Ever since we realized we could really develop accurate metrics that detailed how readers were responding to ads, data analysis has become something of an obsession for us.
If there ever was a holy grail for advertising, it's the ability to accurately compare rates and ad performance. We're finally reaching the point where that goal is attainable.
What we track
When you're talking about Internet-based advertising, there are two elements you want to track: how many times an ad is shown and how many people click on the ad once shown. Once you have those two pieces of information, you're able to determine how much an ad costs and the door opens for you to determine your return on investment.
"DominoPower advertisers smart enough to take advantage of this program will be paying only 2/10ths of a penny each time their ad is seen."
|
Mailing opens
For our mailings, we track both the number of messages that are opened by readers as well as the ads that are clicked on. Whenever one of our HTML email messages is opened in an email client (and the email client allows it), the act of opening that message triggers a request to download an image from our server (called a "Web bug") and that act of downloading triggers a small tracking script running here on our servers.
Now, the gotcha is that not all email clients will properly display HTML and the newer clients make the display of images an option for the user. As a result, our metrics for email opens is conservative. The number our tracking system reports is the number our server records. The actual number is likely to be higher.
Mailing ad clicks
When a user clicks an ad link in our mailing, the user is briefly transferred to a CGI page on our server, which logs a whole pile of information about the click, including the when and who of the click. It then transfers to the destination Web page URL.
Each advertiser, on their Web server's end, is likely to track those arriving clicks differently. The vast majority track the incoming clicks by scanning through a Web server log file, looking for referrer data. This is not always accurate, especially when the originating referrer isn't a Web page, but an HTML email. Others, who use their own CGIs for tracking, are often somewhat more accurate.
Because each advertiser tracks incoming traffic differently, we always track the who-what-whens of clicks as part of sending them on to the destination pages. This makes for far more reliable metrics.
Web server page views
On the magazine Web pages, we track how many times ads are shown on the site. Unlike many Web sites who track "hits," we track impressions. Here's the difference. A hit is any request of data from a Web server. An impression (also called a "page view") is the number of times the complete page is viewed by a person.
Since a typical Web page is often made up of many chunks of data (each graphic, for example, triggers a new hit), one view of a Web page by one reader is likely to be counted as somewhere between 5 and 50 hits. If we tracked hits, like many of our competitors, this one page you're reading would be counted as 19 hits. But you and I both know that you're the one reading it, the one viewing it.
That's why we track page views, rather than hits. Our metrics let us know each time someone views a page. Even if the page has 19 graphic elements and a bunch of smaller text chunks, it still counts as one view. And that's accurate.
For the ZATZ magazines, every time a page is viewed, we increment a database element on our server. We can track how many times, on a given day, ads are served to our readers.
Web server ad clicks
As we all know, ads have value even if they're never clicked upon. After all, every time someone sees your product, your company name, your branding, and your message, you're creating an impression on his or her psyche. And, as an advertiser, that's a good thing.
That said, it's also nice to know just how many people were transfered from our Web site to yours. And, for that, we track each click. Again, we use a special script that stores data in an object database on our server. As a result, we're able to look at each click for each ad and have the accurate details for further analysis.
Measuring value
Once you're able to track performance, another step is measuring value. And that requires finding a way to compare the costs of ads. Imagine you're trying to choose between two coffee pots and one costs thirty dollars and another costs seventeen bunnies. If you don't know how to translate dollars to bunnies, there's no way to tell which is the better price.
The meaning of CPM
On the Internet, ad agencies have pretty much standardized on the concept of CPM, which stands for "cost per thousand impressions." The "C" and "P" make sense, standing for "cost" and "per," respectively. The "M" makes less sense, until you realize the "M" stands for mil meaning "one thousand."
For those of you who've been around computers since the late 70s, CP/M may also mean "Control Program for Microcomputers," one of the very first PC operating systems. In our case, however, we're talking about ad performance.
By the way, the ad industry tends to use the terms "impression" and "page view" interchangeably. We'll do our best to stick with the "page views" term.
What's nice about CPM is that it's a measure of not only how much you're paying, but how many times your ad is shown. Obviously, to be able to provide a CPM value, you have to be able accurately track just how many times an ad is shown.
This is where you need to be very careful. When we quote a performance number, we take it directly from our server's tracking logs, usually measured to the moment we're getting the data. Some of our competitors aren't so rigorous.
How CPM is calculated
What we do is retrieve the number of page views from our tracking system. Since we're talking about the number of people viewing a page or the number of people we've verifiably proven have opened an email, we can know precisely how much viewage an ad has gotten in a given period.
If an ad is in rotation, like the ads in the online magazines, we'll then divide the total number of page views by the percentage of the rotations. For example, we alway have a maximum of 20 possible banner or leaderboard rotations in the magazine. If there were 100,000 page views, we'd divide that by 20 to find out the number of page views per rotation (and we'd get 5,000).
