For example, say you have a product that you really want to market to people who enjoy social networking. To find out what keywords to use and advertise on, you’ll do a Google Trends search for facebook for, let’s say, last month, June 2008. The results will rank the top regions in the following order:

Region facebook (exported data)

1. Colombia 1.00
2. Croatia 0.98
3. South Africa 0.84
4. Turkey 0.83
5. United Kingdom 0.72
6. Canada 0.65
7. Venezuela 0.585
8. Chile 0.565
9. Serbia and Montenegro 0.555
10. Greece 0.475

But you’ll want to include people who search for all different social networking sites. So you’ll also look at the MySpace search’s list of cities:

Region myspace

1. United States 1.00
2. Australia 0.80
3. United Kingdom 0.435
4. Mexico 0.375
5. Italy 0.23
6. France 0.225
7. Canada 0.155
8. Germany 0.15
9. Spain 0.125

Okay, so do you target Colombia and Croatia, or the US and Australia, or all four? Recall that, to find the list of popular regions, Google first pulls the regions with highest search volume on the keyword and then sorts them by the percent of searches using that keyword, out of all searches from that region. So if Colombia is at the top of the Trends chart, that means that a lot of people searched for facebook from Columbia. But it also means that a lot more of the searches from Columbia are for facebook than they are in regions like the UK and Canada, even though those regions probably have more total searches for facebook.

What does this mean? Let’s assume that about the same percentage of Colombian surfers and UK surfers use Google. Then a higher percent of Colombian surfers Google facebook than do UK surfers. If you have to pay per ad impression, you’ll want your ads on Colombian pages, because for the same number of impressions, your ad will be shown to more people who are searching for facebook and, we assume for the purposes of this post, are interested in social networking. In other words, the percentage of Google searches for facebook out of all Google searches is greater in Colombia than it is in the UK.

But we still have to deal with the facebook/myspace issue. Well, Google Trends supports an OR operator, the pipe |, which we can use to get results for searches that include the keyword facebook or the keyword myspace. We can add as many keywords as we like, so let’s look at the Regions chart for myspace | facebook | linkedin | bebo | flickr | friendster, June 2008.

Region myspace | facebook | linkedin…

1. Ireland 1.00
2. United Kingdom 0.52
3. Croatia 0.45
4. Malaysia 0.45
5. Colombia 0.44
6. Philippines 0.37
7. South Africa 0.36
8. Turkey 0.35
9. Australia 0.33
10. Indonesia 0.33

Ireland! This is because of the inclusion of bebo (bebo.com) which replaced Google as Ireland’s most trafficked site last year (right now it’s the 5th), so bebo combined with the rest of the keywords yields almost twice the percentage that the other regions do.

If you look at the Trends result for all the keywords above except bebo, you get fairly similar rankings, except Ireland’s nowhere to be found and Canada pops up. Generally, as Google says, the order of regions that are very closely ranked can’t be relied upon too much, due to sampling error. So these ten regions, or maybe the top half, are good places to start advertising if you’re targeting social networkers.

In Google Trends, you can’t specify which operators to search for. The keywords are all transformed into lowercase, so the operator OR will look just like the word or. There are, however, operators you can use in your search of Google searches.

Quotation Marks

Compare “space time” and space time. As you would expect, space time ranks higher (5.90 times “space time”), because it’ll count searches in which the user entered the keywords space and time in any order, while finding the Search Volume Index for “space time” only counts searches in which the terms were entered in that order.

Does a Google Trends search for “space time” count only searches done with quotation marks (”space time”), or any search with space immediately preceding time (space time)? My guess is the latter, that the quotation marks you use in Google Trends work like the ones you use in Google Search — they just specify that you’re looking for the two words together, so it doesn’t matter if the original searches you’re counting used quotation marks or not. But there’s no clear documentation on this.

