Call (877) SITE-OPT (748-3678)

The Cost of Banner Advertising on Web Performance

Summary: The cost of banner ads for the average website is about one-sixth more objects and one-third more latency. Learn how online advertising affects web page response times, user behavior, and satisfaction.

Advertising is a multi-billion dollar industry on the Web. The Internet Advertising Bureau reported that web advertising grew by nearly 34 percent in 2006 over 2005, totaling about $16.8 billion (IAB 2007). The yearly growth rate of Internet ads is increasing at 28.6% per year, while the overall ad industry drags along with an increase of only 7.7% a year (Baker 2004).

To capture the limited attention spans of visitors, advertisers have increased the size of their ads. Larger ads, on the order of 360x240 pixels, have been shown to have significantly higher clickthrough rates than do the smaller ads of 468x60 pixels (Robinson, Wysocka, and Hand 2007). Advertisers have resorted to pop-up and pop-under ads to deliver their message, but these have been shown to annoy users and reduce their intention to return (McCoy et al. 2007). The number of ads for the average web page has also increased over time with the average popular page containing over 8 ads. Sponsored advertising has increased, with more than twice the share of ads for paid search as opposed to banner ads, according to projections (Laffey 2007). Off-search PPC ads appear as graphic text ads, adding to the profusion of banner ads. All of these trends mean an increase in the file size and in the number of banner ads for the average web page. This article examines the effect of banner ads on web performance. What price do we pay for banner ads on the Web?

The Effect of Banner Advertising on Web Performance

Popular sites that use advertising to generate revenue do so at a cost of about one-sixth more objects and one-third more latency (Krishnamurthy and Wills 2006). The growth in the number and size of advertisements has caused significant delays for users. In an analysis of the top 100 Alexa sites in 13 categories, Balachander Krishnamurthy and Craig Wills have found that 56% of the pages contained ads or some form of “extraneous content.” Ads accounted for about one-sixth of the objects and bytes in the average web page that has at least one advertisement. The average web page with ads was 295K in size with 48K due to advertising. It contained 52 objects, 8.1 of which were ads, served from 5.7 and 3.2 servers, respectively. Another study of the effects of online advertising found that users not subjected to ads were 11% more likely to return or to recommend the site (McCoy et al. 2007). Subjects exposed to inline ads (i.e., ads shown within a web page) remembered 3.4% more of the material than they did when exposed to pop-ups or pop-under ads. Pop-ups were found to be 24% more intrusive than in-line ads, and pop-unders were found to be 33.1% more intrusive than in-line ads.

There is a tradeoff between the number of ads displayed and the corresponding increase in download time and user frustration. More advertising leads to a greater number of servers referenced per web page. This can reduce performance with additional DNS lookups and the inability to group object retrievals over persistent HTTP connections. On average, each DNS lookup takes about 7.1 milliseconds (Bent and Voelker 2002). News sites had the highest percentage of pages with ads (85%) while reference sites had the lowest percentage (32%). Political sites had the highest number of ads per page with 25% of the external objects and 30% of the bytes due to ads.

Ad Delays by Retrieval Method

In analyzing the news category, Krishnamurthy and Wills tested how the retrieval method impacts the presence of advertising and response times. Under normal conditions (images on, JavaScript and Java on) they found that the average news page was a whopping 341K in total, with 84K worth of ads, composed of 48 objects served from 7.7 servers, with a mean download time of 10.4 seconds. Turning off JavaScript and Java cut the number of ads by 62% and total ad size by 71%. Under the same conditions, the total number of objects dropped by only 12%. This 50 percentage point difference plus the fact that no "class" objects were retrieved, show that webmasters rely mainly on JavaScript to include ads in their pages.

Turning off images and scripting cut the number of ads nearly to zero (0.5), the number of objects from 48 to 3.2, and total page size to 87K, nearly four times smaller. The median download time was cut by a factor of ten under these conditions, dropping from 10.4 to 1.3 seconds. Even though page size was reduced by a factor of four, download time was reduced by a factor of ten. You can see the influence of object overhead as the number of objects increases.

Turning off ads with Adblock cut the number of objects and bytes served by about 27%, and the median download time by 33%. These results show that ads play a significant role in increasing the number of objects per page as well as download times. Since the number of ads dropped by 62% when turning off JavaScript, webmasters may want to consider alternative ad display techniques, like XSSI.

Advertisements Distract Readers

Advertisements have been shown to be distracting to readers. One eye tracking study found that ad graphics significantly increase regressions and re-reading of web page copy (Beymerl, Orton, and Russel 2007). On-task images also slowed readers down by decreasing first pass reading speed. The researchers attribute this effect to the "extra effort the reader is making to relate the pictures to the text - the cognitive effort to relate pictures and text is slowing down the reader." The researchers recommend that to minimize distractions, e-learning sites should avoid using advertisements and perhaps relegate them to separate pages.

Conclusion

The performance cost to the average web page with banner advertisements is about one-sixth more objects and one-third more latency. Internet advertising is a multi-billion dollar industry with advertisers competing to capture the attention of users. While the percentage of banner advertisements on the Web has shrunk by more than half from 2000 to 2004 (48% to 21%, Laffey 2007) since that time online advertising spending as a percentage of total media ad spending has more than doubled from 3.6% in 2004 to 7.4% in 2007 (Hallerman 2007). So the number of banner ads is still growing, as graphic search marketing ads appear on "content networks." The average web page had 48K of banner ads in 8.1 objects. Optimizing banner ads for file size will help reduce your size overhead while restraining yourself on their number will reduce object overhead. It would be interesting to see a study testing the tradeoff between the number and size of banner ads and bailout rates.

Further Reading

Baker, S., "The Online Ad Surge: Brand Advertising Online has Taken Off - and It is Shaking Up Madison Ave,"
Business Week, Nov. 22, 2004, 76.
Bent, L., and G. Voelker, "Whole Page Performance,"
in WCW'02 (Boulder, CO: August 2002),
Beymerl, D., Orton, P., and D. Russel, "An Eye Tracking Study of How Pictures Influence Online Reading,"
in INTERACT 2007, LNCS 4663, Part II (2007), 456-460.
Hallerman, D., "eMarketer US Advertising Spending"
(New York: eMarketer, Inc., Nov. 2007), 33, http://www.emarketer.com (Jan. 31, 2008).
Interactive Advertising Bureau, "IAB/PwC Release Fourth-Quarter and FY 2006 Internet Ad Revenue Figures,"
March 7, 2007 (available at http://www.iab.net/).
Krishnamurthy, B., and C. Wills, "Cat and Mouse: Content Delivery Tradeoffs in Web Access,"
in WWW 2006 (Edinburgh, Scotland: May 23-26, 2006), 337-346
Laffey, D., "Paid search: The innovation that changed the Web,"
Business Horizons 50 (2007): 211-218.
McCoy, S., Everard, S., Polak, P., and D. Galletta, "The Effects of Online Advertising,"
Communications of the ACM 50, no. 3 (2007): 84-88.
Robinson, H., Wysocka, A., and C. Hand, "Internet advertising effectiveness - The effect of design on click-through rates for banner ads,"
International Journal of Advertising 26, no. 4 (2007): 527-541.

By website optimization on 31 Jan 2008 PM

Copyright © 2002-2013 Website Optimization, LLC. All Rights Reserved - Free website speed test - Privacy Policy
Last modified: August 26, 2013.

Follow us on: Twitter, Google+, Facebook, Linked In