My last few columns have focused on the accuracy (or otherwise) of Google AdWords click-tracking and billing system. AdWords earns Google over $1 million per hour, 24/7, which makes it the most important web analytics system on earth. The accuracy (or otherwise) of that system underpins the most important financial institution on the web.
In the first article I described problems I had personally encountered. On some occasions AdWords was billing me for much greater volumes of traffic than I could account for using both my own and Google's own web analytics systems. Two different systems that confirmed each other's numbers, yet disputed the AdWords' numbers, raised the issue of how accurate Google AdWords was when it came to counting clicks and generating bills.
In the second article I covered Google's response, which was that the AdWords system is always 100 percent accurate. According to Richard Holden, Google's director of product management, the click-tracking and billing system of Google AdWords never makes a single mistake-- ever. Holden explained that, in Google's view, if you complain about billing accuracy it is merely a sign you need to be educated as to the error of your ways. Since, in Google's view, AdWords can never make a single mistake, if your numbers disagree with its, your numbers must be wrong.
Google's role is therefore to explain to you why your system has come up with different, incorrect, numbers. As a consequence of this, Google simply does not have a mechanism for checking the accuracy of AdWords. Where most companies would have a complaints' investigation procedure, Google has a customer re-education process.
Google's primary educational tool is a single web page on its website. This document seeks to explain why AdWords and Google's own Google Analytics may report different numbers. Some, if not all, of the arguments contained in this page will apply to other web analytics systems you may be using.
In this article I will list each of the reasons on Google's web page, and analyze its validity.
Why AdWords and other web analytics systems may differ (according to Google)
"Make sure that you're comparing equivalent items. Google AdWords tracks clicks, while Google Analytics tracks visits. If a user clicks on your ad twice within thirty minutes without closing his or her browser, this will be registered by Analytics as one visit to your site, even if the user left your site and then returned shortly after. For example, if a user clicks on your ad once, clicks the 'back' button, and then clicks your ad again, AdWords will register two clicks while Analytics will register one visit."
This is perfectly valid reasoning and this situation will lead to Google AdWords' count being higher than yours. The formal definition of a visit is "a series of page requests with a gap of no more than 30 minutes between each one." If someone leaves your site, then returns within 30 minutes, it's still the same visit. Technically it's two separate sessions, but most analytics systems don't report sessions.
It's not uncommon for this to happen, especially if you're selling online. People will want to compare your site against your competition before making a decision. In fact, it is rare for a purchase to happen on a first visit. Whether someone is likely to return within 30 minutes or not depends largely on the nature of the purchase. The more expensive the purchase, the longer it will take for them to weigh up the opposition and make a decision. Sites selling mortgages and loans may see repeat visit cycles measured in days or even weeks rather than minutes, whereas the DVD sales cycle can be fairly short.
You are best placed to understand what behavior your visitors are most likely to exhibit in this regard. However, in my experience the degree of discrepancy between Google AdWords and your own analytics system due to clicks versus visits is unlikely to be more than a few percentage points.
"AdWords automatically filters certain clicks from your reports, while Analytics will report on the resulting visits to your website. The clicks we filter from your AdWords reports are the occasional instances of someone clicking repeatedly on your ad in order to increase your costs or to increase your clickthrough rate. AdWords considers these clicks to be invalid and will automatically filter them out of your AdWords reports. You are not charged for these potentially invalid clicks."
Google will not count as two clicks someone who double-clicks hyperlinks because they think you have to double-click everything. Google will also filter out any intentional click fraud they detect.
Click-fraud can happen for two reasons: a competitor may wish to blow your budget and kill your ad, or someone might be trying to make some money off you. Affiliate sites get 50 percent of the revenue Google earns if you click a Google Ad they are running. So it's in their interest to boost the clicks. There are companies (mainly in India) who boast sophisticated technology to do this for the affiliate in a manner which is hard to detect. Affiliates do deals with them and split the revenue.
