with itsown label and button, and will filler text in it.

    There may be good reason to explicitly support such searches, if users commonlyuse product codes to search for items on the site. However, consider if a separatesearch box is required. A single search box could be used, with text beneathindicating the types of queries the box supported, including product codes.

    When one user saw the Product Code search box before she saw the regular sitesearch, she said, “You have to type in a product code. Oh, no you don’t.” Userslooking for a standard search may be confused by seeing multiple options, whileusers looking to enter a product code should know to put it in the only search field,particularly if the field is accompanied by sample searches listed underneath thefield.

    Multiple search boxes at the top of the page caused one user to think she needed aproduct code to conduct a search.

    A single search box with explanatory text can be enough to indicate what types of

    searches user can conduct. Oriental Trading used filler text to indicate users couldsearch by item number or keyword. We advise against using filler text (see guidelineon page 28), and recommend that the information be placed below the field instead.The site also had a link to Catalog Quick Order , knowing many of their sales camefrom recipients of their catalogs, which also led to a specialized search allowing usersto enter an item number and quantity. This added items directly into the shoppingcart.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    45/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 45

    Oriental Trading allowed users to enter an item number in the main search box andalso provided a separate Catalog Quick Order  option, which allowed users to enteran item number and quantity and add items directly to the shopping cart.

    A user on 1000bulbs.com appreciated the option to search by image, since she had alight bulb that she wanted to replace. She quickly identified her bulb as a halogenmini and saw the options the site had to offer. This was much simpler than herexperience on other sites, where she tried to read tiny numbers and letters on thebulb to enter in site search engines.

    In this case, it made good sense to offer a visual search. This shows that the site

    realized users often don’t know the details of a light bulb they’re trying to replace,but that they shop by sight, looking for one that matches the burned out bulb.

    The placement of the link to Search by Image was problematic, though. The userwho relied on this search successfully stumbled upon it the first time she searched,but when she tried to do the same search again, she accidentally clicked Search rather than Search by Image. It was helpful to place the button next to the standardsearch, as that’s how the user originally stumbled upon it. However, it was presentedin the same way as the standard search with little space between the buttons,making it easy for users to confuse the buttons or accidentally click the wrong one.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    46/153

     

    46 [email protected] Supporting Users’ Searches

    1000bulbs.com offered an image search, allowing users to navigate to the rightproduct type based on what the light bulb looked like. This showed awareness ofhow shoppers look for replacement light bulbs.

    Navigating to a category showed the various types of bulbs available, along withdetailed images to help the user select the appropriate bulb.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    47/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 47

    15. 

    Consider displaying suggested search terms as the user

    types his query. (Ensure suggested terms return relevantresults.)

    It can be helpful to show users suggested search terms as they are entering a query.This can help users pick an appropriate term which will show them results, ratherthan making up a query which might not be successful. Such results typically appear

    beneath the search box and change as users type each letter of their query.

    This can be particularly helpful on travel-related sites, where the site relies onaccurate location names to return results. One user on Orbitz, however, wasconfused when he was searching for a hotel in Chicago. Search options appeared ashe entered “Chicago” and he wasn’t sure which to click. He said, “When I wroteChicago, a few different places came up. I’m not sure which one, but I’ll just go forChicago, Illinois.” He was British and unfamiliar with U.S. states, so wasn’t sure hepicked the correct option.

    A user in London used the suggested search terms presented by Orbitz to select hisdestination.

    If suggested queries are used, they must return relevant results. Suggesting that auser might be searching for “candy bars” is not helpful unless selecting that termreturns all available candy bars on the site. Suggested terms that return zero resultsor one result aren’t helpful.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    48/153

     

    48 [email protected] Supporting Users’ Searches

    Foot Locker provided suggested searches based on the user’s entry in the searchbox.

    Some sites provided search results as users typed. This can be a good shortcut forusers who know what they were looking for or se an item of interest. Other shopperswho are looking for all items that match their search query prefer to see the full listof results so they can browse and compare. Allow users to see a full list of resultseither by clicking the Search button, using the Enter  key, or clicking to View AllResults.