Or (|)

Google Trends gives you an OR operator, the pipe |. If you want to count all searches that included orange or purple, you’d search for orange | purple. The values of the resultant graph should be less than or equal to the values of the orange graph added to those of the purple graph. (They’ll be less than the sum if a lot of people search for both orange and purple in the same Google search, meaning the two overlap.)

Not (-)

There’s also a NOT operator, a minus sign -. To view a graph of searches for stars that don’t relate to Star Wars, search for star -wars or star - wars (the minus sign can have a space after it). The example on the Google Trends About page suggests that there shouldn’t even be a space before the word to be excluded, such as star-wars, but this is clearly wrong. Look at the graph of star, star -wars to see how removing “wars” affects the results. Then, try searching for just star-wars, as Google suggests. It looks for the actual search term star-wars, which few people would type in, so there’s not enough data to return results. (This isn’t the only issue on the About page that I’ll be posting about.)

Grouping (())

Say you want to see how many searches for indiana are truly searches for the state, not for the recent Indiana Jones movie. You’ll do a comparison search for indiana, indiana -(indiana jones). Since indiana jones is grouped, the entire phrase will be excluded. The difference between the two lines shows how much search traffic Indiana Jones generated. While the numbers are about equal before the latest movie, searches for indiana shoot up in 2008, but getting rid of the movie gets rid of that effect.

Similarly, you can group around the |. Google’s example uses (snow boots) | sneakers. Check out the graph when you compare (snow boots) | sneakers, “snow boots” and sneakers. For the combination search, sneakers provide the bulk of the search volume while snow boots provide a large winter spike. Cool, huh?

Hope this helps in your quest to analyze Google searches.

Facebook

For those of you who really don’t know, Facebook is a social networking tool that transports your circle of friends to an online arena. Says their About page, “Facebook is a social utility that connects people with friends and others who work, study and live around them.” One of the most basic feature is that every user has a Wall on his profile page, and his friends come by, write a message on his Wall and read all the other messages on his Wall. When a friend writes specifically on one user’s Wall, all of the user’s other friends can see that post.

In addition, Groups and Events have Walls that anybody in that group or event, or anybody on Facebook depending on the group or event settings, can write on. So it acts as sort of a community forum.

Lexicon

Facebook Lexicon, like Google Trends to which I have already introduced you, shows the frequency of different words or phrases on Wall posts. The key difference here is that Facebook Lexicon looks at posted words, words that other users write in messages, rather than words they use on a search. It’s what they’re talking about, not what they’re looking for.

Lexicon uses a different trick to normalize its data, according to the Help page. It uses absolute numbers and doesn’t worry about any scaling techniques. The graph shows the number of users who posted that term on a given day. It doesn’t count the number of occurrences of a term, just the number of users who use that term — meaning multiple uses of the same term by the same user on the same day are not counted. Says Facebook, “This is to account for the seasonality of Wall posting in general; for example, there are fewer overall posts in the month of December.”

Here’s the graph of, well, facebook. By the way, you have to be logged in to use this tool. Anybody can sign up, though, so there’s no reason not to.

Facebook Lexicon: facebook

There’s no option to see actual values, view the scale of the y-axis or export any data.

Unlike Google Trends, Lexicon is a Flash object. Although Flash leads to compatibility and download speed issues, this means you can easily adjust the time range.

Time Range

This is a fully zoomed out graph. Facebook Lexicon only has data going back to September of 2007.

To zoom in, use the bar at the bottom, dragging the right or left ends to zoom in or out and dragging the scrollbar to… scroll.

Facebook Lexicon: facebook (Nov 4 - Dec 7)

When you zoom in, you can see each individual data point. Hovering over it will tell you the exact date if you’re unsure.

Each data point covers a full 24-hour period, always starting at midnight, Pacific time. Since I’m on the East Coast, I see 3:00AM as the start time. Just remember that the start time is fixed based on Pacific time, and that’s the time frame they use to get the number of users. For example, if a user posts about his favorite banana pudding at 8 PM Saturday, 4 AM Sunday and 11 AM Sunday (Pacific time), he would contribute 1 to the value of the term banana pudding on Saturday and only 1 again on Sunday, sense the value is based on the number of users who post a term on a given day, not the number of times the term was posted. It’s a measure of reach.

