Databases Reference
In-Depth Information
Valid Clicks
(~24%)
Visits that do not result in a
click on a sponsored link
(~70%)
Invalid Clicks
(~6%)
Identifiable Void
and
Identifiable
Fraudulent Clicks
(~5%)
Unidentifiable Void
Clicks (~0%)
Unidentifiable Fraudulent
Clicks (~1%)
Figure 7.2. The click space when discussing click fraud.
aggregate level and the individual level.
Figure 7.2
illustrates the relationship among
valid clicks
,
invalid clicks
, and
overall visits
, based on previously reported data.
If we view the entire space of visits to the search engine (i.e., the click space) as
100 percent, we know that at least 70 percent of these visits do not result in a click on
a sponsored link. This figure is based on a number of user studies and popular press
reports. Therefore, valid and invalid clicks on sponsored links make up about 30
percent of all visits. A variety of popular press reports [
58
] had reported that search
engines are screening about 15 percent of invalid clicks (both void and fraudulent
clicks) or approximately 5 percent of all clicks.
The accuracy of search engines in filtering these invalid clicks is not precisely
known; however, we can reasonably estimate that it is 80 percent or higher. Taking
the conservative end with 80 percent accuracy, this leads to the conclusion that inva-
lid clicks make up 6 percent (i.e., 20 percent of 30 percent) of all clicks on sponsored
links. Assuming that the unidentified void clicks are nearly zero, this means that
slightly more than 1 percent of fraudulent clicks go undetected (i.e., the slip-through
rate). Although a low percentage, this can translate into real cash when one considers
that the search engines are earning billions of dollars per year (e.g., if a search engine
earns $8 billion, one percent translates into $80 million).
These percentages are only for sponsored links off the SERP. There are no com-
parable numbers for sponsored links from contextual sites (i.e., where search engines
display sponsored links on specific Web sites). However, one suspects that these inci-
dents of click fraud are much higher. Complaints of invalid clicks and poor traffic
from content providers seem to substantiate this higher rate of fraudulent clicks on
these contextual ads [
59
]
