Our No. 1 guideline for SERP design is to mimic the SERPs on the major Web-wide search engines. Provide a linear list of search results, with the most recommended on the top.
There is no need to number search results because all users start scanning from the top. There is also no need to annotate search results with relevancy rankings or other estimates of goodness. Users start scanning from the top and give up once the search results don't seem promising enough, based on the content, not where it rates in relevancy.
Each search result should start with a clickable headline, followed by a two- to three-line summary. The headline is the most importantin fact, the first words of the headline are the most important of all, since users don't even read an entire headline as they scan the search results. It's important to start the headline with the most information-carrying words that indicate what the page is about. The summary should elaborate on the headline without repeating it.
This SERP is easy to scan and has good headlines and reasonably good summaries. It is easy to page through the results because the arrow symbols at the top and bottom of the page provide nice big target links. The design of this page could be improved by removing the relevance percentages. People are focused on the titles and descriptions, not the percentages. Also, the user's query is repeated in the site's standard Search box in the upper-left corner, but on the SERP it's better to show the query in a wider Search box right on top of the results, so that the user will be more likely to spot it and use it to reformulate the queries.
On some sites, you can add an indication of the address of the document below its summaryeither the URL or the name of the area of the site that contains the document. But on most sites, this information is irrelevant and should be omitted in the interest of brevity.
You can also list the date when the document was updated, provided that this was in fact when real writing or editing was performed. If the dates only indicate when a typo was fixed or the document was moved to a different server, the feature is more misleading than useful.
This multi-national company's SERP does some things right but most things wrong. What's good: It repeats the user's query in an editable Search box and each Search hit has the expected format, with a headline, a summary that highlights occurrences of the query terms, and the URL. There's also an easy way to page through the results for those few users who want additional SERPs.
What's bad: There is no need to offer features to search for "all or any words." This should be relegated to Advanced Search instead of cluttering up the default Search page. The ability to search either all servers or regional servers is also an unwanted complication, and it should not be necessary to specify e-commerce as a special option. Often people searching for something that's for sale want the product page where they can buy it.
Because of the many extra features, the actual search results are pushed too far down the page. You can't even see the first hit in its entirety above the fold if you are using a screen with a resolution of 1024 by 768 pixels. There is no need to show the relevancy ratings, and it would be better to integrate the line for an item's location with the URL line in order to fit more search results on the screen. Finally, of course, the German headline in the fourth hit should not be used. English-speaking users are quite likely to skip this item, particularly since the summary doesn't make it clear that the article is in English.
Users rarely reformulate their queries if their first ones don't succeed. In the study we conducted for this book, users made do with a single query 83 percent of the time. In only 17 percent of searches did users try alternate queries. Still, it is helpful to make sure that users can perform query reformulation on the SERP: Include a Search box and prepopulate it with the query the user just issued so that they can modify it by simple editing.
Finally, the SERP can provide a link to the Advanced Search page. This link can be repeated at the bottom of the page, where users will see it if they scan to the end of the list and still haven't found what they were looking for.
Ideally Search software would always place the most important page for any query on top. In practice this doesn't happen because the computer is just a stupid machine that doesn't understand your business. It can only score pages by indirect measures, such as how frequently they use various words and how frequently other pages link to them.
In order to have your important pages rank above your secondary pages in Web-wide search engines, you have to fiddle with these secondary markers. This is known as Search Engine Optimization (SEO) and it is discussed in more depth later in this chapter.
Unfortunately you can't tell an external search engine what pages on your sites are the most important for each querythe search engine won't trust you and will try to second-guess you. But you can tell the search engine on your site what pages to put on top for important queries. This functionality is often known as "best bets," because you are betting that the pages you designate are going to be the best matches for your users' queries.
When you add best bets to your search engine, you must manually construct a list of the best page or few pages for various user queries. If a user enters one of these queries, the search engine will simply place the best bets on top and then follow with its own algorithmically sorted results. It is best not to differentiate between your manually selected best bets and the computer-selected ones. Since users start scanning the SERP from the top, the best bets automatically get special prominence. If you put them in a box or otherwise highlight them, users may think that they are ads and ignore them.