Since adjusting the time window gives you individual data points, we could come up with a neat trick to export the data. If you wrote up some code to look at where those points are relative to the horizontal lines and set up a scale that way. Which I’m not going to bother with yet.

Post Terms

Now, Lexicon only tracks words that use the English alphanumeric character set. Any words that include non-English characters are ignored. And the post terms are “cleaned” before being tracked, making them all lowercase and removing apostrophes, so case and punctuation don’t matter in your Lexicon query.

Comparing Post Terms

Like with Google, you can compare up to five terms.

Facebook Lexicon: simpsons, family guy, south park, american dad

Family Guy is more popular overall, but South Park spikes the most after an episodes airs. This implies that generally, people write “family guy” on others’ Walls fairly often, but after an episode you’re more likely to see “Dude, did you catch today’s South Park?” than “Hey man, did you see today’s Family Guy?” (Or maybe it’s, “Dude, are you gonna watch South Park/Family Guy tonight?”)

Privacy

Just like Google Trends, Facebook Lexicon will only show results for a given post term if that term has enough post volume. If it doesn’t exceed the threshold, it won’t show up at all.

Bug

If you zoom in too far, to the point where the time range is only a couple of days, the Flash application — and the entire browser — will freeze up for a while. You can zoom in to the point where you can actually see hours on the x-axis, like “Wed 4am 8am 12pm 4pm 8pm Thu 4am 8am…”. But they would have to recalculate the data to get results in that kind of a time frame, so my guess is they just interpolate a smoothly curved line in between. It’s hart to really say because the app fails so badly when you’re zoomed in that much.

The Bottom Line

Facebook Lexicon is a very interesting tool that approaches web trends from a content-based perspective. It tracks what users are talking about, which may very well turn out to be a better buzz metric than what users are searching for.

Unfortunately, the tool itself is lacking. With absolutely no option to see numerical data, and some quirky bugs, it’s not as useful as it could be.

The post brings up a good question about the Web that all search engines must consider: what defines a unique page? Two different URLs could link to or be copies of the same page. Dynamic applications might have a “next” feature that could continue ad infinitum (Google’s example is a calendar), but does that mean there’s really an infinite number of (useful) unique web pages? And plenty of base URLs with a different query string would show totally different content, meaning the base URL itself is not enough to define uniqueness.

The post itself is worth a read. It’s both historic and interesting.

Today I’m just going to talk about exporting data. This is a new feature in Google Trends and is only available if you sign in with your free Google account.

To export data, expand the “Export this data as a CSV file” node at the bottom of the page, and you’ll see two links: one for fixed scaling and one for relative scaling. (CSV files are lists of “comma-separated values” that load easily into spreadsheet applications like Microsoft Excel.) We’ll start with relative scaling, since that’s what we’re used to.

First, you get a list of search terms with their average Search Volume Indices. Remember, these are all scaled such that the first term has an SVI of 1.00. Here’s what we see for our facebook, myspace comparison search across all years and regions:

facebook facebook (std error) myspace myspace (std error)

1.00 0% 2.25 1%

Then we get the list of data points, the same ones you see in the actual graph. Here’s a sample of some consecutive data points:

Week facebook facebook (std error) myspace myspace (std error)

Apr 6 2008 4.15 2% 4.05 2%
Apr 13 2008 4.1 2% 3.95 2%
Apr 20 2008 4.45 2% 4.05 2%
Apr 27 2008 4.65 2% 4 2%
May 4 2008 4.7 2% 3.9 2%
May 11 2008 4.9 2% 3.85 2%

Yep, you only get one data point per week, starting on Sundays. But I’ll explain how to get around this later. ;-)