    Some sites, such as Urban Outfitters, automatically listed search results as userstyped. This was a handy shortcut for users looking for specific items, but stillallowed users who wanted to browse results to use the search box in the standardway. Urban Outfitter’s design would have been better if the search results weremore clearly differentiated from the background, with a strong border around theresults.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    49/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 49

    16. 

    Carefully categorize products to ensure solid search results.

    Users were frustrated when site searches returned items that clearly did not matchtheir inquiries or when products the site offered did not appear in results.

    On 1-800-Flowers, a user searched for “yellow” in order to find yellow roses. Thesearch returned 21 yellow items – flowers, candles and such – but no yellow roses,even though they were available on the site. Unable to find yellow roses, he said, “IfI were doing this at home, at this point, I’d say screw it. I’d have gone to my florist.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    50/153

     

    50 [email protected] Supporting Users’ Searches

    When users search for a color or finish, they expect search results to reflect theirquery. These results on Build.com for “red” include items available in that color, butthe default product image is shown, which is not always red.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    51/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 51

    In Urban Outfitters’ search results, (mostly) yellow products are returned for asearch for that color (there is one lavender shoe returned in the first row of results.)Rather than showing the default image, the search results return images of theyellow version of each matching item.

    A user on Disney’s site carefully narrowed the scope of her search to entertainmentfor a youth boy . When the search results came back, she immediately noticed thatthe website had not adhered to her search criteria. “I searched for something for aboy. This shows a girl with a Dalmatians pajama set,” she complained.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    52/153

     

    52 [email protected] Supporting Users’ Searches

    17. 

    Accept synonyms commonly used by customers.

    Customers may not use the same language to describe the same product. Forinstance, users looking for a car seat searched for “child seat,” “children’s seat” and “child restraints.” A user looking for bunk beds searched for “children’s bed.” Makesure the search engine accommodates synonyms, even if those words do not appearon the website.

    A user on Pottery Barn’s site searched for “drawers” when looking for a dresser, andreceived results that included no dressers. He never successfully located the dresserssection of the site, so wasn’t sure there were any available. He searched twice for “drawers” and meticulously went through the results trying to find something tomeet his needs, but no dressers were listed. Poor search results made him doubt thesite carried the item he was looking for.

    A Pottery Barn search for “drawers” returned no dressers, but plenty of other itemswith drawers, such as file cabinets and bookcases. The site offered dressers, butnone appeared in the search results.

    A user looking for printer ink on a variety of websites was stymied because he keptsearching for “red ink,” when the ink was actually magenta. Consider how your usersdescribe your products or their characteristics. While ink may come in cyan andmagenta, for instance, users think of those same ink cartridges as blue and red.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    53/153

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 53

    18. Accommodate misspellings and variant forms (spaces,

    plurals, etc).

    Users appreciated when search engines accommodated typographical errors ormisspellings. When searches failed due to users’ typing errors, they assumed the sitedid not offer the products they were looking for. For instance, a user was looking forthe song Guantanamera on the Sony Connect site. He did not know how to spell it,

    so searched for “juan tanamera,” which returned no relevant results. He searchedagain for the misspelling as one word, “juantanamera,” and received results.

    He said, “I think that either way, I should have received results. Sony Connectshould improve their search engine. It would be nice if they had a contact buttonwhere consumers could tell them about problems encountered when searching forspecific songs.”

    Users appreciated when searches simply worked, regardless of misspellings. A usersaid of Home Depot’s site: “I mistyped dehumidifiers as ‘dehumidifer’ with no third ‘i', but they still came up with results.” He made the same spelling error at Lowes,Sears and Walmart. Only Walmart failed to return results for his misspelling: “I’mreally surprised that Walmart’s site is not programmed to search for similar products

    or look for misspellings.”

