Please note: This site's design is only visible in a graphical browser that supports Web standards, but its content is accessible to any browser or Internet device. To see this site as it was designed please upgrade to a Web standards compliant browser.
 
Signal vs. Noise

Our book:
Defensive Design for the Web: How To Improve Error Messages, Help, Forms, and Other Crisis Points
Available Now ($16.99)

Most Popular (last 15 days)
Looking for old posts?
37signals Mailing List

Subscribe to our free newsletter and receive updates on 37signals' latest projects, research, announcements, and more (about one email per month).

37signals Services
Syndicate
XML version (full posts)
Get Firefox!

37BetterGoogle

20 Mar 2003 by Scott Upton

37BetterGoogle

Google is the search expert. Yet as good as their search engine is, the results are only as good as the search terms you enter. How can you know if “african coffee trade” is a better search than “africa’s coffee trading” when you’re doing research? You can’t possibly know, but Google can.

Our idea for 37BetterGoogle was to go beyond Google’s standard search and offer alternatives based on the words you originally entered. By searching for different forms of the same words and comparing those results to your intial query, Google does the legwork for you, helping you see results you may never have discovered the first time around.

So have a look around and be sure to click the links to the alternative searches. Our hope is that Google will see value in adding the search options described above to its basic search functionality.

55 comments so far (Post a Comment)

20 Mar 2003 | Toby said...

Brilliant. Youse gots to stop given these gems away.

20 Mar 2003 | JF said...

And don't miss our other 37Better Projects. We'll be adding 37BetterGoogle to that page eventually.

20 Mar 2003 | Tibloto said...

Great idea. Almost as good as Google's "Did you mean:" feature....almost. ;)

20 Mar 2003 | pb said...

Is that worth quintupling search volume? "Did you mean" is much, much better because it has a substantially higher success rate. For me it's probabloy 40-60%. 37BetterGoogle's would be closer to 10-25%.

How about a 37BetterYodlee?

20 Mar 2003 | JF said...

We see 37BetterGoogle as an optional feature Google could offer. It wouldn't be the default.

20 Mar 2003 | Paperhead said...

1. The logic of this would seem to be that you can make a decision about which terms you should have searched for just by looking at how many results are returned for each term. That's not true.

2. How many of these things would you have, the original search term could produce dozens of variations, you can't waste time generating all of them. And if you can't waste time generating all of them then there's a good chance that you won't generate the most helpful one, so what's the point.

3. You people clearly have fast connections. Most people are still on dial-up. People like to use Google because it's fast, even on a dial-up. You just slowed it down by making it do six searches instead of the one I wanted it to do. How does that help me?

that's the first three off the top of my head.

20 Mar 2003 | Mark Hurst said...

i'd have to disagree on this one, guys. of course i have no data to back this up - this is just my $0.02 - but i really like seeing the results start right at the top of the page.

i also never liked seeing how many results a search turned up. i don't want 3,000 results, i want ONE. MINE.

maybe if your feature was at the bottom of the page... then if i got to the bottom and was still unsatisfied, i could click into the alternate links.

20 Mar 2003 | Steven Garrity said...

It's unfortunately that Google doesn't offer valid XHTML/CSS. They're layout/design is already quite simple and would be relatively easy to move to valid markup.

20 Mar 2003 | scottmt said...

Great idea. I was thinking about how google is limited by your creativity in searching sometimes.

The only real way to search for a hard to find piece of information is to think of a phrase that would appear on the page and try it out.

If Google could get even deeper than this and offer me phrases loosely based on what I was searching for, maybe categorized .. as a tool/feature of course, not standard search.

20 Mar 2003 | fajalar said...

I was thinking about Google today. I don't like that the buttons on the initial search page don't work when there is nothing in the text box. I click either one and go to a page with nothing (of value) on it.

I know there are good and bad points to disabling buttons, but I don't think it is much help that the search button just takes you to a blank search results page.

The reason I got to thinking of this was I was looking at the main page and just clicked the "I'm feeling lucky" button. I don't pay much attention to it, and forgot that it didn't do anything without a search term. But, I think it should take you to a random site on the Web even without a search term.

20 Mar 2003 | JF said...

i really like seeing the results start right at the top of the page...i also never liked seeing how many results a search turned up. i don't want 3,000 results, i want ONE. MINE.

