Monday
An Internet Advertising Primer
Web advertising is typically sold on a "Cost Per Thousand Impressions" basis, or CPM. An impression (also referred to as an "exposure" or "page view") occurs when a visitor to a Web site views a page where an ad is displayed, whether the ad is seen or not. Whenever a page is "served" to your computer screen, measurement software counts the "impression."
Ads can be placed anywhere on a page, although ads are more successful when they are displayed near the top of the page("above the fold"). This ad will deliver higher click-throughs than the same ad placed at the bottom of a page. Some sites use frames to display advertising more prominently.
A common banner ad package consists of 100,000 impressions costing anywhere from $20 to $100 per thousand ($25-$70 CPM is average for popular sites). Keep in mind that most sites have repeat visitors, and most visitors view more than a single page of information, so your 100,000 impressions could actually represent 10,000 to 40,000 unique visitors. In some cases, impressions are guaranteed, in others, estimated. Some sites will charge a flat rate for estimated impressions. If the actual delivered impression count is higher, some will let you keep the "overdeliveries" for no extra charge. Most popular sites offer ad agency discounts and volume or frequency discounts. The ad agency rate, known as the net rate, is typically a 15% discount from the gross rate, which is the typical rate if you purchase advertising directly instead of through an agency.
"Hits" and "page hits" are not favored as measurement units because a hit is registered every time any text or graphic file is delivered, whether advertising is displayed or not. A page hit, or HTML hit, is registered every time a complete HTML page, including text and graphic files is delivered. The ratio can easily be 10 or more hits or 2-4 page hits to actual impressions.
Advertising measurement is performed primarily by using tracking software and server activity logs, which meter, but understate, impressions. This undercounting can be caused by "caching", "site mirroring" and "firewalls." With caching and site mirroring, the page displaying the ad is sometimes delivered once, then stored locally on the visitor's computer or the ISP's computer. When the page is viewed later -- by the same visitor or other visitors -- it is retrieved from local memory rather than from the Web site. It's faster for surfers, but prevents the tracking software from counting it as an impression delivered from the Web server.
A firewall with proxy server is used by corporate intranets for security, but it also can cause impressions to drop because it's possible for 10 or 10,000 corporate employees to be identified by the same corporate visitor address.
Tracking software can't measure the number of unique visitors so statistics, surveys, and subscription information is used to supplement the estimates. It's conceivable that a single visitor, who visited a Web site twice per hour, every hour, for a month, could register 1,440 impressions according to the tracking software.
An undercounting twist to this scenario: many ISPs assign the same IP string to each new user as the previous user logs off. So, for example, a new Chicago-area visitor from the Netcom service who visits a Web site could be identified as chi-il11-13.ix.netcom.com, even if a previous Netcom user with the same temporarily assigned IP string just left the same site minutes ago. Proxy servers and local caching also hide portions of IP strings from the server log access files. So, during a very busy time period, more than one unique visitor could visit a site with the same reported IP string, and the log will consider it a single visitor.A solution that many sites use to correct their traffic measurement is to set a cookie that uniquely identifies each visitor. This is assigned, and stored in the browser's cookie file the first time you visit the site. On subsequent visits, the unique cookie helps identify the visitor in conjunction with the IP address. Cookies are only partially helpful in identifying unique visitors because they can be turned off or the visitor may use different computers and browsers with differing cookies, which makes them a less than perfect confirmation tool.
An intelligent agent, or "bot," could automatically visit a site every hour and inflate a visitor count. Variations on intelligent agents include "Spiders" and "Crawlers," which are software programs that visit virtually every page on the Web to create indexes for search engines, but cause overcounting of traffic in the process. Spiders and crawlers are much more interested in text files than graphics files such as banner ads. If the tracking software is counting requests for a text page, rather than the ad banner, this could result in overcounting actual ad impressions. Some flawed spiders and crawlers continually request the same text page.
Virtual includes and filters can also cause misrepresentation with visitor counts. "Virtual includes" are techniques that are used to capture only a desired image, such as a weather map, without a user having to visit a page and view banner advertising. Filter software such as "Junkbuster" enables a user to strip banners and cookies from accessed pages in order to speed up page delivery. Virtual includes and filters cause a reduction of delivered ad impressions but not necessarily accurate tracking of the reduction.
Browser bugs are constantly changing but are constantly present. A browser bug can cause a request for the text portion of a page twice, but the graphics only once. The impact on the ad impression count depends on the tracking software.
Rogue bots are designed to deliberately inflate the traffic count on CGI-generated in-line GIF advertising images. These graphics are usually only displayed, and counted, when a specific Web page has a legitimate visitor as determined by the http_referer field sent from the client to the server during http transactions. A rogue bot acts as a "visitor" to a web site and sends a fictitious http_referer and a fake IP address. Rogue bots are rare.
Internet bottlenecks can cause the visitor to request the same page repeatedly until the full page with graphics is actually displayed. But the traffic counting software is counting all requests for the full page, thereby overcounting impressions.
Bottlenecks affect larger banner ad graphics more than smaller banner ad graphics. If a banner ad is large, it takes longer to load, giving the visitor a chance to move on to another page or site before it is displayed - yet the impression is still counted. But larger banner ads generally deliver higher click-throughs to the advertiser's site, so there is a trade-off. Smaller banner ads may load faster, ensuring the visitor sees them, but they may be too small to capture the visitor's interest.
Some Web users choose to run their browsers with "graphics off" to speed up delivery of pages. For these users, a text advertising message is delivered instead of a graphic banner ad. Advertisers are more frequently demanding that they only pay for ad impressions when the actual banner ad graphic is fully loaded, not just the alternative text version. Industry estimates place "graphics off" browser usage at 6 percent. Requiring the visitor to navigate a site using only graphics and not text links can significantly improve this problem.
