Hack51.Calculate Click-to-Visit Drop-off


Hack 51. Calculate Click-to-Visit Drop-off

Some advertising systems may report far more clicks than ever reach your web site, a frustrating proposition for anyone paying for clicks. Learn where those clicks might be going and how to bring them back.

A common challenge in Internet marketing is reconciling differences in reported numbers from system to system. This becomes especially obvious when you're trying to compare a system that measures "clicks"such as Google and Overturewith a system that is designed to measure "responses." The essence of the problem is that Google will report that you had 1,000 clicks during a month, while your web measurement system reports only 750 paid visits from Google. The question then becomes "Where did they other 250 clicks go?"

If you're unable to find these clicks in the pay-per-click model, then you have to absorb the loss as a cost of doing business. Using the Google example where you lose 250 clicks, to calculate the true cost-per-click for the campaign, you need to add 25 percent to the cost of every click to absorb the missing traffic. Most marketers agree that click-based advertising is expensive enough already, and with keyword costs projected to increase by over 30 percent by 2009, wouldn't it be nice to figure out where those clicks are going?

Needless to say, there are a number of problems associated with counting clicks on the Internet, well beyond the scope of this hack (or even this book). To keep things practical and actionable, let's walk through the steps you should follow to determine if your tracking strategy is working properly and go over some possible causes for the loss.

3.16.1. Make Sure Your Measurement Strategy Is Working

The first step in determining where the clicks went is validating that your measurement and tracking strategy is working properly. We'll use Google AdWords as an example because it is so popular and easy to set up.

To ensure your measurement strategy is working properly, follow these steps:

  1. Since every incoming advertisement should have its own unique tracking URL [Hack #42], create a URL like test_page.html that you can reference with a campaign ID (for example, test_page.html?campaignid=114). No other link should use this URL and campaign ID to ensure that the clicks from this test can be tracked in isolation.

  2. Test the landing page to make sure it is tracking correctly by clicking on the link you created in step one. Does it record the page view, visit, and visitor properly? Check for JavaScript errors when the page renders, and make sure that the click is shown in your web measurement application. Because some applications process data in "real time" and others depend on batch processing, you may need to rerun your reports to make sure the click was reported. If your system depends on being rerun, follow these steps on a test server to cut down on processing time.

  3. Test the same landing page again within 30 minutes. This may seem strange, but you want to ensure that your system records another click on that URL, but not another visit or visitor. If you don't see the second page view, there may be some sort of click defeating system so that if the same person clicks on a URL in a given time, they are not counted twicemake note of this! It is not uncommon for marketing applications like Google to code rules that prevent fraudulent clicks from being recorded. If you have any questions about how the marketing application you're exploring treats successive clicks from the same visitor, ask!

  4. Create an ad in Google using this test URL, search for the appropriate keyword or phrase, and click the advertisement when it is displayed by AdWords. Check to see if your click was recorded by your web measurement system, as evidenced by both a page view and a respondent to the campaign ID. If you don't see the click and response, clear your browser cookies and browser cache, restart your browser application, and click one more time.

Sometimes cookies or the browser cache get in the way of tests like this. Plan on clearing your cookies and browser cache frequently.


At this point, you should ideally see three page viewsone visit, one visitor associated with your test_page.html page, and one respondent to your test campaign ID. There may be more visits and visitors, depending on whether or not you had to clear your cookies and your specific system, but you should have three page views, letting you know that the system was getting the information.

Assuming you've set this all up properly, you're ready to deploy your pay-per-click tracking in the real world. However, the real world is not an elegant four-step test; the real world is an ugly place where clicks get lost for a variety of reasons.

3.16.2. The System Is Working, So Where Did the Clicks Go?

Once you deploy your campaign tracking and start getting clicks, if it still appears that you're still missing visits, based on the number of clicks reported, you'll need to take a long, hard look at your site and see if the problem is on your end. Here are a handful of questions you should ask yourself to further diagnose the problem.

3.16.2.1 Is your landing page a bandwidth hog?

When you clicked your advertisement, were you able to make a steaming cup of espresso before your page fully loaded? Just because you have a broadband connection doesn't mean that everyone clicking to your site does. Some of these "missing" visits may simply be a result of users bailing out before your tracking system is able to measure them. Many industry analysts believe this to be the number-one source of loss in click-to-visit drop-off.

Especially when your visitors are coming from the search enginesweb applications that spend millions of dollars to optimize their sites and information deliverypeople develop an expectation for how quickly a page should load. If a user is happily and quickly searching along before she finds your link, but when she clicks your link she gets put on perpetual hold while your 127-KB home page loads, this click and many like it may back up before it's measured as a "visit."

A good site for checking your web page size is WebSiteOptimization.com (www.websiteoptimization.com/services/analyze, Figure 3-17). The service will tell you the total size in bytes of your landing page, along with number of objects, HTTP requests, and basic recommendations for making your site faster. You may also want to read Web Performance Tuning (O'Reilly) for more insight into how to optimize your web site architecture or leverage your web and performance measurement applications to create a unified view of your site's response [Hack #68].

Figure 3-17. Report from WebSiteOptimization.com.


3.16.2.2 Is your tracking code in the optimal location (for tag-based solutions)?

Even if your landing page isn't a true bandwidth hog, you might be missing visits from the clicks you're paying for because the visitor has moved on before your page tag can be executed [Hack #14]. If you currently have your tags deployed at the bottom of your web pages, move the script to the top and seeif that reduces the percentage-wise loss. If it does help, perhaps your visitors are very fast readers and some are making it through without being counted (but there is nothing you can do about that).

3.16.2.3 Is your tracking code still deployed properly (tag-based solutions)?

Sometimes changes get made to sites that affect the data collection capabilities, and measurement can get lost in the shuffle. Always verify that visits are being tracked on your landing page correctly. We recommend setting up a monthly reminder to audit your response tracking code. This will save you many headaches in the long run, especially with web sites in which many people are working on the code.

3.16.2.4 Are you tracking page views, visits, or visitors?

If your web measurement solution reports on the number of page views, visits, and visitors from a referring URL or campaign, you are in luck. Clicks are page view analogsone click at Google should result in one page view in your system, every time, in a perfect world. However, you want clicks to be a visit analogyin which one click results in one visit to your web sitebecause if someone has to click two or three times to visit your site, you are paying extra for that visitsomething you don't want to do!

Make sure you understand what both systems are reporting to you. Most pay-per-click systems are reporting exactly that, clicks; discrepancies are introduced at the measurement end. Call or write your measurement application vendor and ask them what the system is reporting, especially if the system refers to what it reports as "responses," a term that may be interpreted differently by different vendors.

Because the Internet and HTTP are not perfect, clicks will inevitably become lost. The best guidance is to work to minimize the number of lost clicks and, if large discrepancies continue, consider spending your hard-earned advertising dollars elsewhere.

Dylan Lewis and Eric T. Peterson



    Web Site Measurement Hacks
    Web Site Measurement Hacks: Tips & Tools to Help Optimize Your Online Business
    ISBN: 0596009887
    EAN: 2147483647
    Year: 2005
    Pages: 157

    flylib.com © 2008-2017.
    If you may any questions please contact us: flylib@qtcs.net