There’s really nothing different here. The data points are scaled, relative to facebook having an average SVI of 1.00. Standard error is an estimate of the standard deviation of the error — Google only uses a small sample of all of its searches to calculate the Trends numbers, so they estimate a 2% standard deviation of error, which is small enough for our purposes. If you download the CSV yourself, you’ll notice that back in 2004, the standard error for facebook was listed as “>10%”. This went down to 10% in the fourth quarter of 2005, then down to 5% at the end of 2006 and down to 2% in April of 2007 as more and more people searched for facebook and the size of the sample increased. The standard error of myspace went down to 5% much earlier, in May 2005, which makes sense because it had more search volume at that time.

We can easily take these numbers and make a line graph that looks very similar to Google’s:

Google Trends Exported

The spreadsheet also comes with the Region, City and Language data, which works the same way as the data shown above.

Now, like I said, this is the same old relative scaling we’ve been dealing with all along. What’s fun is that, only when exporting data, Google Trends offers fixed scaling. Whereas relative scaling scales the data to the entire time frame selected — if you look at data from 2007, it’ll be scaled such that the average Search Volume Index across 2007 equals 1.00; if you look at data from October, 2007 it’ll be scaled such that the average SVI across October, 2007 equals 1.00 — fixed scaling scales all data so that the average SVI equals 1.00 at some early, fixed point in time, for every time frame. According to Google, the fixed point in time is “usually January 2004.” And this only works for keywords with enough historical data to even have data in that fixed time frame.

This is useful because it allows us to compare a keyword across different time frames. For instance, say we want to see how facebook was doing in January 2007 and compare that with its search traffic in December 2007. If we just export the relative data, both sets of data will be scaled so that the average SVI for their respective time frames equals 1.00 — in other words, your data sets will both average to the same number. That’s not very useful for comparing across time. Instead, if you export data with a fixed scale, the January 2007 data and the December 2007 data will both have an average SVI scaled by setting the data from January 2004 to 1.00. Thus, we’ll be making a valid comparison.

Of course, you could just export the entire time frame with a relative scale, and comparing January 2007 and December 2007 would be very easy. So why bother exporting the two months separately with a fixed scale?

Well, remember how I told you that the exported data only gave us one data point a week, every Sunday, but that I’d show you a way to get around that? This is it. Exporting data by month gives you one data point per day. So by exporting fixed-scaled data month by month, we could eventually compile a list of search volume data for every single day from January 2004, or whenever our data starts, to the present.

And there you have it. Google Trends now allows you to export search volume data, down to daily points, without having to estimate from the pixel height of the lines on the graph you get online. You’re free to apply any statistical methods you do (or don’t) want to use. This means we can have actual science behind our trend analyses, making our conclusions much more credible.

This was Part Three of my three-part Introduction to Google Trends. Part One. Part Two.

Regions, Cities and Languages

When you look at the results for facebook over the past twelve months, you’ll see three bar graphs at the bottom of the page. These charts are very minimalist:

Google Trends Regions: Keyword facebook, Past Year

South Africa? Yes, back in October 2007, Facebook rose above Google as the number one most trafficked site in South Africa, according to Alexa, although it’s back to number 2 now (with Google.co.za at the top and Google.com as #3). And if you look at the past month graph instead, it’s Columbia and Croatia that are at the top. Intuitively, it makes sense that countries experiencing a Facebook boom would show up higher than countries like the US, even though you might expect that countries like the US use Facebook more (especially since it originated here in the States). The reason boom-countries show up higher, I speculate, is that if you’re going to Facebook, you go to facebook.com — you don’t type “facebook” into Google. But if you’re hearing that your friends, or your children, or your neighbors are all on this “facebook” you keep hearing about, you’re gonna type it into Google (or whatever search engine). With a big enough boom, that’ll generate enough search volume for that region and beat out regions like the US and the UK.

