How to Stop Searching and Start Finding

How to Stop Searching and Start FindingTired of wading through useless search-engine results? We test 20 engines to pinpoint the ones that really deliver the goods.Kim Zetter and Harry McCracken

Ask a couple of strangers for street directions, and you never know what you'll get. They may be rushed and give you only partial instructions; they may lead you astray with incorrect information; or they may not know the way at all and simply shrug you off with a smile. Still, when you find yourself going past the same mini-mart for the seventh time, you may have no choice but to ask for guidance. But who do you approach: The twenty-something gas-station attendant? The woman washing her car? The man walking his dog?

Searching for anything on the Web carries the same uncertainty. You may know exactly what you're looking for, but you don't know where to find it or who to ask. And when you do ask, you can't be sure you'll get the right answer.

The Web is crawling with search sites vying to show you the way. Conservative estimates put the number at over 8000, including pure search engines, general- and special-interest directories, and metasearchers (which query multiple search sites at once). In addition, there are numerous alternative search methods you could opt for, including utilities that work with your browser to mine the Web for information (see "Search Utilities,") and "expert" sites where real people field your questions.

With so many choices, how do you decide which search site to use, and when? We tested over 20 engines, directories, and expert sites to see which ones produced the best results--that is, the most relevant links in the most logical ranking, with the least effort on our part.

We ran a series of queries, ranging from broad to specific, on each site. Some were business-related, such as one that tried to track down certified public accountants in California; others were scholarly, including one that sought information about a lost Hemingway manuscript (his wife left it on a train in 1922). We also hunted for information on pop-culture topics ranging from Pokémon to Queen (Freddie Mercury's band, not Victoria, Elizabeth, or Latifah).

Some sites did surprisingly well; others forced us to wade through swamps of irrelevant links or pointed us to pages that no longer existed. (For highlights of what we found, see "Search Superstars" and the features chart.) Overall, we discovered that search technology, while far from perfect, has made great advances in helping us find our way efficiently through the Web.

Engines and Directories: The Basics

Despite the profusion of methods for searching the Net, the most common tools remain search engines and directories, each of which has its own merits. The main difference between the two is human intervention. Search engines use automated software "spiders" that crawl through the Web to collect and index the full text of pages that they find.

Directories, by contrast, rely on human editors to sift through pages, winnowing out inappropriate ones and categorizing sites by subject. Nothing goes into the directory unless an editor approves it, so you're unlikely to find Christmas recipes in a Yahoo category about ham radios. But since directories are crafted by hand, they are far less comprehensive than search engines. Conservative estimates put the current size of the Web at a billion pages (and gauge its growth at a million pages per day). Most engines don't come close to indexing the whole Net, but at press time Google claimed to have indexed a billion pages either partially or completely. The largest directory, called the Open Directory Project, is tiny in comparison, with a current index of only about 2 million sites.

Hybrid Search Sites

Directories and search engines have different virtues, so most major search sites meld aspects of each. AltaVista, for example, supplements its search engine with a directory that uses listings from the LookSmart and Open Directory indexes. Similarly, if you search at LookSmart or Yahoo, the site first provides results from its own directory, and then passes your query to a search engine.

Of course, many search sites do a lot more than provide an engine and a directory--Excite, Go.com, Lycos, Yahoo, and others have morphed into portals offering an array of services, including stock quotes, news, e-mail, shopping, and anything else to dissuade you from going elsewhere on the Web. That scattershot strategy keeps visitors glued to pages long enough to notice the banner ads, helping the sites make a buck. But while much of the stuff portals offer is useful or fun, these sites can become so cluttered with peripheral features that the search tools seem an afterthought.

Some search sites are returning to the basics. Google started the trend last year: Its sleek home page consists of little more than the Google logo, the field where you type in your queries, and a couple of buttons. More recently, AltaVista launched Raging Search, which offers the AltaVista search engine without the clutter of the AltaVista site. AltaVista says that searches on the two sites can yield different results, but our tests produced identical listings each time. In any event, pure search sites are excellent news for Web surfers who want to avoid the distractions of a full-service portal.

Search Engines: The Race for Relevancy

Logically you'd expect the search engine that indexes the most pages to have the best chance of finding what you need. But if an engine doesn't properly organize the sites it finds, the one you want may be buried beneath thousands of irrelevant links. What good does it do for an engine to deliver 6000 environmental links when you really want a site for the rock group Green Day?

At Lycos, for example, we searched for Ford Motor Company: The first three links consisted of one to Lycos's own motorcycle section, one to the Nature Company, and a third to Living.com, an online furniture store. At Ask Jeeves, which lets you pose queries in the form of a full statement or question, we asked Who is John Kerry? in hopes of finding a biography of the Massachusetts senator. But Jeeves' first link led to geographical information on Kerry, Ireland.