Google states this is a minor problem which they are on top of. Others suggest it is a major industry earning millions, if not billions, of dollars. Google replies these people are inflating numbers for their own purposes. Google refuses to disclose the exact scale of the problem, but where they have been sued over this, they invariably settle out-of-court for multi-million dollar sums. If you believe click-fraud has cost you money I suggest you sign up with your nearest class action-- there's bound to be one available.
If your AdWords are being filtered by Google for invalid clicks, this will push the AdWords numbers down. Invalid click filtering will never cause AdWords to have numbers higher than your analytics system.
URLs Not Tagged
"If auto-tagging is turned off and the Destination URLs do not contain manually tagged campaign tracking variables, the visit will not be marked as Google CPC, but instead may be attributed to Google Organic. Please ensure that your AdWords' account either (1) has auto-tagging turned on or (2) has campaign tracking variables appended to the end of every destination URL."
While this rule is phrased in a way that refers to Google Analytics, the logic holds for any other web analytics system. If you don't add some form of tracking parameter to the link in your ads that leads to your website, your system won't be able to tell the visitor came from that ad. This also requires that your system is set up to count and report these parameters. This should be self-evident, but the majority of Google Ads do not contain any tracking parameters (see my article "What Analytics Do You Use?").
If you are complaining to Google about the AdWords billing and you are guilty of not using tracking parameters, Google is quite right-- you do need educating.
Not just any old tagging will do if you want to argue with Google. You need to ensure your tagging matches their billing. If you are being billed separately for five different ad groups, you can't put together a strong case if you're using a single parameter that amalgamates them all into one. If your total doesn't match theirs, you need to be able to drill down and pin-point exactly which ad group you believe is being over-charged. In other words, you need a tracking parameter that corresponds to each individual billing line in your AdWords statement.
Google's webpage would appear to be the definitive list of all the possible reasons Google can think of for discrepancies between AdWords and another analytics system. If Google staff are trying to educate you, their procedures dictate that you be referred to this document. You may also, in my experience, receive emails from them containing points copied from it. I have yet to see any examples of Google staff coming up with any possible reasons that are not contained in this web page in writing.
There is another reason that may cause a discrepancy between AdWords and your system: People may click the ad, and then fail to arrive at your site. They may change their mind and hit the back button before your page is loaded. The Google re-direct may be slow and people may give up waiting (I've personally experienced this many times). I am sure there are other reasons as well.
Richard Holden mentioned this as a cause of discrepancy in our interview, and I've heard it from other Google staff in conversation. One wonders why Google have not put this point in their web page? Could it be they have no desire to publicize the fact that even if someone never sees your site, you still have to pay for it?
All of the points Google does make are perfectly valid, and can lead to a discrepancy between your numbers and theirs. However, no one complains to Google if the numbers vary by a few percentage points. The issue is whether these reasons can account for discrepancies large enough to trigger a billing dispute. In my opinion, if your system is implemented correctly, the maximum possible variation should be no more than five percent. I have personally seen variations of 25 percent with some regularity and recently a consistent difference of 70 percent. Nothing in Google's documentation can possibly explain such variations on sites where the analytics systems have been implemented correctly.
The key factor controlling the process of AdWords billing disputes is Google's unquestioning belief that the AdWords click-tracking and billing system is 100 percent perfect. According to Google your system has to be wrong because AdWords click-tracking is the most perfect computer system ever created by the human race. In this sense it is the absolute pinnacle of technical evolution.
In writing this article I have tried very hard to think of any other technology that even vaguely approaches perfection, but I can't. When you consider the amount of data it processes, you realize Google AdWords' click-tracking must be more accurate than an atomic clock! It defies the laws of physics in its perfection. Of course, you just have to take Google's word for it because it's so wonderful they can't tell anyone else anything about how it works to prove their claims.
Google does have a point; even if your analytics system is correctly implemented, some variation in numbers will occur. But if that variation is very large, you can take it from Google-- your system is faulty. Because Google AdWords is absolutely perfect.