It’s not quite that simple, though. Google normalizes and scales these rankings, too, and for good reason. If the top cities were simply those with the highest number of searches originating there, the most populated cities would always be at the top, even if the search isn’t particularly relevant to the city itself. After all, if only 5% of New York City residents search for “facebook”, that’s still going to be almost half a million searches, which will trump 50% of Virginia Beach residents making the same search, amounting to a mere quarter million searches. But surely if half of a city searches for a certain term, that term’s going to be more relevant for that city than for a city in which only five percent makes that search. So Google normalizes as follows.

First, they get a list of the top cities, simply using the number of searches. So New York might be at the top. But then they take that list and re-order the cities by the percentage of all searches from that city that used that term. So even if New York City is at the top of the first list, and Virginia Beach is at the bottom, the list will be re-ordered such that Virginia Beach is above New York City because a higher proportion of all searches from VA Beach are this specific search.

But VA Beach can only make it into the list in the first place if it’s got enough searches to be returned in the first result set, ordered by the raw number of searches. Otherwise, we’d have the reverse of the problem we discussed above. If, in some rural town with twenty residents, only three have ever used Google, and that was only to search for “facebook”, then, granted, 100% of the searches from that town will have been for “facebook”. But putting those three searches at the top of the list, above NYC, would be ludicrous. Risible. Absurd. Et cetera. “Facebook” may only take up 5% of NYC’s Google searches, but that’s still way more relevant than 100% of a town with only three technologically adept citizens.

The regions are all links — clicking them simply refines the search by region, which can also be done with the dropdown menu at the top-right like we did last time. Once you search within a particular region, you then see a chart of popularity by subregion or state in place of the region chart. And once you drill down to a particular subregion or state, that chart goes bye-bye. But you can still look at cities, if you want to drill down even further.

The city and language charts are similar to the region chart shown above. A user’s city and region are determined by his IP address, which really just allows Google to make educated guesses about the location of a search. Language, on the other hand, is directly determined by which version of the site is used.

Comparing Search Terms

Want to compare the popularity of Facebook and MySpace as Google Search keywords?

Google Trends: Keyword facebook, All Regions, All Years

It’s pretty straightforward. You just get two lines instead of one, and they’re scaled by the same amount. See how facebook is scaled to be 1.00, according to the top bar? Since there are multiple terms, Google only scales the first one so that its average for the time period equals 1.00. So the number you see for myspace, 2.25, means that myspace had an average search volume of 2.25 times the search volume of facebook. In other words, all other terms are scaled relative to the scaling of the first term to equal 1.00. If you look at the graph of just myspace, the scale on the y-axis is different, because now the “myspace” values are the ones scaled to an average of 1.00.

This enables us to look at around September 2007 and say that the term myspace was about twice as popular as facebook. In general, as time increased until about March of 2007, the myspace:facebook ratio grew and grew. Then Facebook really took off worldwide — MySpace’s curve is concave down while Facebook’s is concave up — and the ratio shrunk until that fateful day in March 2008, when searches for Facebook became more popular. Although if you look at the March graph for yourself, you’ll see that Facebook doesn’t usurp MySpace quite as dramatically.

The regions, cities and languages charts also let you compare search terms. Notice how in the graph below, the cities are all ranked by the blue bar — facebook — and the red bars lack any order. Since the data is scaled by facebook, the first term, Google uses that to determine the city rankings, just as it would if only one term were entered. The bars, or search volume, for the cities for the term myspace is shown just for comparison, scaled relative to the first term.

Google Trends Cities: Keywords facebook, myspace; All Regions; All Time

What if we want to see the cities ranked by myspace search popularity? Just use that as your first term. Alternatively, there’s a dropdown box that says “Rank by facebook” — you can quickly change that to “Rank by myspace”, and it would recalculate and redisplay all the graphs, scaled to myspace. Such a graph shows you that, with myspace scaled to an average Search Volume Index of 1.00, facebook’s average SVI is 0.44. This ratio, as you would expect, is approximately equivalent to the old ratio of 2.25:1.00, when we scaled by facebook.

