compare-stuff.com FAQ

compare-stuff.com frequently asked questions

What's going on?

Have you seen the help page yet?

Can I see all the random tips before I switch them off?

What can I compare?

To give you an idea, try the examples which you see when you reload the main page - but please bear in mind that companies and trademarks have generally been avoided to avoid creating the impression that compare-stuff.com is out to get them (but we are, of course). If you need more inspiration, check out these suggestions.

I've checked the results pages like you said, and found that the search terms are miles apart on the page. What's wrong?

Yes, we can't stop this happening at the moment. In general we hope that the sheer number of web pages will "average out" problems like this. In general this seems to hold true (try some comparisons where you already know the answer). When you don't have "large numbers of web pages" then take extra care.

The AltaVista search engines offer a NEAR operator (distance = 10 words). If you select the "AltaVista (near)" search engine from the pop up menu in the advanced options, we'll combine all your search terms with NEAR. You can try this, it should work well when there are enough web pages, but we haven't had much experience with it yet.

I checked the results and found that the web pages were totally unrelated to the comparison I was making. Do you really expect this to work?

Well it seems to work quite well with many types of comparison (see previous question), but of course not all. We have to rely on large numbers of web pages and a carefully constructed, unambiguous set of query terms. And remember, it's just a bit of fun...

I got completely the opposite results to what I expected. Why?

Consider this example, comparing CRT and LCD displays with the comparison term "eye strain". You might expect web pages to say "CRT monitors cause eye strain". However, you might find that, in fact, people write "I have no more eye strain problems with my new LCD display." or "Many people choose LCD displays to avoid eye strain" or "I had bad eye strain so I bought an LCD display", and so on. This demonstrates that compare-stuff.com simply measures word co-occurrences. Computers are not able to read text and understand its meaning. Well, not yet anyway. So for now, if we want to measure and compare things from the web, we'll have to form careful queries, check the results, and treat them with caution.

If I don't get a warning about statistical significance, does that mean my results are significant?

No, sorry. The statistical test we perform is only really suited towards detecting uncertainty when the numbers of web pages are small. You must decide for yourself what the results mean. Try some different query terms, do a few controls, try another search engine. Then decide. But remember, we don't guarantee compare-stuff.com results in any way (see the disclaimer).

What is automatic correction?

We have found that many of the search engines give document totals that consistently favour one particular "thing" over another during a compare-stuff.com comparison. You can see this by turning it off (in the advanced options) and comparing "California" with "Ohio" you will see that California "wins" for almost any comparison term you try, even meaningless words like "because", "Susan" and "hello". If you don't see this, switch to another search engine.

We use exactly this information (percentages for meaningless words) to correct the percentages shown in the chart. The correction is "improved" a certain number of times as you resubmit the form, by including more meaningless words.

What are the "meaningless" words you use?

We will have a web page showing these soon. We currently have control words for English, Swedish, French, Spanish, Italian, Portuguese and German. More will be added as soon as possible. Numbers are used when your language is not supported and for "Any Language".

What is manual correction?

It's just like automatic correction, except that you choose the control word. If control words are not available for your language (see above) you can enter one here. You can also check for the presence of your comparison term "against a background" of the control term if you know what you are doing.

What do the search engines think of all this?

We keep the load on the search engines as low as possible by keeping a cache of all the recent document totals. Also, requests are sent in series (not in parallel). We also provide links to the search engine results (remember to check them!) - so that extra human traffic should go their way. Further load control measures will be taken as necessary.

We have sought permission from each search engine to make use of their services - in some cases no reply has been forthcoming. Search engine owners should email us with any queries.

Why isn't search engine "X" available?

Email us if your favourite search engine is not available, and we will try to remedy the situation. But, please check that the search engine has the following features before emailing:

returns total number of documents found
pages contain all terms (boolean AND)
terms can be excluded (boolean NOT)
language specific search

For biomedical science comparisons, there is a special PubMed version which is not accessible from the main page.

Why is it sometimes very fast, and sometimes not?

We cache all the recent document totals, so for example when you clone a page or email your friends the link, the results come direct from the cache very quickly. Response times in general depend on search engine load and network traffic.

How do I permanently set my language preference?

You can bookmark compare-stuff.com using a URL like this http://compare-stuff.com/?Language=French

What's "AltaVista (near)"?

It's just AltaVista but we use the NEAR operator to combine the search terms. That means that the words have to be found within 10 words of each other on the page. It should work well, although bear in mind that the numbers of web pages will be lower. It hasn't been tested as much as the other search engine options; if there are any problems, please let us know.

Your questions...

If the answer to your question is not here, please send an email and we'll aim to send you a personal answer within a week or two.