If, by now, you still stubbornly refuse to be re-educated as to Google's Absolute Perfection, Google will happily remind you that if you want to use AdWords, Terms & Conditions stipulate you have to accept their numbers anyway. So tough!
Brandt Dainow is an independent web analytics consultant. Read full bio.
Landing Page Tracking Code
"If the landing page for your ads is not being tracked, your campaign information will not be passed to Analytics. Please ensure that you are tracking all landing pages for your AdWords ads."
Web analytics systems know how you got to a page by asking your browser what page you were on previously. Your browser has a built-in capability to report this. This is known as the 'referrer field' and it has been around on the web since Day One. The referrer field contains the full URL of the page containing the link you clicked to get to the current page.
On the other hand, the browser can only report the referrer field for the page it is actually on. There's no way to ask the browser for the referrer field for something three pages back. Therefore the only way to tell where someone came from is at the point of entry. This means you have to get the referrer field for the first page in your site that they land on. If you don't get it then, you can never tell where they came from. You have to ensure all landing pages contain tracking, and that tracking needs to get the referrer field.
To many of us, this seems self-evident. However, the fact that Google needs to mention it suggests many people are not aware of this. It suggests that at least some people creating online marketing material may not be aware of the site's tracking system. I've seen this happen myself when outside agencies are producing the online material, especially if they're hosting it themselves. You need to ensure that everyone who creates pages for the site is fully briefed on the tracking requirements.
If you're not tracking landing pages fully, your own numbers will be under AdWords. The more successful the campaign, the greater the discrepancy.
Another factor that can create the same problem is that some users can prevent their browser from providing the referrer field when asked for it. Some anti-spyware products do this, and it is an option in a few browsers. However, in my experience the percentage of people doing so is rarely more than one percent. It largely depends on the audience; if you run a conspiracy theory site for paranoid techies, it could be higher.
Visitor Browser Preferences
This can, and probably does, contribute to some discrepancy between AdWords and your analytics system. However, I have personally conducted extensive research into how common this type of blocking is. Among people using Linux-based browsers, it's around 10 percent. Approximately 0.5 percent of PC users and less than 0.25 percent of Mac users also block tracking in this way. In total, the discrepancy in numbers between your system and AdWords as a result of browser preferences is unlikely to exceed one percent.
Unable to Load Code
"Clicks reported on Google AdWords but not on Google Analytics may be the result of obstruction between the Google AdWords click event and the ability to load the tracking code on the landing page. If this is the case, ensure that your web hosting servers are functioning properly, the page is loading for all possible users and IPs, and the tracking code is installed correctly on your web pages."
Alternatively, the sequence of "image call -- server response -- cookie setting" may not have completed before the visitor leaves the page. It can depend on the location of the visitor and the tracking server. Tracking servers in the northern hemisphere can have problems with visitors in Australia or parts of Asia. This very much depends on the number of hops the signals must go through and the speed and quality of the networks involved. I have seen situations where additional tracking servers had to be placed closer to the user-base to handle this.
While these situations will be the cause of discrepancies between AdWords and other systems, the number of times such things will occur is fairly low. As with browser preferences, the variation from such cases is extremely unlikely to exceed one percent.
Redirects in Landing Pages
This is not necessarily true, although the reason depends on the type of redirect. 30x redirects are produced by the server. When a specific page is requested, the server sends a message back to the browser saying "page not here, get it at XYZ." The browser has to then try to get the page from the new location. RFC 2616, which defines HTTP 1.1, doesn't specifically state how referrers should or should not be handled by 30x redirects. However, since the referrer is the page that generated the original request, the fact the server redirected does not mean the page from which the click originated has changed. Browsers will maintain the originating referrer during a redirect. Provided the page the visitor lands on contains tracking code, you will still accurately attribute that visitor to the AdWord.
In both the above cases, you need to ensure the tracking code runs correctly. It should be obvious that if your code doesn't run, your numbers will be off. Beyond that, there's no valid reason why redirection should cause differences between your analytics package and AdWords.