A particularly fertile source of best bets is the query terms that your users enter most. Check your Search logs for the 100 or so most popular queries, enter each into your search engine, and see what comes up. If the top hit is not the most important page for that term, add it to your best bets list.
Once you have checked the 100 most frequent queries, continue down the list and check the next 100 queries. Continue until you have constructed best bets for all frequent queries.
Hewlett-Packard's SERP calls out a categorized set of related items in a column to the left of the main results. This is an interesting way of supplementing the best bet, which is the product page for the item the user searched for. For example, when we searched for a popular printer, the search engine appropriately placed the printer's product page on top of the main listings, but it also placed links to printing supplies for this printer in the left column. The page also uses the international dating convention, even though the search was performed on the U.S. area of the site. It would have been even better to spell out the name of each month.
Sorting the SERP
Usually the SERP should be sorted by relevance, and no other options should be made available, since they will only confuse the user. For some types of sites, however, it makes sense to make other sorting criteria available. For example, sorting by date instead of relevance is useful for people who want to monitor recent events or find out what has changed on a site. Most people don't care about your periodic changes or updates, but for some site genres recent developments are of special importance. This is certainly true for news sites.
Giving users the flexibility to sort search results by attributesuch as pricecan be very helpful. This is often the case on e-commerce sites or other product-heavy sites where users are searching for products rather than language documents.
Sorting by date can be useful on several types of Web sites beyond news sites. For example, users who want to assess the future product strategy of Yahoo! can search the U.S. Patent and Trademark Office to find out what patents the company has applied for recently.
In most cases, it's best to offer users both ascending and descending views, and it's common to be able to toggle between the two. The default direction of the sort should be the one that makes the most sense to most users. For example, when sorting by date, place the newest materials on top for the search results. If the user clicks the date column, switch to showing the oldest stuff first. When sorting by price, show the cheapest items first and the most expensive items second.
(Facing page) This SERP allows for sorting by three attributes of the products: name (which would mean brand if all products are named with the brand as the first word), delivery date, and price. Since the user in this screen shot searched for memory cards, it would have been nice if the SERP also allowed for sorting the list by storage capacity. Here the user has just clicked on the pricing column. A second click on "Sorted by Price" will reverse the sort order and place the most expensive products on top. (Why would anybody want the most expensive items? In this case, you may want to get the highest-capacity cards, and price is a way of identifying them when the system doesn't recognize capacity as a sortable attribute.)
The product thumbnails are not that useful for the type of products in this screen shot, though they do provide a fast way of scanning for brands. For items that are more visually differentiated, however, the thumbnails can be helpful. The main problem on this SERP is that the items are spaced too far apart because rebate offers and other features take up a lot of room. It is more difficult to use a SERP if users can see only a small number of items at a glance.
No Results Found
When no results to a user's query are found, a special-case SERP is necessary. The first requirement for this page is that it clearly state that no results were found. Returning a blank page may make users sit and wait for the search to finish because there is no indication that it's already completed. Blank pages are also likely to make users think that something is broken.
Second, a no-results SERP should help users modify their queries to get better results. As with all SERPs, it should repeat users' queries in a Search box so that they can edit their queries easily, but it should also provide explicit hints for how to correct typical search mistakes. If you have a spelling correction featureas we recommendit should display suggestions for revised spellings prominently.
Finally, end the page with a link to your Advanced Search, if you have it. Although Advanced Search is usually more confusing than helpful, anything is better than finding nothing, and it's possible that the Advanced Search features will help users come up with a query that will work.
One Result Found
Another special case is when only one result is found. Some sites dispense with the SERP in this case and take you directly to the page that was found. We recommend against this for most sites because it violates user expectations for Search, which include seeing a SERP after issuing a query. Users can get confused when they see a content page instead of a listing of links after they have clicked the Search button.
Another reason to show the SERP even if there's a single hit is that users may want to modify their query once they see how little it found. You also need the SERP to show users alternative spellings or other features you have for query refinement.
The SERP can be skipped when users conduct a known-item queryfor example, if they enter a SKU, part number, or other unique identifier that can't be considered full-text retrieval. In such cases, go straight to the page for that SKU.