I agree, Mark -- I like the results at the top of the page too. That's why we provided a link to "Turn this feature off." The feature would also be optional -- it would only appear if you wanted it to appear (it would start as a link and then expand into what we've shown).

Regarding your desire to see that ONE result that best matches your request... I couldn't agree more. That's the point of this feature. In order to get that ONE result that satisfies you, you have to know exactly what the best Google search is. You need to know the exact order, the exact tense of the combination of words you are looking for. This feature helps you find exactly what you are looking for by suggesting slight modifications to your original search.

20 Mar 2003 | fajalar said...

Also, here's a paper (link is also posted on my Web site) on cognitive walkthroughs for the Web. In it, they talk about the use of "cognitive models of user search behavior." It's a short papr (only 8 pages) and worth a read. They talk about understanding the "corpus" of words that a given sample of users have (usually based on age) to match the most likely ways that users will go about looking for information. They have a tool on their site that you can use as well.

(I'm in a class for 3 days looking at much of the HCI literature that has been published in the last year, so my brain actually gets to work!)

21 Mar 2003 | slvrfrg said...

This reminds me a lot of the Network Solution's domain search. I would never register a domain through them, but whenever I am searching for a domain, it's the first place I go. Check it out - a great tool. Network Solutions

21 Mar 2003 | alisha said...

would it then produce different results than if I had just continued to pages 3, 4, 5...? Maybe google does that already by searching for every possible word combo(?).
---
"i also never liked seeing how many results a search turned up."
---
why? I like to know what I´m working with.
---
"i don't want 3,000 results, i want ONE. MINE."
---
would be nice but it´s unrealistic.

21 Mar 2003 | p8 said...

I use Google's cache a lot to browse the different search words. Google highlights the search words with different colors.
I think it could be very helpful, if you could click on a search
word to go to the anchor of the next search word on the page. Especialy with long texts.

For example:
These search terms have been highlighted: <a
href="#google1">google</a>

This is an example for the word
<a name="google1" id="#google2">google</a>. When clicking on ths link the page will jump to the next examples of the word <a id="google2"
href="#google3">google</a>.

21 Mar 2003 | p8 said...

"i don't want 3,000 results, i want ONE. MINE."

But sometimes I find myself not even knowing what I'm looking for precisely and using google to help me find better search criteria. It's an iterative process.

I don't want google to tell me: "This is what I think you were looking for. Even if you wanted to know something else you should be looking at this."

BTW Has anyone ever looked passed the first 100 results?

21 Mar 2003 | Tony said...

I like the idea. But I also depend upon Googles "Did You Mean" option. (I'm a horrible speller...that helps).

I don't see real estate for both options, as I want to see my results quickly.

I think I'd like the ability to automatically roll these results up into my search. An "Expanded" search, a "Roll-up" search, a "Total" search...(whatever you want to call it.) Be able to switch it off and on easily...a visible toggle. You could easily show the variations on the words I think on the same page.

I just think that if you have both variations and Did You Mean, it gets confusing.

21 Mar 2003 | Ed said...

I manage the search engines at a large corporation, and I can tell you that I get a LOT of negative feedback from folks complaining about the number of results that are returned.

The #1 complaint is often "Too many results! For example, 5,462 results for the word 'policy'?!?" How do you expect me to go through that many results?!?"

In my next revision I plan on removing the search stats, reserving them for advanced searches only.

21 Mar 2003 | alisha said...

The #1 complaint is often "Too many results! For example, 5,462 results for the word 'policy'?!?" How do you expect me to go through that many results?!?"
---
are they so daft that they think there are only 20 websites with the word "policy" in it´s contents out there? then they should narrow down their search with more words. jeez.

21 Mar 2003 | fajalar said...

are they so daft that they think there are only 20 websites with the word "policy" in it´s contents out there? then they should narrow down their search with more words. jeez.

So at what point do we make the diliniation between making the system work for the user, and allowing the user to grow their ability to define what they are looking for before they click the search button?

I actually think there is a lot of value in raising the skills and knowledge of the user through the interface. I agree that users should come to a better understanding on how to search for information. But, we certainly should display all the information in a way that is not overwhelming, or defeating, or makes the user feel "dumb" about their search criteria.