Note that right now, you can only compare up to five terms at the same time.

So there you have it. Now we can have fun comparing different search terms. Here are some examples to get you started:

• windows, apple, linux
• microwave oven, toaster oven, convection oven
• laptop, desktop, tablet
• blueberries, pears, oranges, apricots
• wordpress, movable type

Go back to Part 1 to figure out how the crazy normalizing and scaling stuff works. Part 3, about exporting data, is on the way.

Before getting into a detailed explanation of Google Trends, I’ll show you the graph of searches for the search term facebook, an example we’ll be using throughout this article:

Google Trends: Keyword facebook, All Regions, All Years

This chart, like all on Google Trends for Searches, includes two parts: Search Volume Index and News Reference Volume. The top one displays search volume, and that’s the one we’ll focus on here. In a nutshell, the search volume graph shows the popularity of Google searches for the term facebook over time. Those flags, A-F, give various news headlines that may have contributed to spikes in the news volume graph.

Relative Data

The main problem lies in defining the “popularity” of a search term. According to the About page, all data from Google Trends are normalized. They’re unclear about exactly how they do that, but basically, the data is all relative. The chart above shows the relative popularity of the term facebook, not just the number of searches in which the term is used. So if the popularity of the search for facebook increases, the graph is going to go up. Now, if people stop searching for other terms like myspace and linkedin, then the total number of Google searches is going to decrease and the graph for facebook is going to increase, even if the number of searches for that particular keyword don’t change, because it’s relatively more popular; it’s takes up a higher percentage of all Google searches.

Relative Scale

The scale is also a bit funky. First of all, there is no scale — none that you can see here. If you sign in with your Google account, you can then see some numbers along that y-axis, a feature Google just added.

Google Trends: Keyword facebook, All Regions, All Years

Again, we must refer to the About page to understand the scale. On all the graphs Google Trends produces, the scale is relative. Specifically, it’s relative to the average value for the time period, and that average value is set to 1. In other words, Google Trends takes all the data you see, calls the average height “1″ and then adds a fitting scale to the graph.

Here, for example, the time ranges from 2004 to the present. First, Google scales the data so that the average Search Volume Index from 2004 to the present equals 1.00. Then it simply draws lines at 2.00, 4.00 and 6.00.

Right at the start of 2008, the Search Volume Index was almost 4 times the average SVI for all time. The current value is just above 6.00, so the SVI for facebook has doubled since October 2007, when it was about 3.00. This doesn’t mean that search traffic has doubled, remember; this means that relative search traffic has doubled. In loose language, the term “facebook” takes up about twice the percentage of all Google searches that it had in October 2007. (That’s pretty good for nine months.)

That’s about it for the math behind the data. We’ll get into different techniques for manipulating the data later on, including exporting data and investigating fixed scaling. There are a few other features I want to introduce you to first.

Drilling Down: Regions

Go ahead and play with Google Trends yourself, if you haven’t yet (how could you resist?). At the top-right is a dropdown box for regions. The graphs I’ve shown you have all been worldwide; take a look at the United States-only graph.

Google Trends: Keyword facebook, United States, All Years

The first thing you’ll notice is that search traffic has been increasing much more steadily in the United States. Again, the Search Volume Index is a relative number; relative to all Google searches in the US, searches for facebook have approximately doubled over the past year.

We can also drill down by sub-regions — in the United States, we like to call our sub-regions, well, states. Check out the graph of searches made in West Virginia.

Google Trends: Keyword facebook, United States/West Virginia, All Years

Those random dips to 0 in 2007? That’s just where Google Trends didn’t have enough sample data to report a Search Volume Index. They only report numbers once they cross a certain threshold; says Google, “As an additional measure [to protect privacy], Trends only returns results for terms that receive a significant amount of search traffic.” And Google Trends only includes a sample of their actual searches anyway.