"Cache-Busting" refers to the ability by ad management systems such as MatchLogic and Imgis to "bust" cached banner ads (to "view" inside a cache and count the number of ads delivered from the cache rather than from a server). This improves accuracy of reporting delivered advertising impressions. These cached ads previously were not counted. A controversy arises because historically, no advertiser had to pay for cached ad impressions, they were a "bonus." Now, the same advertiser finds that they may have to pay more to advertise on the same site, because the "bonus" ads are now being counted and charged for.
Some web sites are now also offering "cost per click" or "cost per click through," which is a direct measurement of response and ad performance. A click through occurs when the visitor sees or reads the ad and clicks on it, taking them directly to the advertiser's Web site. Sometimes a visitor will click, but quit before the other Web site is displayed. Click through averages anywhere from 1 to 25% of total impressions, but generally 1 to 3% is used as a rule of thumb. The overall average for click-throughs is falling to around 1 percent. Web sites try to avoid selling advertising this way because poorly designed ads won't be clicked on, reducing ad revenue. Numerous other advertising models require performance results before the Web advertising site is paid. Examples include banner ads sold based on percentage of sales, which has been done by Nabisco, and based on a bounty for every software program download, as has been done by Microsoft.The common log format is the standard for most Web servers where your Web site might reside. It exists as a text file in the log directory and can be viewed on-screen or downloaded by anyone who has FTP or Telnet access to the server files. Numerous log formats exist, each of which can be used by themselves to analyze Web site traffic. Traffic analysis software then takes this raw data and plots it for easier reading. The common log will report the following information on every visitor and what activity they performed on your site:
host/ip
• RFC name
• logname
• datestamp
• retrieval
• code
• bytes.
They look like this:
unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:14 -0800] "GET /login_on.gif HTTP/1.0" 200 8452 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:15 -0800] "GET /ad.gif HTTP/1.0" 200 22443 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)"
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:20 -0800] "GET /bg.jpg HTTP/1.0" 200 4377 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:12 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/3.0 (Win95; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:45 -0800] "GET /ord.com HTTP/1.0" 404 240 "-" "Mozilla/3.01 (Win16; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:14 -0800] "GET / HTTP/1.0" 200 7047 "http://www.netgambit.com/" "Mozilla/3.01 (Win16; I
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:20 -0800] "GET /links.html HTTP/1.0" 200 7404 "http://www.positionagent.com/" "Mozilla/3.01 (Win16; I)
" burn-mx0508707.smartt.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:02:05 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/4.0b2 (Win95; I)
Host/IP
The first field in the record is the host/IP. It indicates the hostname, or RFC name, of the visitor or an IP number if domain nameserver lookup is not enabled for that visitor. The example above shows the activity of three host/ip visitors. Visitors can have .com, .edu, .net, .uk and other extensions, indicating whether they originate from a company, an ISP, a university, or another country.
Datestamp
The next important field is the datestamp, which is useful when running the log figures through analysis software. The datestamp information can be used to graph peak activity throughout the day, week and month. This will show you whether traffic picked up after your banner ad started appearing.
Retrieval
The retrieval method is posted. GET /links.html indicates that the link resource page was retrieved by the visitor. Most visitors retrieve multiple files which make up a single Web page. Some counters would indicate three hits to the Web site. Others would indicate a single HTML page hit. If a banner ad also appeared on the page and was retrieved with the other three files, an appropriate counting device would register an impression.
Code
The code field indicates whether a retrieval was successful.
Bytes
This indicates the size of the file retrieved. This is important because many ISPs charge a Web site based on traffic activity as measured in bytes.
Browser and Platform
Other data in this log shows which browser and computing platform were used by the visitor.
Referrer URL
This file usually resides in your Web site log directory along with your activity file (common log file or other). If you don't have a referrer log, you will know because their will be no field in any activity log that indicates where visitors came from. Since this is one of the most crucial items of information for any Web site, you should ask your ISP to create a referrer log file if you don't have one. The set-up cost should be minimal, and you can utilize the raw file by viewing it in your word processor, although reporting software will help make it easier to follow. Often, the referring URL will be a search engine or directory. In this case, the field will also show what "keyword phrase" the visitor used to find your site in the search engine listings. Use this information to fine-tune the keyword usage in your home page.
'140.161.64.126','Student-5-2.Lab-4247.Douglas.BC.CA','http://Webcra wler.com/select/market.09.html','Mozilla/3.0 (Win16; I)','859944487'
'195.37.0.67','ext3.fh-brandenburg.de','http://www.botspot.com/newbots/ newbots4.htm','Mozilla/3.01Gold (Win95; I)','859944829'
'139.92.89.97','slip139-92-89-97.tel.il.ibm.net','(null)','Mozilla/2.0 (Win95; I; 16bit)','859944864'
'204.32.199.230','ftl-fl11-06.ix.netcom.com','http://www.exposure-usa.co m/exposure/970321.html','Mozilla/2.01E-NC250 (Win95; U; 16bit)','859944993'
'204.32.166.173','chi-il11-13.ix.netcom.com','http://personal.netscape.co m/custom/page/show_page.html','Mozilla/3.01 (Win95; I)','859945061'
This is a custom referrer URL log which provides a line showing each visitor to a Web site, their host IP and RFC name if available, the site they were referred from (any site which offers a link to your site) and the browser and computing platform. If the field before browser and computing platform (in our log, it's the third field) reports "NULL" instead of a URL, it usually means that your visitor typed in your URL directly or has it saved in their bookmark or hotlist file. Knowing the difference between a visitor who was "referred" by another site, and a visitor who came on their own will help you in your effort to determine new visitors vs. repeat visitors. "NULL" in the referrer field could also mean that the visitor typed in your URL from a printed listing in a magazine or newspaper.
