digiterrati
July 5, 2007 10:47 PM Subscribe
Ad server/landing page discrepancies: what causes them?
If an ad server records that an ad got a click, but no request to the landing page is recorded, then what most likely happened?
Conversely: what if the number of clicks the ad server records is *fewer* than the number of requests that the landing page receives?
Suppose that in either case, the discrepancy is in the tens of thousands...
If an ad server records that an ad got a click, but no request to the landing page is recorded, then what most likely happened?
Conversely: what if the number of clicks the ad server records is *fewer* than the number of requests that the landing page receives?
Suppose that in either case, the discrepancy is in the tens of thousands...
It's possible someone clicks on an ad and bails out (closes browser, hits the back button, loses network connection), or that the user has a routing or network latency issue that affects the ability to successfully request the landing page, but not the click tracking page on the ad server. These could partially explain lower numbers on the landing page.
Many visitor reporting systems may also be trying to give you a count of "unique" visitors. Some do this simply by discarding repeated hits from the same IP address in a certain time period, some use cookies (which your browser may be configured to reject), some embed a one-use token in the URL you've clicked, and some use some hybrid approach or something altogether different. Often a discrepancy between two systems which are ostensibly recording the same traffic is caused by different methods of identifying "uniqueness."
Visitors reloading the landing page or navigating the site and returning back to the landing page may also cause a mismatch.
There are also different methods of tracking visitors that may not work in all browsers, such as an inline image tag which loads a single-pixel image tied to a counter or tracking system. If each system tracks differently, browser differences and privacy settings may cause one system to undercount.
Some ad servers also have anti-fraud systems which attempt to throw away clicks which it thinks are part of a pattern of fraud such as automated or repeated clicks, clicks from geographical areas with high instances of fraud, or even very complex statistical models that may just catch some false positives. This may cause the number they report to you to be lower than the number of HTTP requests their ad server actually received.
Lastly, it's possible there is a bug in either tracking system, but consider all of the above before jumping to this conclusion.
posted by hutta at 11:34 PM on July 5, 2007
Many visitor reporting systems may also be trying to give you a count of "unique" visitors. Some do this simply by discarding repeated hits from the same IP address in a certain time period, some use cookies (which your browser may be configured to reject), some embed a one-use token in the URL you've clicked, and some use some hybrid approach or something altogether different. Often a discrepancy between two systems which are ostensibly recording the same traffic is caused by different methods of identifying "uniqueness."
Visitors reloading the landing page or navigating the site and returning back to the landing page may also cause a mismatch.
There are also different methods of tracking visitors that may not work in all browsers, such as an inline image tag which loads a single-pixel image tied to a counter or tracking system. If each system tracks differently, browser differences and privacy settings may cause one system to undercount.
Some ad servers also have anti-fraud systems which attempt to throw away clicks which it thinks are part of a pattern of fraud such as automated or repeated clicks, clicks from geographical areas with high instances of fraud, or even very complex statistical models that may just catch some false positives. This may cause the number they report to you to be lower than the number of HTTP requests their ad server actually received.
Lastly, it's possible there is a bug in either tracking system, but consider all of the above before jumping to this conclusion.
posted by hutta at 11:34 PM on July 5, 2007
Response by poster: The situation has nothing to do with Google. This is an ongoing issue with various third-party ad servers such as Doubleclick and Atlas.
Whether or not certain clicks should count as freebies is uninteresting to me; I am well versed in the business implications. I've been an online ad salesman, an online media planner, and an ad trafficker. I want to know what is actually happening.
The issue is not the definition of uniques. The ad server claims to record the number of times upon which the ad was clicked. The number of hits to the landing page is simply the number of times that a request was made for the landing page URL.
This is not a special problem that I've observed in a particular context (although I am trying to troubleshoot a particular instance at the moment). It's an industry-wide problem that occurs constantly.
hutta's explanations about users who bail out, and landing page URLs that are refreshed, do seem like the most obvious explanations, and I've considered those before. But it seems to me that those things are unlikely to occur often enough to cause the huge discrepancies that I see as a matter of course, across multiple campaigns.
posted by bingo at 6:22 AM on July 6, 2007
Whether or not certain clicks should count as freebies is uninteresting to me; I am well versed in the business implications. I've been an online ad salesman, an online media planner, and an ad trafficker. I want to know what is actually happening.
The issue is not the definition of uniques. The ad server claims to record the number of times upon which the ad was clicked. The number of hits to the landing page is simply the number of times that a request was made for the landing page URL.
This is not a special problem that I've observed in a particular context (although I am trying to troubleshoot a particular instance at the moment). It's an industry-wide problem that occurs constantly.
hutta's explanations about users who bail out, and landing page URLs that are refreshed, do seem like the most obvious explanations, and I've considered those before. But it seems to me that those things are unlikely to occur often enough to cause the huge discrepancies that I see as a matter of course, across multiple campaigns.
posted by bingo at 6:22 AM on July 6, 2007
Response by poster: p.s. Before anyone else says it, I'm aware that Google now owns Doubleclick, but for the purpose of this question, that doesn't matter. I'm talking about old-fashioned banner ads trafficked through DFP (for example), not AdWords.
posted by bingo at 6:24 AM on July 6, 2007
posted by bingo at 6:24 AM on July 6, 2007
« Older If you were doing the Peace Corps, what reading... | I just want a more classic Vista. Newer »
This thread is closed to new comments.
If the landing page receives MORE requests against it than the ad server records, I'd consider them freebies. Since the only way to reliably ascertain the source of a clickthrough request (that is, to assure that the URL wasn't simply abused outright, though you can still do this through the ad-requesting link), is by going through the ad server, that's the price you pay.
Depending on the nature of the landing page, you could verify the referring page, but that's not always provided by all user agents.
What most likely happened? How is the script written? What exactly occurs on the backend? If it's a simple database call and then a header() away from the ad request, then perhaps the script failed for some reason. If it's occurring randomly, then perhaps the landing page isn't actually logging the incoming requests as it should be.
If the ad server is something like Google AdSense, and Google tells you they received a click, but your server logs don't show anything, and the problem is in the tens of thousands, are you certain that your refer-to URLs are configured properly? Google has metric-tracking *AFTER* a clickthrough, so you can follow a user's actions and determine clickthrough conversions to purchases, or in your case, ANY clickthrough activity as having been legitimate.
You need to give much more specific information about the set up you have for us to be of absolutely any help at all, and even then, we can only speak in terms of likelihood and probability. (For instance, it's MORE likely that you screwed up the configuration of your campaign than it is for Google to be recording non-forwarded clickthroughs.)
Meanwhile, click-fraud is something Google takes very seriously, and you should consider their metric tracking and again, read your terms of service.
Of course, you never once mention Google and the devil is in the details...
posted by disillusioned at 11:34 PM on July 5, 2007