21 Mar 2003 | Charlie Park said...

Regarding Steven's point, about Google and valid XHTML/CSS, I totally agree. It wouldn't be too hard for Google to set its site up using standardized markup. I was kind of hoping the 37s code would be valid. Sadly, it's not even close.

21 Mar 2003 | JF said...

Charlie, our code isn't valid because we used Google's code. This wasn't an excercise in rewriting their code, it was an excercise in coming up with a new idea to make a Google search more productive and useful. Surely we could have spend a bunch of extra hours to make sure the code validated, but remember, we did this for free and we wanted to spend the time that we had on the idea, not the code.

21 Mar 2003 | Ben said...

"Google highlights the search words with different colors.
I think it could be very helpful, if you could click on a search
word to go to the anchor of the next search word on the page. Especialy with long texts."

Doesn't this feature already exist in the Google toolbar? There is a "search site" button that functions just like the "Find" function in your browser, based on the keywords you entered for your web search.

As far as the ideas for functionality goes, it's an interesting simple idea that seems like a natural evolution of their "Did you mean" spell checker. Perhaps instead of having it be toggled, it would just be a linked option after the number of results are given so that it only has to run the multiplied search when specifically asked to. (Something like "See variations of your search" or "See similar searches".)

Does anyone actually use the "I'm feeling lucky" button? For me, the novelty wore off after using it once.

21 Mar 2003 | Ben said...

Since we're talking about Google... is there an official (or non-official, who cares) Google suggestions repository? I know Google has a suggestions e-mail, but threads like these resonate so well with people, and there are a million other ideas out there, so it would be nice to be able to see them all and talk about them and have a communal place to suggest new ideas.

And speaking of ideas... the one I always thought would be useful would be to combine their searching with their translating tool. Like if you're looking for "african coffee trade", why can't it also look for each of those specific words in other languages and find me those pages (which Google can already translate for me)? I would assume some site in Africa may be able to tell me more about the african coffee trade than many others...

21 Mar 2003 | fajalar said...

Here's another searching study. It is a survey of recent literature, as well as some conclusions on guidelines.

Not that this one isn't valuable of course. ;)

21 Mar 2003 | fajalar said...

SearchEngineWatch.com

Heh, now we are talking about your search report in class.

21 Mar 2003 | JF said...

Which class are you taking?

BTW: Gordon Whyte, About.com's E-Commerce expert, just reviewed our E-Commerce Search Report. In summary:

"Overall this is an impressive piece of work. The testing methodology is explained well, the results are clear and the conclusions are persuasive. This report should be required reading for anyone interested in Web usability or involved in the design of E-Commerce sites."

Thanks!

21 Mar 2003 | fajalar said...

(I'm in a class for 3 days looking at much of the HCI literature that has been published in the last year, so my brain actually gets to work!)

From my post toward the top. It's a class at my company. We are reviewing about 100 papers that have been published in the last year, and while I was writing the last post, I turned the page in my class binder and lo and behold... 37Signals.

It was a favorable review by the person giving the class.

21 Mar 2003 | Ed said...

Alisha--
My metrics data shows that the average number of terms entered for a search is 1.41. In other words, most folks enter a single keyword and expect accurate results (from a corpus of 500,000 files) on the first page.

It's that first page that counts (actually, the first 10-12 results), and that's where Google really shines.

Fajalar--
When more than 1,000 results are displayed, we display a message that they should consider including more, specific terms in the search.

Users seem to have a fundamental problem with the behavior of search. Furthermore, most of my experiences with search engine vendors has not been good. You often need to do an awful lot of modifications to the out-of-the-box systems to get a decent user experience (such as the best practices described on this site). The vendors really don't get it.

21 Mar 2003 | fajalar said...

Users seem to have a fundamental problem with the behavior of search. Furthermore, most of my experiences with search engine vendors has not been good. You often need to do an awful lot of modifications to the out-of-the-box systems to get a decent user experience (such as the best practices described on this site). The vendors really don't get it.

We have been looking at a lot of studies on searching. One of the reasons that these out-of-the-box options are not able to be used out-of-the-box, is because the specific sites have specific domain and product/service information. Some of that needs to be set by the user when the context of the search parameters (the site) is set. In other words, I wouldn't look for a CD on REI.com (even though they may sell CDs).