Activity logs can show information that a Web advertising site may not want a potential advertiser to know. This could include high levels of activity from students and other visitors who are not prospects for the advertiser's product, short attention span (where a site gets many visitors who spend very little time at the site, which suggests a low interest level), and visits by "bots" which count an impression that was not actually viewed by a human. Good activity tracking software is available to provide a potential advertiser with everything they need to know to make a wise advertising decision.Advertising measurement is performed primarily by using tracking software and server activity logs, which meter, but understate, impressions. This undercounting can be caused by "caching", "site mirroring" and "firewalls." With caching and site mirroring, the page displaying the ad is sometimes delivered once, then stored locally on the visitor's computer or the ISP's computer. When the page is viewed later -- by the same visitor or other visitors -- it is retrieved from local memory rather than from the Web site. It's faster for surfers, but prevents the tracking software from counting it as an impression delivered from the Web server.
A firewall with proxy server is used by corporate intranets for security, but it also can cause impressions to drop because it's possible for 10 or 10,000 corporate employees to be identified by the same corporate visitor address.
Tracking software can't measure the number of unique visitors so statistics, surveys, and subscription information is used to supplement the estimates. It's conceivable that a single visitor, who visited a Web site twice per hour, every hour, for a month, could register 1,440 impressions according to the tracking software.
An undercounting twist to this scenario: many ISPs assign the same IP string to each new user as the previous user logs off. So, for example, a new Chicago-area visitor from the Netcom service who visits a Web site could be identified as chi-il11-13.ix.netcom.com, even if a previous Netcom user with the same temporarily assigned IP string just left the same site minutes ago. Proxy servers and local caching also hide portions of IP strings from the server log access files. So, during a very busy time period, more than one unique visitor could visit a site with the same reported IP string, and the log will consider it a single visitor.A solution that many sites use to correct their traffic measurement is to set a cookie that uniquely identifies each visitor. This is assigned, and stored in the browser's cookie file the first time you visit the site. On subsequent visits, the unique cookie helps identify the visitor in conjunction with the IP address. Cookies are only partially helpful in identifying unique visitors because they can be turned off or the visitor may use different computers and browsers with differing cookies, which makes them a less than perfect confirmation tool.
An intelligent agent, or "bot," could automatically visit a site every hour and inflate a visitor count. Variations on intelligent agents include "Spiders" and "Crawlers," which are software programs that visit virtually every page on the Web to create indexes for search engines, but cause overcounting of traffic in the process. Spiders and crawlers are much more interested in text files than graphics files such as banner ads. If the tracking software is counting requests for a text page, rather than the ad banner, this could result in overcounting actual ad impressions. Some flawed spiders and crawlers continually request the same text page.
Virtual includes and filters can also cause misrepresentation with visitor counts. "Virtual includes" are techniques that are used to capture only a desired image, such as a weather map, without a user having to visit a page and view banner advertising. Filter software such as "Junkbuster" enables a user to strip banners and cookies from accessed pages in order to speed up page delivery. Virtual includes and filters cause a reduction of delivered ad impressions but not necessarily accurate tracking of the reduction.
Browser bugs are constantly changing but are constantly present. A browser bug can cause a request for the text portion of a page twice, but the graphics only once. The impact on the ad impression count depends on the tracking software.
Rogue bots are designed to deliberately inflate the traffic count on CGI-generated in-line GIF advertising images. These graphics are usually only displayed, and counted, when a specific Web page has a legitimate visitor as determined by the http_referer field sent from the client to the server during http transactions. A rogue bot acts as a "visitor" to a web site and sends a fictitious http_referer and a fake IP address. Rogue bots are rare.
Internet bottlenecks can cause the visitor to request the same page repeatedly until the full page with graphics is actually displayed. But the traffic counting software is counting all requests for the full page, thereby overcounting impressions.
Bottlenecks affect larger banner ad graphics more than smaller banner ad graphics. If a banner ad is large, it takes longer to load, giving the visitor a chance to move on to another page or site before it is displayed - yet the impression is still counted. But larger banner ads generally deliver higher click-throughs to the advertiser's site, so there is a trade-off. Smaller banner ads may load faster, ensuring the visitor sees them, but they may be too small to capture the visitor's interest.
Some Web users choose to run their browsers with "graphics off" to speed up delivery of pages. For these users, a text advertising message is delivered instead of a graphic banner ad. Advertisers are more frequently demanding that they only pay for ad impressions when the actual banner ad graphic is fully loaded, not just the alternative text version. Industry estimates place "graphics off" browser usage at 6 percent. Requiring the visitor to navigate a site using only graphics and not text links can significantly improve this problem.
"Cache-Busting" refers to the ability by ad management systems such as MatchLogic and Imgis to "bust" cached banner ads (to "view" inside a cache and count the number of ads delivered from the cache rather than from a server). This improves accuracy of reporting delivered advertising impressions. These cached ads previously were not counted. A controversy arises because historically, no advertiser had to pay for cached ad impressions, they were a "bonus." Now, the same advertiser finds that they may have to pay more to advertise on the same site, because the "bonus" ads are now being counted and charged for.
