Sounds spooky, huh? Well, this was a question posed to one of our senior folks, we’re not sure if this was just another question by the coffee maker or one at a discussion table subject to serious deliberation. We don’t know that yet, but the interesting bit is that the answer turned out befitting the latter, and with the simplicity intact. It also happens to be up on the author's blog. Read on...

Excerpts from the blog:

"If we try to figure out the situation here, the story goes like this. This questioner was surfing on the internet for wrist watches in the evening and went on a couple of websites to see products, reviews or deals. Then, the same night, he was again surfing the internet for something else and realized that he was being shown ads for wrist watches when he didn’t even intend to. The question here is, how on earth, the internet knows his interests, or what he surfed just few hours back and now displaying similar advertisements when he’s not even opened any page related to it.

So here is my analysis / explanation on that 'How?'

The free websites, which show product reviews and deals, like the one this gentleman visited, are funded by the ads they show. These ads are typically received by an Ad Syndication Network / Provider like Google. Internet advertising, if you don’t know, is a multibillion dollar business, Google itself made close to $28 billion last year by its online advertising division with products like (AdSense, AdWords, DoubleClick). One more thing about this market you should know is that by now it has gone through multiple cycles of consolidation and they are left with only a few big time providers (Google, Yahoo, WPP etc). This simply means you are highly likely to stumble upon multiple websites who use the same ad provider.

Now let me explain the technical aspects of the above scenario.

Let's call the first website which the questioner surfed for Wrist Watches as 'ABC.com' and the second one (surfed the same night) as 'XYZ.com' and the ad company as 'ADCompany.com'.

ABC.com signed up with ADCompany.com as it wanted to make revenue for ads. ADCompany.com's server listed ABC.com as its client and crawled ABC.com. Crawling is a process of reading and parsing a website using automated software without human intervention and a real web browser like what you and me would. The Crawler extracts useful tagging information from pages and URLs. So this is what the ADCompany's database would be like;

Company:	ABC Inc
Website:	www.abc.com
Description:	Men's Accessories Manufacturer
Crawler:	www.abc.com/shop/watches = shopping, wristwatch www.abc.com/shop/sunglasses = shopping, sunglasses
Client ID:	ADCompany allocated a client ID to ABC.com, which it will use for future reference and billing purpose. In this case, it is 'CA-PUB-2777731555298173561'.

The developers of ABC.com went ahead and made changes to their website to show ads from ADCompany.com.

The other website XYZ.com (which was surfed in the night) also had followed the same process however XYZ.com had nothing to do with Wrist Watches maybe it was a pizzeria.

Now here is the story of that evening:

Our questioner googled Wrist Watches and landed up on ABC.com website and his URL said 'www.abc.com/shop/watches'. He was very busy looking at one of the expensive wrist watches on that page and wondering when he would afford it. Meanwhile, he did not notice that ABC.com served him with some ads on the right hand side of the page. Usually, we see these ads and never bother to pay attention and hence are ignored. But the ads have your full attention, if you do a 'View Source' of the page you would have found something like this;

This tiny and innocent looking 'show_ads.js' script did its trick. So, our questioner’s browser not only downloaded it from 'Adcompany.com' website, but it also executed it. In this JavaScript, with one very small HTTP request, which no one noticed, the ADcompany got the following information about our questioner:

ad_client = "ca-pub-2777731555298173561"

Your browser's URL = www.abc.com/shop/watches

Apart from this, the HTTP response put a cookie in the browser which looks something like this

adcompany.com | TrackerID | TRACKER123

ADcompany.com's website also made a database entry somewhere saying

TRACKER123 = Shopping, Wristwatch

(It knows 'Shopping, Wristwatch' because sometime ago it had crawled abc.com/shop/watches and figured out the keywords for it.)

The questioner simply saw a few lame ads about some enlargement products and ignored. Now after a few hours passed, questioner thinks of ordering a pizza online. He grabs his laptop and goes on XYZ.com and starts choosing the pizzas. Now as I said, internet advertising is really a small word and XYZ.com also has Ads coming from AdCompany.com. Same thing happened when he was staring at the Pizza! A small section on the right hand side corner started showing Ads from a script exactly like the one above on ABC.com.

Now because 'show_ads.js' in both cases are on AdCompany.com , both Ad spaces have access to the the cookie set by AdCompany.com. However, this time 'show_ads.js' found a cookie in the browser for it called 'TRACKERID = TRACKER123' which it sent along with other information to Adcompany.com website while fetching the Ads. The Ad Server found TRACKER123 in its database and retrieved the keywords 'Shopping, WristWatch' stored against it. The Ad Server realized the viewer (our questioner) can be targeted for Wrist Watches. It now filters its advertisements that it is about to send to be shown on XYZ.com. The Server delivered only ads that were selling Wrist Watches which made it more likely that the viewer would click on one of them and that would be more revenue for the Ad company.

Here is a somewhat pictorial view of what happened.

Web Advertising Tracking Browser Activity

- Ashish Nangla
Senior Manager: Technology.