So those search engines are set up for a general lexicon, but need to be tailored to the specific lexicon of the site. Which most sites do not seem to bother to do.

21 Mar 2003 | dmr said...

FYI, Google toolbar is for Windows only. =[

While I appreciate the efforts to narrow results, couldn't this kind of thing get out of hand with prefixes/suffixes for multiple words? Offering too many alternatives surely isn't a good idea either.

21 Mar 2003 | gianni said...

havent read the other comments but did you guys know you can make applications from google. You can extend google's engine to do exactly what you are mentioning above...

http://www.google.com/apis/

22 Mar 2003 | Eric Scheid said...

Suggesting alternative words for the thing I'm looking for has limited value for me -- unless I get extremely few results I usually review the likely looking results, and from that small excursion I typically pick up a few extra search terms, which I then go back and reiterate.

My top tip is: if you're searching for academic or similar in depth commentary on a subject, include the word "bibliography" in your search. "African coffee trading bibliography" will find a much smaller set of results, but you're pretty well assured there'll be meat on those bones.

Of course, I'm not directly interested in reading bibliographies ... I'm interested in reading documents that are substantive enough to have bibliographies attached.

For what it's worth though: bibliographies do contain a very juicy list of extra keywords to try, so it's not a wasted read.

22 Mar 2003 | monk said...

For the guy that said that the Google Toolbar was Windows only, I know for a fact there is a Mozilla extension that ports the toolbar over to Mozilla, which means it should also be available for Linux and Mac OSX.

Googlebar

23 Mar 2003 | Joshua Kaufman said...

If this could be done without increasing the response time and was turned off by default it would be an okay feature. I suspect a lot of users won't see the immediate value of alternate searches and instead just want to see their results start at the top.

My top suggestion for a truly better Google is to apply more of the Parc's search research to search results. Google already highlights search terms in cached pages, but I want that taken one step further with the live pages. I've got the bandwidth and I'm willing to wait a little bit longer for more a more useful interface to browse the results.

24 Mar 2003 | Jonathan said...

Don't know. How do you know what (changed) query gives you the best result (the answer to the question you're looking for)?
Good idea. But easy on the features I'd say. (cfr. Word)

24 Mar 2003 | Brandie said...

"How do you know what (changed) query gives you the best result (the answer to the question you're looking for)?"

When you find the best result then you know. The point 37signals is trying to make, I think, is that Google should provide some slightly different alternatives to help you find what you're looking for. Without these suggestions you don't even have a chance if nothing shows up that's useful on your original search. I mean, if you search for something, and the results aren't useful, then what? People hit a dead end. This idea at least gives them somewhere to turn next.

27 Mar 2003 | p8 said...

I love how smart Google is getting. The AI is amazing.
http://www.google.com/search?q=french+military+victories&hl=en

31 Mar 2003 | jupiter said...

What about a 37BetterHotmail?

04 Apr 2003 | james said...

fajalar:

if they are too much relevant documents found, you shouldn't reduce the founded documents to 20, but explain the user (how) to narrow down their search with more words.

05 Apr 2003 | Danielle said...

Google could offer the same option for those of us who really want it, and avoid consuming bandwidth for those that don't, by supporting truncation.

For example, I could type in "africa* coffee trad*" and get results for all 5 of the alternative searches, plus the original, in one.

Much as I appreciate Google, this is one search feature they lack that the experienced web searcher really misses. If AltaVista can do it, why not Google?

11 Apr 2003 | Aaron Swartz said...

I suspect Google may not do this because a decent page on "african coffee trade" (their example search) would probably also include the words africa, africa's, and trading (their corrections). If this is true, the high quality pages would come up no matter which term you used and the pages that don't come up are probably about something else (and only mention coffee trade casually).

11 Apr 2003 | Alex Piner said...

Google API is a good start as Gianna said.
Maybe Tara from Research Buzz can build it?

12 Apr 2003 | Tom said...

The other thing Google should do is remove the difference between numeric and written numbres, so searching for "thirty-seven signals" would find this site.

14 Apr 2003 | kami said...

That's great, but I still can't sort by date.

22 Apr 2003 | José de Menezes said...

Hi, someone knows how to do a toolbar for www.detetive.net like toolbar.google.com??

24 Apr 2003 | Pablo Impallari said...

Very god idea...