Some web sites are now also offering "cost per click" or "cost per click through," which is a direct measurement of response and ad performance. A click through occurs when the visitor sees or reads the ad and clicks on it, taking them directly to the advertiser's Web site. Sometimes a visitor will click, but quit before the other Web site is displayed. Click through averages anywhere from 1 to 25% of total impressions, but generally 1 to 3% is used as a rule of thumb. The overall average for click-throughs is falling to around 1 percent. Web sites try to avoid selling advertising this way because poorly designed ads won't be clicked on, reducing ad revenue. Numerous other advertising models require performance results before the Web advertising site is paid. Examples include banner ads sold based on percentage of sales, which has been done by Nabisco, and based on a bounty for every software program download, as has been done by Microsoft.The common log format is the standard for most Web servers where your Web site might reside. It exists as a text file in the log directory and can be viewed on-screen or downloaded by anyone who has FTP or Telnet access to the server files. Numerous log formats exist, each of which can be used by themselves to analyze Web site traffic. Traffic analysis software then takes this raw data and plots it for easier reading. The common log will report the following information on every visitor and what activity they performed on your site:
host/ip
• RFC name
• logname
• datestamp
• retrieval
• code
• bytes.
They look like this:
unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:14 -0800] "GET /login_on.gif HTTP/1.0" 200 8452 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:15 -0800] "GET /ad.gif HTTP/1.0" 200 22443 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)"
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:20 -0800] "GET /bg.jpg HTTP/1.0" 200 4377 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:12 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/3.0 (Win95; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:45 -0800] "GET /ord.com HTTP/1.0" 404 240 "-" "Mozilla/3.01 (Win16; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:14 -0800] "GET / HTTP/1.0" 200 7047 "http://www.netgambit.com/" "Mozilla/3.01 (Win16; I
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:20 -0800] "GET /links.html HTTP/1.0" 200 7404 "http://www.positionagent.com/" "Mozilla/3.01 (Win16; I)
" burn-mx0508707.smartt.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:02:05 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/4.0b2 (Win95; I)
Host/IP
The first field in the record is the host/IP. It indicates the hostname, or RFC name, of the visitor or an IP number if domain name server lookup is not enabled for that visitor. The example above shows the activity of three host/ip visitors. Visitors can have .com, .edu, .net, .uk and other extensions, indicating whether they originate from a company, an ISP, a university, or another country.
Datestamp
The next important field is the datestamp, which is useful when running the log figures through analysis software. The datestamp information can be used to graph peak activity throughout the day, week and month. This will show you whether traffic picked up after your banner ad started appearing.
Retrieval
The retrieval method is posted. GET /links.html indicates that the link resource page was retrieved by the visitor. Most visitors retrieve multiple files which make up a single Web page. Some counters would indicate three hits to the Web site. Others would indicate a single HTML page hit. If a banner ad also appeared on the page and was retrieved with the other three files, an appropriate counting device would register an impression.
Code
The code field indicates whether a retrieval was successful.
Bytes
This indicates the size of the file retrieved. This is important because many ISPs charge a Web site based on traffic activity as measured in bytes.
Browser and Platform
Other data in this log shows which browser and computing platform were used by the visitor.
Referrer URL
This file usually resides in your Web site log directory along with your activity file (common log file or other). If you don't have a referrer log, you will know because their will be no field in any activity log that indicates where visitors came from. Since this is one of the most crucial items of information for any Web site, you should ask your ISP to create a referrer log file if you don't have one. The set-up cost should be minimal, and you can utilize the raw file by viewing it in your word processor, although reporting software will help make it easier to follow. Often, the referring URL will be a search engine or directory. In this case, the field will also show what "keyword phrase" the visitor used to find your site in the search engine listings. Use this information to fine-tune the keyword usage in your home page.
'140.161.64.126','Student-5-2.Lab-4247.Douglas.BC.CA','http://Webcra wler.com/select/market.09.html','Mozilla/3.0 (Win16; I)','859944487'
'195.37.0.67','ext3.fh-brandenburg.de','http://www.botspot.com/newbots/ newbots4.htm','Mozilla/3.01Gold (Win95; I)','859944829'
'139.92.89.97','slip139-92-89-97.tel.il.ibm.net','(null)','Mozilla/2.0 (Win95; I; 16bit)','859944864'
'204.32.199.230','ftl-fl11-06.ix.netcom.com','http://www.exposure-usa.co m/exposure/970321.html','Mozilla/2.01E-NC250 (Win95; U; 16bit)','859944993'
'204.32.166.173','chi-il11-13.ix.netcom.com','http://personal.netscape.co m/custom/page/show_page.html','Mozilla/3.01 (Win95; I)','859945061'
This is a custom referrer URL log which provides a line showing each visitor to a Web site, their host IP and RFC name if available, the site they were referred from (any site which offers a link to your site) and the browser and computing platform. If the field before browser and computing platform (in our log, it's the third field) reports "NULL" instead of a URL, it usually means that your visitor typed in your URL directly or has it saved in their bookmark or hotlist file. Knowing the difference between a visitor who was "referred" by another site, and a visitor who came on their own will help you in your effort to determine new visitors vs. repeat visitors. "NULL" in the referrer field could also mean that the visitor typed in your URL from a printed listing in a magazine or newspaper.
Activity logs can show information that a Web advertising site may not want a potential advertiser to know. This could include high levels of activity from students and other visitors who are not prospects for the advertiser's product, short attention span (where a site gets many visitors who spend very little time at the site, which suggests a low interest level), and visits by "bots" which count an impression that was not actually viewed by a human. Good activity tracking software is available to provide a potential advertiser with everything they need to know to make a wise advertising decision.
Web advertising is typically sold on a "Cost Per Thousand Impressions" basis, or CPM. An impression (also referred to as an "exposure" or "page view") occurs when a visitor to a Web site views a page where an ad is displayed, whether the ad is seen or not. Whenever a page is "served" to your computer screen, measurement software counts the "impression."