So, if you've paid $250 for a rotation of a banner, and you got 5,000 page views per rotation, you could then divide 250 by 5,000 to find out that you're paying a nickel per page view. Because CPM is measured in thousands, you'd then multiply 0.05 times 1,000 to determine that you're getting a CPM of 50. All that means is you're paying $50 to show your ad to 1,000 people.
Using CPM to compare ad programs
Let's take a few different ZATZ ad programs and use CPM to determine how they really compare when it comes to price. A mailing ad in one of our publications can cost $1,000 if it's in the top position and $100 if it's in Position 10.
How many of us have a stack of subscribed to, but unopened magazines? Did the advertisers in those unopened magazines gain any value at all? Of course not. It's only when the magazine is opened and read that the ads are seen.
A value example based on ads seen
For the months of July through October 2004, the average number of people who opened the DominoPower emails each week was 2,973. If we calculate CPM of the top ad based on the number of people who opened the email, a Position 1 ad has a CPM of $336 (meaning it costs $336 for a thousand people who you know looked at your ad) and a Position 10 ad has a CPM of $34.
Obviously, if you're selling products and services that cost thousands of dollars, a CPM of $336 might be a good deal, although a CPM of $34 is a better deal.
A value example based on ads mailed, but not necessarily seen
By the way, our DominoPower and OutlookPower competitors claim CPM rates of $50-100, which makes it appear they're competitive and may even have better rates. The difference is that we're quoting the CPM of actual, verifiable opens while they're usually quoting CPM values based on the number of messages mailed out.
Take a look at how that little trick makes them (and us, if we did the same thing) look much better than they are. What follows is what our CPM values look like based on counting the number of messages sent rather than the number of messages opened.
For the months of July through October 2004, the average number of people we mailed DominoPower to was 66,599 per week. If we calculate the CPM of the top ad based on the number of people we mail to, a Position 1 ad has a CPM of $15 (meaning $15 for a thousand people mailed to) and a Position 10 ad has a CPM of $1.50.
Instead of a CPM of $336 for the top position, we're now showing a CPM of $15. And instead of a CPM of $34, we're now showing a CPM of $1.50. Isn't it amazing how the numbers change?
By doing these metrics, you can really see how the value formulas compare. This is worthy of repetition: be sure you know what's being measured. Is it actual opens or messages sent?
Using CPM to value our extended advertising program
While we're always willing to sell a single ad or two, we've worked hard to develop an advertising program that generates tremendous value for you as an advertiser. And CPM can help prove that value.
We now offer a program whereby you can nail down your entire 2005 advertising plan, run an ad in each and every issue of our wildly popular OutlookPower, DominoPower, or WebSpherePower mailings, and have a constant presence on our OutlookPower, DominoPower, or WebSpherePower Web sites in two ways, all for a single payment of $6,000.
You buy this program for one of the magazines, and when you do, you get the following in that magazine:
- a Weekly Update Text Ad (floating position) with graphic (48 run - 4 per month, $17,800 value);
- constant presence with 1 rotation of Web page text ad throughout the year (12 run, $3,000 value);
- constant presence with 1 rotation of Web page banner throughout the year (12 run, $3,000 value);
- the confidence that comes from having an active, constant ad program throughout the entire year;
- ongoing free marketing support to help you tune your message and even create your ads and graphics, if needed;
At first, it may seem like $6,000 is a lot of money. But when you use the CPM system to do a value comparison, the resulting accurate value become very interesting.
First, of course, we need accurate tracking data. We took the data gathered from July through October of 2004 and used that to predict the total number of page views and opens for 2005. Obviously, performance data from the summer months are going to be a bit lower than the rest of the year, which makes these calculations even more conservative.
Then, we made sure we're calculating page views and opens based on actual views and opens, not based on the number of items mailed or on hits.
Once all the numbers are crunched, you get the results shown in Table A.
TABLE A: Comparitive CPM of ZATZ Extended Ad Progam
| Metric |
DominoPower |
OutlookPower |
WebSpherePower |
| Total Page Views |
2,801,904 |
785,052 |
229,584 |
| Total Email Opens |
137,883 |
456,750 |
107,601 |
| Total Views |
2,939,787 |
1,241,802 |
337,185 |
| Approx cost per view |
$0.0020 |
$0.0048 |
$0.0178 |
| Approx CPM |
$2.04 |
$4.83 |
$17.79 |
DominoPower advertisers smart enough to take advantage of this program will be paying only 2/10ths of a penny each time their ad is seen. OutlookPower advertisers are paying only about half a cent each time their ad is seen, and even WebSpherePower advertisers, in that new and wide open market are paying less than a two cents.
And, when you look at the CPMs, you begin to see how we're able to make truly tangible comparisons. Our competitors are quoting CPMs of $50 to $100 and yet this program has CPMs running from $2.04 to $17.79.
What's even more amazing is how the value formula differs even from a single ad buy. Instead of spending a CPM of $34 to $336 on DominoPower, you're spending a CPM of $2.04. The savings are enormous, the long-term benefits are huge, and, for the first time, you can tangibly measure and prove the value.
Print-friendly version of this page
|