Looking at other regions can help show how Facebook’s popularity has increased, at least among those who search for it in Google. Its SVI has exploded much more recently in the United Kingdom than it did in the US. In the Mexico graph, notice that the scale goes up to 10.00; the relative popularity of the term facebook has seriously increased. In fact it seems to be about 5 times more popular than it was a year ago. Facebook added the Spanish language in February of 2008, where you can definitely see facebook gaining momentum.

Why does it drop, though, at the end of 2007? Many non-seasonal terms tend to do this. Remember that the Search Volume Index is a relative index. When Christmas rolls around, a lot of search terms spike, so even if other keywords get the same number of searches they always get, they’ll drop down lower relatively. Just look at Trends results for gift ideas, digital camera and ipod, to name a few examples.

Drilling Down: Time

Yes, there’s also the option to choose your timeframe. Google doesn’t let you specify your own start and stop dates yet, but you can choose to see Search Volume-Index over a specific year (2007, 2008, etc.), over a specific month (Jan. 2007, Feb. 2008, etc.), over the past 30 days (month-to-date), over the past year (year-to-date) or for all time (which starts in 2004), all from the same dropdown box.

Looking at a graph from the year 2007 scales so that the average Search Volume Index equals 1.00 for that year. This means that you can’t just place the 2007 and 2008 graphs side by side and compare them; the Search Volume Indices themselves will have different meanings.

What you can say, if you notice that the SVI equals 2.00 at a certain point in a graph from 2007, is that at that time the relative Google search traffic was twice the average for that year.

Regions, Cities, and Languages, and Comparing Terms

This post is intended as the first part of a three-part series. Next time I’m going to briefly explain the Regions, Cities and Languages bar graphs that appear below the main chart, and then walk you through comparing multiple search terms, which adds a whole new layer to the information Google Trends provides. In Part Three I’ll go into exporting the actual data, which, until Google releases an API for Trends, is good enough for now.

I hope this post has been informative. Most of the information here is also available on Google Trends’ About page, but this stuff is going to be essential to properly interpreting the conclusions we’ll make later on about trends.

The tubes, of course, abound with valuable information (and plenty of less-than-valuable content). Like in a university library, visitors can search for, access, and digest all the knowledge they want. Like in an international newspaper, subscribers can browse and read articles about all sorts of current events. Like on cable television, viewers can select from and watch millions of channels and videos. Like in a community forum, participants can debate, discuss and dispense their opinions.

But unlike these physical information sources, the tubes give us something even more valuable — meta-tubes. Who knows which books students read in libraries? The best we can do is look at their checkout records, which isn’t usually legal. Who knows which newspaper articles the average homeowner reads? Who knows which videos couch potatoes like best? On the Internet, we know. Usage can be logged, pageviews can be recorded, searches can be aggregated — if we know where to look.

Google Trends offers detailed information on which keywords searchers use on The Google. What can we do with information on who’s looking for what? For starters, candidates can see how successful their campaigns are. Coke can compare itself to Pepsi. Defense attorneys can finally even define community standards in the Miller test for obscenity, because Google knows people are really interested in, what words are searched for and how those searches are made, an invaluable metric in the online world. It’s like Google can read your mind — and everyone else’s.

Google’s offerings are particularly relevant due to the search engine’s massive popularity, reaching almost a third of the global Internet population daily. But there are other tools out there. If Google shows you what people are looking for, Facebook can show you what they’re talking about. Facebook Lexicon, like Google Trends, graphs the popularity of words and phrases. But where Google Trends is query-based, showing data about what people are researching, Facebook Lexicon is content-based, giving insights into what people are talking about. Will it be Obama or McCain? Who’s got the most widely discussed electronics? What’s going to be a box office hit?

The Internet delivers a radically new medium of content. Every action can be tracked, and those actions, properly aggregated and sliced, can give advertisers, politicians and citizens unprecedented amounts of actionable data. But before we can take action, we must first understand this data. Scope must be determined; limitations must be acknowledged. Sources must be mined; observations must be corroborated.

And trends must be analyzed.