Ads can be placed anywhere on a page, although ads are more successful when they are displayed near the top of the page("above the fold"). This ad will deliver higher click-throughs than the same ad placed at the bottom of a page. Some sites use frames to display advertising more prominently.
A common banner ad package consists of 100,000 impressions costing anywhere from $20 to $100 per thousand ($25-$70 CPM is average for popular sites). Keep in mind that most sites have repeat visitors, and most visitors view more than a single page of information, so your 100,000 impressions could actually represent 10,000 to 40,000 unique visitors. In some cases, impressions are guaranteed, in others, estimated. Some sites will charge a flat rate for estimated impressions. If the actual delivered impression count is higher, some will let you keep the "overdeliveries" for no extra charge. Most popular sites offer ad agency discounts and volume or frequency discounts. The ad agency rate, known as the net rate, is typically a 15% discount from the gross rate, which is the typical rate if you purchase advertising directly instead of through an agency.
"Hits" and "page hits" are not favored as measurement units because a hit is registered every time any text or graphic file is delivered, whether advertising is displayed or not. A page hit, or HTML hit, is registered every time a complete HTML page, including text and graphic files is delivered. The ratio can easily be 10 or more hits or 2-4 page hits to actual impressions.
Advertising measurement is performed primarily by using tracking software and server activity logs, which meter, but understate, impressions. This undercounting can be caused by "caching", "site mirroring" and "firewalls." With caching and site mirroring, the page displaying the ad is sometimes delivered once, then stored locally on the visitor's computer or the ISP's computer. When the page is viewed later -- by the same visitor or other visitors -- it is retrieved from local memory rather than from the Web site. It's faster for surfers, but prevents the tracking software from counting it as an impression delivered from the Web server.
A firewall with proxy server is used by corporate intranets for security, but it also can cause impressions to drop because it's possible for 10 or 10,000 corporate employees to be identified by the same corporate visitor address.
Tracking software can't measure the number of unique visitors so statistics, surveys, and subscription information is used to supplement the estimates. It's conceivable that a single visitor, who visited a Web site twice per hour, every hour, for a month, could register 1,440 impressions according to the tracking software.
An undercounting twist to this scenario: many ISPs assign the same IP string to each new user as the previous user logs off. So, for example, a new Chicago-area visitor from the Netcom service who visits a Web site could be identified as chi-il11-13.ix.netcom.com, even if a previous Netcom user with the same temporarily assigned IP string just left the same site minutes ago. Proxy servers and local caching also hide portions of IP strings from the server log access files. So, during a very busy time period, more than one unique visitor could visit a site with the same reported IP string, and the log will consider it a single visitor.A solution that many sites use to correct their traffic measurement is to set a cookie that uniquely identifies each visitor. This is assigned, and stored in the browser's cookie file the first time you visit the site. On subsequent visits, the unique cookie helps identify the visitor in conjunction with the IP address. Cookies are only partially helpful in identifying unique visitors because they can be turned off or the visitor may use different computers and browsers with differing cookies, which makes them a less than perfect confirmation tool.
An intelligent agent, or "bot," could automatically visit a site every hour and inflate a visitor count. Variations on intelligent agents include "Spiders" and "Crawlers," which are software programs that visit virtually every page on the Web to create indexes for search engines, but cause overcounting of traffic in the process. Spiders and crawlers are much more interested in text files than graphics files such as banner ads. If the tracking software is counting requests for a text page, rather than the ad banner, this could result in overcounting actual ad impressions. Some flawed spiders and crawlers continually request the same text page.
Virtual includes and filters can also cause misrepresentation with visitor counts. "Virtual includes" are techniques that are used to capture only a desired image, such as a weather map, without a user having to visit a page and view banner advertising. Filter software such as "Junkbuster" enables a user to strip banners and cookies from accessed pages in order to speed up page delivery. Virtual includes and filters cause a reduction of delivered ad impressions but not necessarily accurate tracking of the reduction.
Browser bugs are constantly changing but are constantly present. A browser bug can cause a request for the text portion of a page twice, but the graphics only once. The impact on the ad impression count depends on the tracking software.
Rogue bots are designed to deliberately inflate the traffic count on CGI-generated in-line GIF advertising images. These graphics are usually only displayed, and counted, when a specific Web page has a legitimate visitor as determined by the http_referer field sent from the client to the server during http transactions. A rogue bot acts as a "visitor" to a web site and sends a fictitious http_referer and a fake IP address. Rogue bots are rare.
Internet bottlenecks can cause the visitor to request the same page repeatedly until the full page with graphics is actually displayed. But the traffic counting software is counting all requests for the full page, thereby overcounting impressions.
Bottlenecks affect larger banner ad graphics more than smaller banner ad graphics. If a banner ad is large, it takes longer to load, giving the visitor a chance to move on to another page or site before it is displayed - yet the impression is still counted. But larger banner ads generally deliver higher click-throughs to the advertiser's site, so there is a trade-off. Smaller banner ads may load faster, ensuring the visitor sees them, but they may be too small to capture the visitor's interest.
Some Web users choose to run their browsers with "graphics off" to speed up delivery of pages. For these users, a text advertising message is delivered instead of a graphic banner ad. Advertisers are more frequently demanding that they only pay for ad impressions when the actual banner ad graphic is fully loaded, not just the alternative text version. Industry estimates place "graphics off" browser usage at 6 percent. Requiring the visitor to navigate a site using only graphics and not text links can significantly improve this problem.
