philippe::niquille | regular niche market thoughts

Google appears to have reached an enormously high penetration level throughout the world when talking about search engines. That seems enough of a reason to give Google Trends a shot and tickle some of its potential.

The tool allows a usage analysis of search keyword volumes. A Google quote states:

Google Trends aims to provide insights into broad search patterns. Several approximations are used when computing your results. Please keep this in mind when using it.

I was wondering wether the tool was capable of being more than “just a neat feature” and tried several searches and interpretation attempts, as well as a commercial usage scenario. But please always keep in mind, that this is not a science and probably statistically incorrect.. Read on:

Let’s start off with Swiss politics.
In december 2007 our parliament (re)elected the seven members of our Swiss General Council (Bundesrat). The parliament itself consists of members of either one of the stated parties.

As you can see the curves indicate a relatively quiet year until october 2007 which must have been the pre phase of the december elections. Furthermore the right SVP and left SP draw most of the web-attention whereas the rest resides more or less unvisited. The last red peak is most interesting, since one of our Federal Councelors (Christoph Blocher) wasn’t reelected. The Google Search history does show this event as a result of people searching for svp. By the way, I almost got Google to display the correct political colors..

Since these results aren’t overwhelming and solely reprint a past event, let’s try comparing world topics such as HIV, cancer and tsunami.

The keyword tsunami did draw attention big time at the end of 2004. As you may remember a huge tsunami did in fact hit several coasts in the indian ocean with a 30 metre wave. But let’s have a look at the more interesting blue curve: Peaks repeatedly appear shortly before Christmas every year. Do people start to think about donating money to HIV charities mostly around Christmas times? Also have a peek at the regional distribution and look at the bars behind South Africa. Isn’t Africa known for a large penetration of the HIV virus? If you then click on South Africa, the peaks disappear and the curves becomes flatter (but still more volume than the cancer keyword). If you do the same narrowing down with the US you may observe that the keyword cancer shows way more volume and the HIV peaks persist.

But wait, the service has more to offer than just a basis for historical analysis and interpretation. What about keyword optimization for AdWords campaigns, slogans, product names, you name it. Picture yourself advertising for cell phone accessories. Which keyword would you choose in Switzerland and Germany?

Handy (which is common around here) for all of them? On first sight one would agree, but focus on the regional distribution. The local bar graphs of Switzerland (Cities) show a density of the keyword natel in the french speaking part of Switzerland. Although the keyword is known in the german speaking part of Switzerland as well, Germany isn’t familiar with the term at all. Shockingly though the offical term “Mobiltelefon” isn’t used at all. Welcome common speech. And last, the english term cell phone isn’t spread in europe at all, including (!) the UK.

Last but not least: What programming language seems to be en vogue according to Google?

Strangely PHP is steadily loosing attention whereas ruby is slowly growing. Why on earth is the Czech Republic #1? Do the offer extensive PHP outsourcing?
But watch it: in the news section one sees a link to a story about a python snake. Aren’t those results (due to poor filtering) completely destroying this analysis?

Furthermore you could start stepping on really thin ice and compare brand names such as coca cola and pepsi to each other and take it as a (weak) measure for brand valuation..

But, why it may be stupid to interpret too much.. (downside)

• People don’t just lookup the topic’s title (such as apple) and instead enter words such as iphone or similar combinations. You are never sure what your search base consists of. You could be solving this issue by interacting with users and asking them specific questions about their search terms. Did you mean x? Does it include y? Wouldn’t this be interesting for search applications in general despite Google users being used to typing and immediately getting responses..
• Although Google does a world-wide lookup by default, the tool is not (yet) capable of breaking language barriers by combining translated words. Its use appears to be limited to certain language areas.
• The graphs don’t show any axis values. Although trends is only showing relative movements, the size of the search base would be of interest (especially from a statistical point of view). So basically you’ll never know what those approximations are based on.
• Sorting results by region is nice, but, how many people falsificate these results by using either VPN’s, proxies (ex. in China) and providers with few backbone nodes? Did you note: China never appeared in the top 10 in my searches..

Some links to interesting Google Trends lookups:

• Hartz IV in Germany
• Searches for french, current and past US elections