Of the sites we tested, Google provided the most consistently pertinent results--one reason it's our favorite engine overall. The site uses "page rank" technology to track the number of pages that link to a site. If a lot of pages link to a particular site on a specific topic, the reasoning goes, that site must be relevant to that subject. Consequently, Google gives it higher placement in the results.

Google has such confidence in this theory that it offers an "I'm feeling lucky" button you can click to go directly to the site Google thinks is most relevant to your search. But it might more appropriately be called "We're feeling lucky," since Google is gambling that the site it selects is the one you want. In our testing, Google's optimism was justified in some cases. We searched for Al Gore's official campaign site using the keywords Al Gore campaign site. Google listed algore2000.com as the first item, whereas GoTo.com buried it beneath nearly 70 other links.

Google was less fortunate when we searched with the keywords world beat, hoping for information on music; the engine sent us to the Web site of travel publisher Rough Guides. However, when we put quotation marks around the query ("world beat"), Google's luck (and ours) improved: It sent us to the highly relevant world beat page at the Internet Underground Music Archive--proof that subtle differences in how you enter a query can have a big impact on the results you receive.

Another Google plus: Below each link it finds, Google provides a snippet of text with the word or words you searched for highlighted in bold. That helps you eyeball the results and gauge the relevance of each link quickly. (Most engines simply display the first line or two of text from each linked page, whether the text contains your search terms or not.)

Direct Hit takes a different approach to maximizing the relevance of results by organizing them based on their popularity with previous searchers. For instance, if few searchers clicked the first link for a search about computer chips (say, because it is actually for a site about potato chips), that link will sink lower on the list. Several engines, including HotBot, Lycos, and MSN Search, have adopted Direct Hit's technology. Our tests of Direct Hit itself, however, produced mixed results. When we searched for Pokémon and Queen, the engine served up the official site first in each case. But of ten results for world beat, only four were relevant. Most of the others weren't related to music at all.

Directories: Yahoo Versus Its Rivals

In 1994 when the World Wide Web first took off, your choices for searching the Net were pretty limited: Yahoo and, uh... Yahoo. The site remains a cornerstone of Web research, but as a directory it faces stiff competition from the Open Directory and from LookSmart.

All three are useful, but on balance, we like the Open Directory Project best. Spearheaded by Netscape, this project relies on more than 24,000 volunteer editors worldwide who have indexed nearly 2 million sites in over 200,000 categories. Any other search site can license Open Directory and use its database for their own search results, and many of them do: AltaVista, HotBot, Lycos, MetaCrawler, and some 100 other sites dip into it for links.

"I use HotBot for general, quick searches. It has a simple, clean feel, and pages load quickly. When I need muscle, I use MetaCrawler for broad category searches. When I want to drill down, I use AltaVista for specific topics" --Karl Ujanen, San Francisco network consultant.

You might expect the quality of the Open Directory's results to be erratic, since the site relies on the work of volunteers. But in our tests, it produced well-organized lists of pertinent sites, with clear descriptions of each link. (In some cases, Yahoo offered only cryptic descriptions, or none at all.) And the www.dmoz.org site has a pure-search feel similar to that of Google, with no shopping links or other distractions.

The descriptions at LookSmart are also superior to those at Yahoo. But LookSmart seems to have fewer listings for business sites than either Open Directory or Yahoo--a problem if you're trying to track down an office design firm in Detroit. And though it's not a full-service portal, LookSmart sometimes lards its results with ridiculously off-topic shopping links. Do a search for Buddhist philosophy, for instance, and LookSmart asks if you want "Local sales & coupons for 'Buddhist philosophy' at ShoppingList" before you check its search results.

No matter which directory you choose, all have one advantage over search engines: They're browsable, so you can wander through a series of related topics laid out in the site's menu. For instance, click on "Sports" at LookSmart and you get a drill-down list of topics like Baseball, Olympics, and Motor Sports. Click on "Motor Sports," and you can refine the search to a subtopic such as Formula One Racing or NASCAR.

Or suppose you need a company to make a sign for your Los Angeles-based business, but you know of only one supplier. You can perform a reverse query by running a directory search on the name of the company you know. Yahoo, for instance, will list the category that the company falls under ("Los Angeles>Business and Shopping>Business to Business>Signage"). Click on that category, and you'll get a list of sign manufacturers in LA. This approach can help you track down sites more quickly than you would by guessing at keywords.

Metasearchers

It's bound to happen: AltaVista has a link to exactly the Web site you're looking for. But you're searching at Google, which doesn't know that particular site from Adam.com. Enter metasearch engines (also called metacrawlers). Instead of maintaining their own Web indexes, these sites call on other engines and directories each time you issue a query. Most metasearchers combine results, weed out duplicates, and present you with more links than you'd get if you confined your search to a single engine.