    Costco.com corrects misspellings and indicated that the site was “showing resultsfor ‘dehumidifier’” and provided a link to “search instead for ‘dehumidifer.’”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    54/153

     

    54 [email protected] Supporting Users’ Searches

    Users are so accustomed to search engines correcting their spelling that theysometimes assumed the site had no matching items, even if they knew theymisspelled their search query. One user mistyped a search for “James Bond ultimateedition dved set” on Amazon.com and received no results. She said, “I bet they don’thave it, because it didn’t come up even with my typo.” She ran the search again,removing the extra e in dved  and saw the item she wanted as the first result. She

    simply expected the site to fix her mistake and return her desired item.Some users had work-arounds when they were unsure of the spelling of a word. Auser searched for “Arnold” rather than “Schwarzenegger” when searching for theactor’s movies. Another abandoned a search when he wasn’t sure how to spell JackNicholson’s last name.

    On FYE.com, a user thought the site did not carry the artist he was looking for due toa spelling error. He searched for the artist “littlefeat,” though the artist’s name has aspace between the two words, which he did not include. He received no results. Hesaid, “They don’t have it. It’s not on here, see?” The site didn’t offer anysuggestions, either for the spelling or for similar products he might enjoy.

    A user’s search for “littlefeat” rather than “little feat,” returned no results on

    FYE.com. The site did carry items by that artist.A user on YesAsia.com searched for “dianna krall,” misspelling artist Diana Krall’sname. He saw no results, corrected the spelling, and searched again. He said, “WhenI did the search it said no matches. For some others, like the Yahoo search machine,if you make a typo it tells you ‘are you looking for Diana Krall.’ If yes, you just click itand it takes you to that product.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    55/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 55

    Ensure that the search engine accommodates singular and plural terms as well. OnLiving.com, several users searched for “bunk beds” and received no results. “Bunkbed,” however, returned results. One user said after a facilitator helped with theunexpected problem, “I really disliked having to take away the ‘s’ on bunk beds. Ithought it was almost an insurmountable problem. I would have never guessed.”

    19.  Accommodate searches with multiple words, including

    characteristics such as color, size and brand.

    Users sometimes did not understand the search results they received for multiple-word queries. Some sites looked only for exact matches, and some returned resultsfor each keyword, but did not prioritize results that matched all terms.

    In our first study, 54% of queries consisted of more than one word. In our secondstudy, 71% of all in-site queries consisted of more than one word, with queriesaveraging 2.3 words. Of all sitewide searches, only 4 queries used enclosing quotes.

    A user on Sears.com searched for “10v cordless power drill” and received no results.Sears had 33 cordless drills at the time, but none were exactly 10 volts. However,

    the site could have returned “Nothing found for 10v cordless drills, but here arematches for cordless drills.”

    Similarly, a user searched on Pottery Barn’s site for “espresso dresser drawers,” andreceived no results. She searched again for “espresso dresser” and received results.While it was good the second search returned results, there was no clear reason thefirst did not. Dressers on the site were available in espresso stain and had drawers.

    After the first search, she said, “They probably don’t have it,” and went back tobrowsing the dressers page of the site. Unsatisfied, she returned to search. However,if she were not in a study situation, she likely would have left the site.

  • 8/9/2019 Volume 05 - Search 3rd Edition

    56/153

     

    56 [email protected] Supporting Users’ Searches

    A search for “espresso dresser drawers” returned no results on Pottery Barn’s site.

    A search for “espresso dresser” returned products which weren’t returned for thesimilar search “espresso dresser drawers.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    57/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 57

    Another user searched for “drawers espresso stain” and again for “espresso stain,drawers.” While this search returned results, it returned no dressers. Instead, shesaw a mirror and two file cabinets.

    A search for “drawers espresso stain” returned no dressers, but only a mirror andtwo file cabinets.

    When results were listed for multiple-word queries, users expected items matchingevery word in the query would be listed first. Users on The Container Store’s sitewere frustrated by the site search, which they did not think prioritized results

    properly.Several users on the site searched for the term “wooden box.” The site returned 100results for the search, and apparently first listed items with the word “wooden” in theproduct name, followed by items with the word “box” in the name. This meant thebox users were looking for was returned as result 96 out of 100, though it was awooden box. Several users gave up scanning the results page before they reachedthe result when they did not see the box immediately.

