Introduction to first party cookies
It seems that many people who work in online advertising don’t fully understand how cookies work, so I thought I’d do a post on it since cookies are such an integral part of online advertising.
Hopefully the chart below will help you follow along.

When thinking about cookies, it’s best to think of all the communication as occuring between the browser and the web server. In the depiction above, the browser is the program on a computer that the user uses to surf the web. The web server is a separate machine, which hosts web pages.
When a user enters in a website address, the browser makes a request to the web server that is hosting that website address. When making a request, the browser sends a bunch of information, such as its IP address, the type of browser as well as the requesting domain name and path. The web server will then typically respond by sending the browser some html code, which defines what the user will see in her browser, and maybe a cookie.
In essence, all a cookie is is some text, which the browser saves as a text file. It often consists of a random text string generated by the web server along with an expiration date. If the browser makes another request to the web server (such as requesting another page), the cookie will get sent back to the web server. This happens only when the requested page is on the same domain (or subdomain) as the original request, so if you have a cookie from google.com stored in your browser, that cookie will get sent only when a URL residing on google.com is requested. The web server can then see that the cookie sent with the second request contains the same text string as the response to the original request that the web server sent and will know that the two requests were made by the same browser.
The common applications of first party cookies are site login systems (the user/browser logs in to a site and then on subsequent requests, the web server knows that the requests are coming from the browser that first logged in) and some web analytics programs, which allow for tracking a single browser across a site.
Next up, I’ll discuss third party cookies.
Atlas' Engagement Mapping is out of beta
A year ago, there was a post by Bob Heyman on Media Post Search Insider that I found very interesting. He talks about how search receives too much credit in ad campaigns where the "last click" model is typically used. He points to a case study that showed that when a portion of display advertising yielding the highest CPA had been discontinued, overall conversions dropped along with conversions from search, a sign that the seemingly underperforming media was having a significant impact on brand awareness without receiving proper credit.
Eyeblaster recently launched a tool that allows advertisers to see which ads a user was exposed to prior to converting through paid search. From Marketingvox:
Eyeblaster launched Channel Connect for Search, which enables marketers to track users that are exposed to, or that interact with, display ads — even if they don't immediately convert.
A couple of days ago, Atlas announced that Engagement Mapping was out of beta and available to all existing clients for use. 'Engagement Mapping,' also known as multiple attribution gives credit to every online advertiser touch point before a user converts. Jason Heller wrote a great post about Engagement Mapping and does a better job explaining it than I could.
Atlas is able to track these touch points by placing cookies on user's browsers. To do this, all ads that are being tracked will need an Atlas redirect, which is the point where Atlas loads the cookie. So if a different agency handles search from display advertising, redirects would have to be implemented by the agency handling search.
I look forward to trying this out in the near future.
Want to see who's serving and tracking an ad? Enter Live HTTP Headers
The Firefox plugin, Live HTTP Headers is a great tool if you want to see if certain links are getting redirected the way you want them to be, but you might also be able to see which ad network is serving a banner or text ad and also what ad server the advertiser is using.
Online ads generally redirect to another URL upon click. What users don't see is that often when they click on an ad, they are successively getting redirected to multiple URLs - that is, one URL redirects to another, which redirects to another with the host server of each URL tracking the click. What gets sent to the browser when a URL is requested is called a "header," which sometimes is followed by html code.
The header contains things like the type of server the URL is hosted on, a cookie perhaps as well as something called a status code. When a web page gets loaded, that is a '200 OK' code. Everyone has the experience of trying to load a URL but the server is unable to find that page - that is a '404 Not Found' code. When a URL redirects, it uses either a '301 Moved permanently' or a '302 Found' code. These basically tell the browser to go to another, specified URL. The server will log the request in the process. So I could create a redirecting URL on my server, perhaps agencyvendors.com followed by /redirect/123 and then point that at another URL. Then, I'd be able to see when someone visited my redirecting URL, which I could then use to monitor clicks.
Live HTTP Headers displays the requests that your browser sends to a server and the headers that get sent back. If a particular banner ad or link redirects 5 times before resolving to the landing page, you will see all those redirects. I just visited msnbc.com and found an ad for Infinity, so I clicked on it with Live HTTP Headers turned on. The first redirect is a path on cdn1.eyewonder.com, which redirects to an eyewonderlabs.com URL, which then redirects to an ad.doubleclick.net URL before resolving to infinity.com. I know that eyewonder.com and doubleclick.net both belong to ad serving companies, so I know that the advertiser is using these ad servers to serve and track the ads. Further, I know that if two ad servers are used, they serve different purposes. I know that eyewonder is known for rich media ad serving, so I can tell that it is likely that the advertiser is using doubleclick for their general ad serving needs and eyewonder for rich media. I can then use this same technique to see if an ad network (if any) is serving the ad. It is also useful to see what cookies (if any) are getting served with each requested URL.