In the early days of the Web when no single search engine had indexed more than a few million pages, metasearchers were essential. Now AltaVista, Google, and other engines boast indexes of hundreds of millions of pages and let you comb a significant percentage of the Web all by themselves. Still, the best metasearch engines are worth trying, especially at times when your favorite single-engine search site strikes out. Unfortunately, metasearchers often fail to match the relevance of Google--probably because it's so tough to mesh links from disparate sites coherently.

Today, the Web has numerous metasearcher sites. Among the most prominent are Dogpile, Mamma.com, MetaCrawler, ProFusion, Search.com, and TheBigHub.com. Each one melds results from a lineup of search sites: MetaCrawler, for example, draws from 13 engines--among them AltaVista, Excite, Google, and LookSmart--while TheBigHub.com relies on only eight. But the more engines a metasearcher queries, the longer your search can take. Each time you issue a query, the metasearcher must pass it to the other sites, retrieve the results, and fold them into a single list. By reducing the number of engines it uses, a metacrawler can keep response times to a reasonable length.

The popularity of metasearchers has encouraged many traditional engines to acquire metasearch-like features. Ask Jeeves, for instance, provides results from About.com, AltaVista, Excite, WebCrawler, and 4anything Network in addition to the results from its own database.

Searching for the Best

MetaCrawler, our favorite metasearcher, draws on solid search sites such as AltaVista and Google, which helps it bring back plentiful and pertinent results. It trolls through all sorts of Net resources, too. Search for Beethoven and you can click on tabs to leap between engine and directory results, MP3 music files, images, newsgroup discussions, and even online auctions relating to the great composer and his music.

Another worthwhile metasearch site is Search.com, largely because of its specialized metasearchers on dozens of topics. The Automotive one, for instance, queries Autobytel, Car and Driver, CarPrices.com, and nine other car sites. But Search.com's general-purpose metasearcher often starts by returning a link from GoTo.com, a search engine that invites sites to pay for higher placement in its results. When we searched for Dell Computer at Search.com, it referred us to GoTo's paid link to a company that leases Dell PCs before it identified Dell's own home page. But at least Dell's site was the second link at Search.com; at Mamma.com, it was number 18.

Specialty Search Engines

One of our favorite search sites has indexed a grand total of just 1600 sites. That achievement might seem negligible, except for one thing: Each of those 1600 sites is about beer. So if you seek a home brew club in Portland, Oregon, Beersite should be your search site of first resort.

From TaxTopic.com to the James T. Kirk Search Engine, single-topic search sites cover just about every subject imaginable. We found nine astronomy search sites, half a dozen on politics, and three on fishing. Within their specialty, the best are authoritative in ways that general-purpose engines rarely are, and they helped us find what we were looking for in almost all of our tests.

In addition, because these sites are so focused, the results are highly relevant: If you search for bacon at Recipes-For-All.com, you'll get recipes, not pictures of Kevin or essays by Sir Francis.

"For the past year I've been pretty faithful to AskJeeves.com. It checks I don't know how many search engines, and you can pose queries in natural English, like, Who was a better artist: Sid Vicious or Cyd Charisse?" --David Bock

How do you find out if there's a search site devoted to, say, springer spaniels? Simple: Consult a guide to topic-specific searchers. Two of our favorites are InvisibleWeb.com and Search Engine Guide.

Only a few specialized search sites use true automated engines. (One of these is Ditto.com, which helps you find images on the Web.) Most are actually hand-compiled directories to sites on a specific topic. As such, they're better for topic searches than for highly specific queries. For instance, search All Magic Guide for Houdini, and you'll get 15 useful links about the legendary illusionist. But query on harry houdini straitjacket escape photo, and it won't find anything.

Many specialty searchers are labors of love by one obsessed Netizen. Therefore, a site will likely grow stale if its proprietor loses interest or runs short of spare time. Look for a "Last updated" banner on the home page--and move on if the site seems to be in limbo.

Expert Sites

For years, search-engine companies have tried to make their sites think like human beings (without much success). So why not turn the tables? That's the idea behind expert sites--information centers powered by the knowledge and opinions of real people.

Leading contenders in this area include Abuzz, Ask Jeeves AnswerPoint, AskMe.com, and LookSmart Live. Most of the help at these sites is free. Exp.com's specialty is higher-end, fee-based advice from folks like attorneys and consultants; other sites, such as ExpertCity, devote themselves to computer questions.

The drill at most expert sites is similar: Browse a Yahoo-style topic directory until you find the category you want. Then post your query and wait for answers from site visitors ("experts") who claim knowledge in that subject.

Much of the time, this system delivers savvy, personalized assistance of a sort no search engine can. When we asked about buying a commuting bike, for instance, an AskMe expert shared his favorite urban biking links, and a seasoned cyclist at Abuzz provided a checklist to bring to the bike shop. Though response times from such sites are not instantaneous, we often received answers within an hour or two.