    One user said after the task, “There were very broad results for search. I had to findthe closest things that were coming up. So many results weren’t relevant to what Isearched.” Another said, “I’m searching for wooden boxes, but I’m getting boxes andnot wooden boxes. I shouldn’t be getting white gift boxes.” She later complained, “Ihad over 100 hits, with everything from bamboo to plastic to cardboard, and I

    searched for ‘wooden.’”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    58/153

     

    58 [email protected] Supporting Users’ Searches

    Users searching for this box on The Container Store’s site tried a query for “woodenbox.”

    The top results for the term “wooden box” included wooden items followed byboxes, but wooden boxes were not prioritized. The box users were looking for wasitem 96 out of 100 returned results.

    Users were frustrated when they searched and saw items that seemed irrelevant tothe query they had entered. Another user on The Container Store searched for avariety of different terms describing a set of stainless steel canisters with clear tops.She said, “No matter what I type in, I get the same 100 results.” While that wasn’tthe case, it was true that every search did seem to return 100 results, many ofwhich seemed irrelevant at first glance. For instance, she searched for “steel and

    glass canister” and received results that were “ridiculous – toilet brushes, cracker jars.” The items returned were either stainless steel or glass, but to her they werecompletely unrelated to what she was looking for.

    She searched for “kitchen canisters” and said, “A third of the way down the page,I’m not even getting canisters anymore. This has nothing to do with what I typed.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    59/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 59

    20. 

    Recognize search operators.

    Some people will try to use search techniques they’ve learned from other searchtools on your site. While most user will stick to basic search queries, some mayemploy more advanced search methods. Here are some examples of specialcharacters your search engine should respond to intelligently:

     

    software AND space•  (lincoln county)

    •  +mr. john smith

    •  *price*

    •   “winnie the pooh”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    60/153

     

    60 [email protected] Advanced and Scoped Searches

    Advanced and Scoped Searches

    21. 

    Consider the need for advanced search.

    Most users prefer to use simple search. If advanced search is offered, use a distinctlink that is less prominent than the simple search’s button.

    Rather than offering advanced search, consider expanding the types of searches yoursearch tool can accommodate or using facets to help users narrow results (see page105 for more on facets.)

    22. 

    Provide a clear link to advanced search – and back.

    If your customers require advanced search functionality, offer it; but keep simplesearch as the default and allow an easy return.

    Advanced search offers manual control over many parameters that refine the searchprocess. The main problem with advanced search, however, is that people mayinterpret the additional parameters to mean that they can adjust the searchdifferently than the designer intended. Here are some examples from Reel:

    •  A Danish user thought that if he selected “Country = Denmark,” he could thensearch for the Danish title (“Gøgereden”) of the movie “One Flew over theCuckoo’s Nest.”

    •  A user thought that if he wanted to search for an English actor, he had to set “Country = England.”

    23. 

    Explain the scope of the search.

    A site search may cover the full site or part of the site or site inventory. It may covera particular attribute, such as Artist  or Author . Most users assume that a search will

    cover the entire site and all its inventory. It must be clear when and if this is not thecase. A dropdown menu can indicate the options for scope and search results shouldalso restate the scope of the search.

    It is difficult for sites to predict what type of search users want to conduct on anygiven page. Within a section, they may want to search that section or they may befinished shopping there and want to look for an entirely different type of product.

    For instance, some of the users on the Sears site started looking for a microwaveoven when they were in the Tool  Territory  area of the site. On this site, that meantthe Tools department was the implicit scope of the search. Several users searchedfor “microwave” and received no results.

    One user said, “I know a microwave is not a tool, but I thought maybe it would bring

    me to the appliance section.” Another user remarked, “This reminds me of lookingunder the Yellow Pages. You’re looking under lawyers, and it’s attorneys. I think I’min the tool thing. I have to get out of the tool thing.”

  • 8/9/2019 Volume 05 - Search 3rd Edition

    61/153

     

    © NIELSEN NORMAN GROUP WWW.NNGROUP.COM 61

    24. 

    Default searches to the most lenient category.

    Search boxes should default to the most lenient search option, allowing users torefine if needed. Every selection list should include an option for All  or Entire Site.

    Many users simply ignore options to limit the scope of a search. This is not aproblem, so long as the option defaults to A