Daily Links - July 14, 2008
Weather.com Has Secret Arsenal: Zip Code Data (MarketingVox)
Predictive Modeling Across Channels (MediaPost)
Veoh Debuts Behavioral Ad Targeting System for Online Video (MarketingVox)
Finance Info Sites' Audiences Growing, Yahoo Finance Leads (MarketingVox)
Fledgling Ad Servers Start to Take on the Three Major Players (AdAge)
Trying out Google Ad Planner
I decided to take a look at Google Ad Planner, Google’s new research tool designed for online media planners that seemingly relies on Google Analytics and the Google Toolbar for its site profile data.
The first thing I did after logging in to my new account was to create a “media plan.” Note that this “media plan,” which Google also refers to in their introductory post is not a real media plan because it doesn’t contain information such as flighting, costs, impressions or clicks, placements, ad sizes, creative type or targeting. It’s more like a Nielsen @plan pull (which I haven’t used in years) that shows sites with a high composition of your target audience along with the reach.
The “Research” tab is where most of the action happens. This page displays the functionality of finding sites that match your criteria. You can set the criteria either by demographics (gender, age, education and household income) or by online behavior (you specify a site and it tells you what other sites visitors to that site went to). For sites that match, it gives you the following stats: category, comp index, unique visitors, country reach and page views. It also shows whether you can advertise on the site with Google Adwords along with the specs (text or image, and image size).
I first tried researching by online behavior. I typed in the URL of a client of mine and five sites were spat back – two belonged to my client, one was a competitor’s site and the other two were on topic but have very small amount of traffic.
Next, I tried youbet.com, a horse race gambling site. This time, I got a good number of results. I could apply a filter to show only sites that are in Google’s content network. This means that if I wanted to run advertising for youbet.com, I’ve already found a few sites that are contextually relevant.
I could then go ahead and add these sites so that they appear in the Media Plan tab.
Next, I tried their Demographics research tool. A recent campaign I ran targeted women aged 20-29. They have 18-24 and 25-34 so I selected both of those. What I got back were a lot of parenting sites – nothing too unexpected.
All in all, I’d say that as a research tool, Google Ad Planner seems useful but I would like to see the research criteria get more robust. For those media planners who don’t rely very much on audience profile research, such as myself, the uses are pretty limited.
Daily Links - June 30, 2008
Google To Distribute Content Via AdSense (MediaPost)
Google Uses Previous Searches to Serve Ads on Current Ones (MarketingVox)
Independent Hotels' Online Presence Lagging (MarketingVox)
Lawmakers Shy Away from Online Ad Regulation (MarketingVox)
Daily Links - June 26, 2008
AdPlanner Has Google Edge (Mediapost)
Daily Links - June 25, 2008
Helping A Small Nonprofit To Use Email (MediaPost)
Introducing Google Ad Planner (Google Blog)
Is Google Ad Planner Getting Its Data From The Google Toolbar? (TechCrunch)
Universal Action Tags
I know these aren't anything new, but I recently had a chance to use them for the first time so I thought I'd write about it. The Universal Action Tag (UAT) is a javascript tracking pixel offered by Atlas, Doubleclick and some other ad servers. The beauty of them is that you can piggyback other network pixels on top of them without changing the code on the site. For example, say you have some networks running leadback (or retargeting) for you so that they can retarget people who have already visited the site. In order to do this, they need to place their retargeting pixel on the site. Without a UAT, you would have to have each pixel from your networks hardcoded into the site html by the site admin and then have them remove it once the campaign is over. With a UAT, you just have one pixel coded in and then if you want to add network pixels, you can do it within the ad server console. This isn't just for leadback pixels either - you can use it for optimization pixels as well. This means that theoritically the site needs to get pixels coded into it only once and then they never have to be touched again.
Another benefit of UATs is de-duplication when running CPA campaigns. If you have one or more networks doing CPA, a consumer could potentially view the CPA network ad, and then click on an ad run by a different network and convert. Without the UAT, you would have two networks claiming credit for the conversion. If you use a UAT for counting the conversions, it will ensure that only the last click/view before the conversion is credited.
For more info, you can check out this (and this) literature from Atlas.
Daily Links - June 23, 2008
Google Trends for Websites (Matt Cutts blog)
Short article of free SEO tips (Matt Cutts blog)
Upstart Challenges ComScore, Nielsen Web-Traffic Counts (Ad Age)
Behavioral Targeting Ad Spend Poised to Grow, with Help from Online Video (MarketingVox)
Facebook Launches Profession-Based Ad Targeting (MarketingVox)