Also, You can just create a link called:
"Try this variations of your search" or something... instead of pointing all the variations whit all the result.

So, when de people click that link, then you show all the opcional searches...

Wath do you think of this?

Sory about my little english.... i am from Argentina.
Bye, bye...

12 May 2003 | Anonymous Coward said...

How pretentious are you guys?

13 May 2003 | wne4 said...

As Paperhead (March 20) alluded, the "frequency" vs. "relevance" problem rears its head in this sort of alternative analysis. The search results may include precisely the link one seeks while simultaneously burying it within a list of thousands of links. Trying to find the RELEVANT link in this welter is like looking for a particular sign, in New York City, that you know exists, but that you nevertheless cannot find.

However, unless you've already provided related info about your preferences previously, neither Google nor any other search engine or person can, in principle, divine your intent. Indeed, if this search represents an inquiry into a heretofore unexplored field for you, data on previous searches, if used to enhance the effectiveness of your search, may in accurately skew the results away from the result/s you seek.

What about a "Google hack" that automatically entered each search term into the Google Sets generator http://labs.google.com/sets and asked the searcher if s/he'd prefer any of the alternative words? (Of course, any searcher can do this as a separate operation. The selected words are then entered into a new "search within results".)

No doubt we'd all like to have Google or any other search engine read our minds and give us what we want. Sometimes, as any searcher knows, we're not exactly sure what we want, and the very serendipity of the search--especially the serendipity of following links--brings us unexpected treasures. But, on a more prosaic level, most often we're looking for relevant information on a specific subject, and we want it as fast as possible, without delay or misdirection.

I have found that one or two quick "search within results" with some sort of qualifying additional word can readily get me where I want to go. For example, after getting the 750 results for "africa's coffee trade", I go to "search within results" and enter whatever the most relevant additional word or two to my search, e.g., "NOT Congo", or "trade AND organization", etc. Doing so is likely a way of bypassing the initially raised question of whether word order (or precise wording--"African" vs. "africa's") in the initial search significantly effects results. Using Google Sets takes a bit more time, but are the results worthwhile? A more extensive trial and error comparison of might be telling.

One last comment... I certainly agree with "fajalar" (March 21) :
I actually think there is a lot of value in raising the skills and knowledge of the user through the interface. I agree that users should come to a better understanding on how to search for information. But, we certainly should display all the information in a way that is not overwhelming, or defeating, or makes the user feel "dumb" about their search criteria.

17 May 2003 | radman53 said...

Google is fine just the way it is- If it's not broke- DON'T FIX IT!

23 May 2003 | Analogik said...

Yeah. Great Idea. I am sure, however, that there are many more ways to improve the search on Google. Some things are up to the user but some of them depend of Google features alone... there is a LOT they could improve and MANY features yet to be implemented...

I'm sure you all have some search solutions yourselves!

29 Jun 2003 | David S said...

Not sure how great it would be, but I'd like to comment on what Paperhead said:

You people clearly have fast connections. Most people are still on dial-up. People like to use Google because it's fast, even on a dial-up. You just slowed it down by making it do six searches instead of the one I wanted it to do. How does that help me?

You only do the extra searches by clicking on one of the alternate searches. There is no extra cost to you by having this feature, sans the 4 or 5 links to alternate searches, which is too small to measure.

30 Jun 2003 | Chris McLaren said...

I agree with other who've pointed out functionality and usability problems with this suggestion. You can't be right all the time - nevertheless I applaud the effort and the other "37 better" ideas. However, the quality of this one doesn't match up to the others IMHO.

02 Jul 2003 | Isaac Pinnock said...

How can you know if “african coffee trade” is a better search than “africa’s coffee trading” when you’re doing research?

Simply put, you don't need to know - Google automatically takes care of this issue for you by stemming their index.

Stemming works by stripping word endings from index terms to leave the core, or stem of a word, which will be common to all forms of the word. For example, if a user searches for "analyze", it follows that they may also want documents which contain "analysis", "analyzing", "analyzer" and "analyzed". These various forms are returned as they all have the same core of "analy".

This technique saves the user from having to make multiple searches and indeed from knowing which search term is the most relevant (unfortunately rendering 37BetterGoogle redundant).

03 Dec 2003 | ortho evera said...

I agree ... Google is great!

Comments on this post are closed

 
Back to Top ^