"Cache-Busting" refers to the ability by ad management systems such as MatchLogic and Imgis to "bust" cached banner ads (to "view" inside a cache and count the number of ads delivered from the cache rather than from a server). This improves accuracy of reporting delivered advertising impressions. These cached ads previously were not counted. A controversy arises because historically, no advertiser had to pay for cached ad impressions, they were a "bonus." Now, the same advertiser finds that they may have to pay more to advertise on the same site, because the "bonus" ads are now being counted and charged for.
Some web sites are now also offering "cost per click" or "cost per click through," which is a direct measurement of response and ad performance. A click through occurs when the visitor sees or reads the ad and clicks on it, taking them directly to the advertiser's Web site. Sometimes a visitor will click, but quit before the other Web site is displayed. Click through averages anywhere from 1 to 25% of total impressions, but generally 1 to 3% is used as a rule of thumb. The overall average for click-throughs is falling to around 1 percent. Web sites try to avoid selling advertising this way because poorly designed ads won't be clicked on, reducing ad revenue. Numerous other advertising models require performance results before the Web advertising site is paid. Examples include banner ads sold based on percentage of sales, which has been done by Nabisco, and based on a bounty for every software program download, as has been done by Microsoft.The common log format is the standard for most Web servers where your Web site might reside. It exists as a text file in the log directory and can be viewed on-screen or downloaded by anyone who has FTP or Telnet access to the server files. Numerous log formats exist, each of which can be used by themselves to analyze Web site traffic. Traffic analysis software then takes this raw data and plots it for easier reading. The common log will report the following information on every visitor and what activity they performed on your site:
host/ip
• RFC name
• logname
• datestamp
• retrieval
• code
• bytes.
They look like this:
unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:14 -0800] "GET /login_on.gif HTTP/1.0" 200 8452 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:15 -0800] "GET /ad.gif HTTP/1.0" 200 22443 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)"
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:20 -0800] "GET /bg.jpg HTTP/1.0" 200 4377 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:12 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/3.0 (Win95; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:45 -0800] "GET /ord.com HTTP/1.0" 404 240 "-" "Mozilla/3.01 (Win16; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:14 -0800] "GET / HTTP/1.0" 200 7047 "http://www.netgambit.com/" "Mozilla/3.01 (Win16; I
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:20 -0800] "GET /links.html HTTP/1.0" 200 7404 "http://www.positionagent.com/" "Mozilla/3.01 (Win16; I)
" burn-mx0508707.smartt.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:02:05 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/4.0b2 (Win95; I)
Host/IP
The first field in the record is the host/IP. It indicates the hostname, or RFC name, of the visitor or an IP number if domain nameserver lookup is not enabled for that visitor. The example above shows the activity of three host/ip visitors. Visitors can have .com, .edu, .net, .uk and other extensions, indicating whether they originate from a company, an ISP, a university, or another country.
Datestamp
The next important field is the datestamp, which is useful when running the log figures through analysis software. The datestamp information can be used to graph peak activity throughout the day, week and month. This will show you whether traffic picked up after your banner ad started appearing.
Retrieval
The retrieval method is posted. GET /links.html indicates that the link resource page was retrieved by the visitor. Most visitors retrieve multiple files which make up a single Web page. Some counters would indicate three hits to the Web site. Others would indicate a single HTML page hit. If a banner ad also appeared on the page and was retrieved with the other three files, an appropriate counting device would register an impression.
Code
The code field indicates whether a retrieval was successful.
Bytes
This indicates the size of the file retrieved. This is important because many ISPs charge a Web site based on traffic activity as measured in bytes.
Browser and Platform
Other data in this log shows which browser and computing platform were used by the visitor.
Referrer URL
This file usually resides in your Web site log directory along with your activity file (common log file or other). If you don't have a referrer log, you will know because their will be no field in any activity log that indicates where visitors came from. Since this is one of the most crucial items of information for any Web site, you should ask your ISP to create a referrer log file if you don't have one. The set-up cost should be minimal, and you can utilize the raw file by viewing it in your word processor, although reporting software will help make it easier to follow. Often, the referring URL will be a search engine or directory. In this case, the field will also show what "keyword phrase" the visitor used to find your site in the search engine listings. Use this information to fine-tune the keyword usage in your home page.
'140.161.64.126','Student-5-2.Lab-4247.Douglas.BC.CA','http://Webcra wler.com/select/market.09.html','Mozilla/3.0 (Win16; I)','859944487'
'195.37.0.67','ext3.fh-brandenburg.de','http://www.botspot.com/newbots/ newbots4.htm','Mozilla/3.01Gold (Win95; I)','859944829'
'139.92.89.97','slip139-92-89-97.tel.il.ibm.net','(null)','Mozilla/2.0 (Win95; I; 16bit)','859944864'
'204.32.199.230','ftl-fl11-06.ix.netcom.com','http://www.exposure-usa.co m/exposure/970321.html','Mozilla/2.01E-NC250 (Win95; U; 16bit)','859944993'
'204.32.166.173','chi-il11-13.ix.netcom.com','http://personal.netscape.co m/custom/page/show_page.html','Mozilla/3.01 (Win95; I)','859945061'
This is a custom referrer URL log which provides a line showing each visitor to a Web site, their host IP and RFC name if available, the site they were referred from (any site which offers a link to your site) and the browser and computing platform. If the field before browser and computing platform (in our log, it's the third field) reports "NULL" instead of a URL, it usually means that your visitor typed in your URL directly or has it saved in their bookmark or hotlist file. Knowing the difference between a visitor who was "referred" by another site, and a visitor who came on their own will help you in your effort to determine new visitors vs. repeat visitors. "NULL" in the referrer field could also mean that the visitor typed in your URL from a printed listing in a magazine or newspaper.