Where Are the Experts?

In the end, we liked AskMe.com's easy interface and friendly, responsive community of experts best. But even there, some queries went unanswered, and pop-culture fans greatly outnumber literary historians. We asked one user who touted her expertise on Hemingway about the author's lost manuscript. She had never heard of it and then admitted she could name only two of his works.

When we asked the same question at LookSmart Live and Exp.com, we got accurate responses the same day. (At Exp.com, the answer cost us $3, but LookSmart performed just as well for free.) The lesson: No expert site has authorities on everything, so it makes sense to bookmark several sites.

In fact, that's essential advice for Web searching of all sorts: Don't depend on one site, or even two. Bookmark a bunch of them, get to know their strengths and limitations, and use them all. Even then, they won't always know the shortest route to the knowledge you seek: The Web is rife with detours and dead ends. But with the right sites to guide you, you'll spend less time driving aimlessly on the information highway.

Kim Zetter is an associate editor, and Harry McCracken an executive editor, for PC World; Yael Li-Ron is a California-based technology writer. Liz Garone, a California-based freelance writer, also contributed to this article.The Best Search Engines and Directories (chart)Search Superstars

You've got questions. Who's got answers? Start your quest at these sites--our picks as the best options in five key categories of search tools.

Search engine: Google. Sleek, simple, and fun to use, Google promises the Web's most relevant results--and delivers.

Directory: Open Directory Project. Sorry, Yahoo--this site's breadth, descriptions, and clean feel make it the top Web card catalog.

Metasearcher: MetaCrawler. Our favorite all-in-one searcher combines results from top engines such as Google with a friendly, customizable interface.

Specialty directory guide: InvisibleWeb.com. This "search engine of search engines" catalogs 10,000 special-interest search sites on everything from auto racing to zoology.

Expert site: AskMe.com. Call on this site's community of real people for information and advice with a decidedly human touch.

Top Search Tips

1. Be specific. Search engines work better when you use multiple keywords that are unlikely to pop up on irrelevant pages. The string armstrong nasa apollo moon landing will target your search far better than any one of those words alone.

2. Read the instructions. Every site has its own syntax. Entering a proper query can mean the difference between finding what you're after and being deluged by irrelevant links.

3. Go Boolean. Some sites will permit to you enter Boolean-logic queries such as "Moe AND (Curly OR Larry) AND NOT Shemp." If you learn the lingo, you'll be able to enter searches with pinpoint precision.

4. Get advanced. Many engines come with an advanced search page offering additional query types and sorting options. These are especially helpful when a simple search bombards you with links.

5. If at first you don't succeed ... Try another search site. Keep several of them bookmarked, and try out new contenders from time to time.

6. Save pages. Once you find the facts you need, keep track of them. At the very least, bookmark pages; better yet, copy them to your hard disk with SurfSaver, a free utility.

Search Utilities

Besides using search sites on the Web, you can install search utilities on your hard drive. Here are three we recommend.

Copernic 2000 4.5

Though Copernic isn't a search engine per se, it raises online research to new heights by offering impressive organization and collaboration tools. You can save results to custom folders, e-mail results to others, eliminate broken links by clicking on the Validate button, and sort entries by relevance and other criteria. Like a metacrawler, the free version submits your query to dozens of search engines simultaneously. The Plus and Pro versions ($40 and $80, respectively) work with hundreds of general and specialized engines, and let you perform such advanced options as removing banner ads from results pages and (in the Pro version) running unattended, scheduled searches. Get a free version at FileWorld.

GuruNet

This tiny, free utility (available at FileWorld) lets you find references, definitions, and links to any word displayed on your screen. Launch any application (Web browser, word processor, and so on), and Alt-click any word you want to research. Within seconds, a pop-up box will show you information that's pertinent to the word you clicked. For instance, if you click on a company name, you'll get stock information about that business, a corporate address, and a map displayed on the left-hand pane of the small screen. Click on the name of a city, and you'll get dictionary definitions, encyclopedia entries, traffic information, and weather. Clicking on other terms may generate links to financial information (provided by sites such as Motley Fool) or a stock chart (from StockPoint.com). Simply put, GuruNet is a must-have.

X-Portal

Combining the timeliness of online data and the speed of accessing 22 localized reference tools (dictionaries, encyclopedias, almanacs, and an atlas), X-Portal is ideal for the avid researcher. The $60 CD-ROM program needs about 200 megabytes of hard disk space; it integrates into Internet Explorer (version 4.01 or higher). X-Portal queries multiple search engines and its built-in reference works, then sorts results by relevance. The program weeds out many superfluous and not-so-relevant links from the list it presents to you at the end of the search.

--Yael Li-Ron

Subscribe to the Daily Downloads Newsletter