Activity logs can show information that a Web advertising site may not want a potential advertiser to know. This could include high levels of activity from students and other visitors who are not prospects for the advertiser's product, short attention span (where a site gets many visitors who spend very little time at the site, which suggests a low interest level), and visits by "bots" which count an impression that was not actually viewed by a human. Good activity tracking software is available to provide a potential advertiser with everything they need to know to make a wise advertising decision.Advertising measurement is performed primarily by using tracking software and server activity logs, which meter, but understate, impressions. This undercounting can be caused by "caching", "site mirroring" and "firewalls." With caching and site mirroring, the page displaying the ad is sometimes delivered once, then stored locally on the visitor's computer or the ISP's computer. When the page is viewed later -- by the same visitor or other visitors -- it is retrieved from local memory rather than from the Web site. It's faster for surfers, but prevents the tracking software from counting it as an impression delivered from the Web server.
A firewall with proxy server is used by corporate intranets for security, but it also can cause impressions to drop because it's possible for 10 or 10,000 corporate employees to be identified by the same corporate visitor address.
Tracking software can't measure the number of unique visitors so statistics, surveys, and subscription information is used to supplement the estimates. It's conceivable that a single visitor, who visited a Web site twice per hour, every hour, for a month, could register 1,440 impressions according to the tracking software.
An undercounting twist to this scenario: many ISPs assign the same IP string to each new user as the previous user logs off. So, for example, a new Chicago-area visitor from the Netcom service who visits a Web site could be identified as chi-il11-13.ix.netcom.com, even if a previous Netcom user with the same temporarily assigned IP string just left the same site minutes ago. Proxy servers and local caching also hide portions of IP strings from the server log access files. So, during a very busy time period, more than one unique visitor could visit a site with the same reported IP string, and the log will consider it a single visitor.A solution that many sites use to correct their traffic measurement is to set a cookie that uniquely identifies each visitor. This is assigned, and stored in the browser's cookie file the first time you visit the site. On subsequent visits, the unique cookie helps identify the visitor in conjunction with the IP address. Cookies are only partially helpful in identifying unique visitors because they can be turned off or the visitor may use different computers and browsers with differing cookies, which makes them a less than perfect confirmation tool.
An intelligent agent, or "bot," could automatically visit a site every hour and inflate a visitor count. Variations on intelligent agents include "Spiders" and "Crawlers," which are software programs that visit virtually every page on the Web to create indexes for search engines, but cause overcounting of traffic in the process. Spiders and crawlers are much more interested in text files than graphics files such as banner ads. If the tracking software is counting requests for a text page, rather than the ad banner, this could result in overcounting actual ad impressions. Some flawed spiders and crawlers continually request the same text page.
Virtual includes and filters can also cause misrepresentation with visitor counts. "Virtual includes" are techniques that are used to capture only a desired image, such as a weather map, without a user having to visit a page and view banner advertising. Filter software such as "Junkbuster" enables a user to strip banners and cookies from accessed pages in order to speed up page delivery. Virtual includes and filters cause a reduction of delivered ad impressions but not necessarily accurate tracking of the reduction.
Browser bugs are constantly changing but are constantly present. A browser bug can cause a request for the text portion of a page twice, but the graphics only once. The impact on the ad impression count depends on the tracking software.
Rogue bots are designed to deliberately inflate the traffic count on CGI-generated in-line GIF advertising images. These graphics are usually only displayed, and counted, when a specific Web page has a legitimate visitor as determined by the http_referer field sent from the client to the server during http transactions. A rogue bot acts as a "visitor" to a web site and sends a fictitious http_referer and a fake IP address. Rogue bots are rare.
Internet bottlenecks can cause the visitor to request the same page repeatedly until the full page with graphics is actually displayed. But the traffic counting software is counting all requests for the full page, thereby overcounting impressions.
Bottlenecks affect larger banner ad graphics more than smaller banner ad graphics. If a banner ad is large, it takes longer to load, giving the visitor a chance to move on to another page or site before it is displayed - yet the impression is still counted. But larger banner ads generally deliver higher click-throughs to the advertiser's site, so there is a trade-off. Smaller banner ads may load faster, ensuring the visitor sees them, but they may be too small to capture the visitor's interest.
Some Web users choose to run their browsers with "graphics off" to speed up delivery of pages. For these users, a text advertising message is delivered instead of a graphic banner ad. Advertisers are more frequently demanding that they only pay for ad impressions when the actual banner ad graphic is fully loaded, not just the alternative text version. Industry estimates place "graphics off" browser usage at 6 percent. Requiring the visitor to navigate a site using only graphics and not text links can significantly improve this problem.
"Cache-Busting" refers to the ability by ad management systems such as MatchLogic and Imgis to "bust" cached banner ads (to "view" inside a cache and count the number of ads delivered from the cache rather than from a server). This improves accuracy of reporting delivered advertising impressions. These cached ads previously were not counted. A controversy arises because historically, no advertiser had to pay for cached ad impressions, they were a "bonus." Now, the same advertiser finds that they may have to pay more to advertise on the same site, because the "bonus" ads are now being counted and charged for.
Some web sites are now also offering "cost per click" or "cost per click through," which is a direct measurement of response and ad performance. A click through occurs when the visitor sees or reads the ad and clicks on it, taking them directly to the advertiser's Web site. Sometimes a visitor will click, but quit before the other Web site is displayed. Click through averages anywhere from 1 to 25% of total impressions, but generally 1 to 3% is used as a rule of thumb. The overall average for click-throughs is falling to around 1 percent. Web sites try to avoid selling advertising this way because poorly designed ads won't be clicked on, reducing ad revenue. Numerous other advertising models require performance results before the Web advertising site is paid. Examples include banner ads sold based on percentage of sales, which has been done by Nabisco, and based on a bounty for every software program download, as has been done by Microsoft.The common log format is the standard for most Web servers where your Web site might reside. It exists as a text file in the log directory and can be viewed on-screen or downloaded by anyone who has FTP or Telnet access to the server files. Numerous log formats exist, each of which can be used by themselves to analyze Web site traffic. Traffic analysis software then takes this raw data and plots it for easier reading. The common log will report the following information on every visitor and what activity they performed on your site:
host/ip
• RFC name
• logname
• datestamp
• retrieval
• code
• bytes.
They look like this:
unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:14 -0800] "GET /login_on.gif HTTP/1.0" 200 8452 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:15 -0800] "GET /ad.gif HTTP/1.0" 200 22443 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)"
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:53:20 -0800] "GET /bg.jpg HTTP/1.0" 200 4377 "http://www.positionagent.com/" "Mozilla/3.0 (Win95; I)
" unknown-23-147.pilot.net /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:12 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/3.0 (Win95; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:22:56:45 -0800] "GET /ord.com HTTP/1.0" 404 240 "-" "Mozilla/3.01 (Win16; I)
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:14 -0800] "GET / HTTP/1.0" 200 7047 "http://www.netgambit.com/" "Mozilla/3.01 (Win16; I
" irv-ca28-44.ix.netcom.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:00:20 -0800] "GET /links.html HTTP/1.0" 200 7404 "http://www.positionagent.com/" "Mozilla/3.01 (Win16; I)
" burn-mx0508707.smartt.com /Webadmin/home/pub.d/delph/public_html - [02/Apr/1997:23:02:05 -0800] "GET / HTTP/1.0" 200 7047 "-" "Mozilla/4.0b2 (Win95; I)
Host/IP
The first field in the record is the host/IP. It indicates the hostname, or RFC name, of the visitor or an IP number if domain name server lookup is not enabled for that visitor. The example above shows the activity of three host/ip visitors. Visitors can have .com, .edu, .net, .uk and other extensions, indicating whether they originate from a company, an ISP, a university, or another country.
Datestamp
The next important field is the datestamp, which is useful when running the log figures through analysis software. The datestamp information can be used to graph peak activity throughout the day, week and month. This will show you whether traffic picked up after your banner ad started appearing.
Retrieval
The retrieval method is posted. GET /links.html indicates that the link resource page was retrieved by the visitor. Most visitors retrieve multiple files which make up a single Web page. Some counters would indicate three hits to the Web site. Others would indicate a single HTML page hit. If a banner ad also appeared on the page and was retrieved with the other three files, an appropriate counting device would register an impression.
Code
The code field indicates whether a retrieval was successful.
Bytes
This indicates the size of the file retrieved. This is important because many ISPs charge a Web site based on traffic activity as measured in bytes.
Browser and Platform
Other data in this log shows which browser and computing platform were used by the visitor.
Referrer URL
This file usually resides in your Web site log directory along with your activity file (common log file or other). If you don't have a referrer log, you will know because their will be no field in any activity log that indicates where visitors came from. Since this is one of the most crucial items of information for any Web site, you should ask your ISP to create a referrer log file if you don't have one. The set-up cost should be minimal, and you can utilize the raw file by viewing it in your word processor, although reporting software will help make it easier to follow. Often, the referring URL will be a search engine or directory. In this case, the field will also show what "keyword phrase" the visitor used to find your site in the search engine listings. Use this information to fine-tune the keyword usage in your home page.
'140.161.64.126','Student-5-2.Lab-4247.Douglas.BC.CA','http://Webcra wler.com/select/market.09.html','Mozilla/3.0 (Win16; I)','859944487'
'195.37.0.67','ext3.fh-brandenburg.de','http://www.botspot.com/newbots/ newbots4.htm','Mozilla/3.01Gold (Win95; I)','859944829'
'139.92.89.97','slip139-92-89-97.tel.il.ibm.net','(null)','Mozilla/2.0 (Win95; I; 16bit)','859944864'
'204.32.199.230','ftl-fl11-06.ix.netcom.com','http://www.exposure-usa.co m/exposure/970321.html','Mozilla/2.01E-NC250 (Win95; U; 16bit)','859944993'
'204.32.166.173','chi-il11-13.ix.netcom.com','http://personal.netscape.co m/custom/page/show_page.html','Mozilla/3.01 (Win95; I)','859945061'
This is a custom referrer URL log which provides a line showing each visitor to a Web site, their host IP and RFC name if available, the site they were referred from (any site which offers a link to your site) and the browser and computing platform. If the field before browser and computing platform (in our log, it's the third field) reports "NULL" instead of a URL, it usually means that your visitor typed in your URL directly or has it saved in their bookmark or hotlist file. Knowing the difference between a visitor who was "referred" by another site, and a visitor who came on their own will help you in your effort to determine new visitors vs. repeat visitors. "NULL" in the referrer field could also mean that the visitor typed in your URL from a printed listing in a magazine or newspaper.
Activity logs can show information that a Web advertising site may not want a potential advertiser to know. This could include high levels of activity from students and other visitors who are not prospects for the advertiser's product, short attention span (where a site gets many visitors who spend very little time at the site, which suggests a low interest level), and visits by "bots" which count an impression that was not actually viewed by a human. Good activity tracking software is available to provide a potential advertiser with everything they need to know to make a wise advertising decision.
Comments:
Post a Comment
