JP7702473B2 - Identifying and managing cannibalized advertising to improve the effectiveness of internet advertising - Google Patents

Description

The present invention relates to online advertising technology, and in particular to the ability to identify and manage cannibalistic ad purchases within web pages.

Internet search engines return one or more web pages to a user's browser in response to a keyword search performed by a user. The returned web pages are known as "search engine results pages" or SERPs, and include unpaid search listings, commonly referred to as "organic" search results, and paid advertisements, also referred to as ads. Each listing contains a URL or a link to a World Wide Web page that is related to the search term entered by the user. The web pages that correspond to the URLs returned in the paid advertisements or unpaid listings in the SERPs are often referred to as landing pages.

The goal of online advertising is to induce users to click on paid advertisements and visit the advertiser's website. The effectiveness of an online advertising campaign is usually measured as a function of the number of clicks on paid listings, which correspond to the number of visitors to the website attributable to the advertising campaign, and the amount of revenue generated by those visitors. The effectiveness of an advertising campaign may be measured by the ratio of revenue generated to advertising costs or by the ratio of visitors to advertising costs.

However, the effectiveness of an advertising campaign can be significantly limited by cannibalizing ads, which are ads that lead to additional costs by diverting visitors to paid ads who would otherwise visit the free listing as a result of clicking on the free listing. In other words, cannibalizing ads increase costs for advertisers by diverting visitors from the free listing to the cannibalizing ads.

In its most basic form, a cannibalized ad is an ad purchased by an advertiser that appears on a web page immediately adjacent to a free listing that advertises the same service or product as the paid ad. Users sometimes click on the paid ad instead of the free listing, resulting in unnecessary advertising costs for the advertiser, assuming that in what percentage of cases the user would have clicked on the free listing. In other cases, there may be one or more ads between the advertiser's ad and the corresponding free listing. In such cases, the paid ad may still be considered cannibalized. It is therefore advantageous to identify cannibalized ads so that an advertiser can determine whether to purchase the cannibalized ad.

Because cannibalized ad spend is a new metric for determining the effectiveness of an advertising campaign, in addition to detecting cannibalized ads, it is desirable to determine the amount of money an advertiser spends on purchasing cannibalized ads. Such a value would enable an advertiser to determine the magnitude of the problem and reduce the advertiser's spending on cannibalized ads without compromising the effectiveness of the advertiser's advertising campaign.

A method, system, and device for measuring the amount of money an advertiser spends on buying cannibalized ads (ads) and the amount that could be recouped by not buying the cannibalized ads, called reclaimed ad spend.

In some embodiments, the method calculates a cannibalization score for an advertisement in a Search Engine Results Page (SERP), where the cannibalization score estimates the likelihood that a paid advertisement is cannibalized, i.e., that the paid advertisement appears near a corresponding free listing. In other embodiments, the cannibalization score estimates the reduction in revenue to an advertiser due to the paid advertisement appearing in the same SERP as the corresponding free listing.

Some embodiments are directed to generating a cannibalization score for a paid advertisement in a search engine results page (SERP) by collecting keywords associated with the advertiser, defining rules for calculating a cannibalization score for an advertisement associated with a corresponding free listing, where the cannibalization score estimates a reduction in revenue to the advertiser due to the advertisement appearing in the same SERP as the corresponding listing, providing the keywords to a search engine, receiving the SERP from the search engine, locating a first advertisement placed by the advertiser among one or more advertisements in the SERP, locating the corresponding free listing among multiple free listings in the SERP, and applying the rules to the advertisement and the corresponding free listing to obtain a cannibalization score for the advertisement.

Some embodiments are directed to generating an estimate of recovered advertising costs for an Internet advertising campaign by receiving a set of keywords, each keyword corresponding to a paid advertisement served to a search engine; collecting a cannibalization score for each paid advertisement, the cannibalization score indicating that the presence of a specified paid advertisement in the SERP reduces the likelihood that a user will click on the corresponding unpaid listing; estimating the recovered advertising cost as the difference between actual revenue reported for a period of time and an estimate of expected advertising cost for a comparable period in which non-cannibalized actions were taken; and reporting the estimated recovered advertising cost.

An embodiment is a computer-implemented method for estimating the effectiveness of an Internet advertising campaign, comprising the steps of receiving a plurality of keywords, each keyword corresponding to a paid advertisement served to a search engine, the paid advertisement including a link to a web page, and in response to receiving the keywords from the web browser, the search engine returning a search engine results page (SERP) including (1) the corresponding paid advertisements and (2) at least one unpaid listing, the unpaid listing including a link to the web page, the web page being within the domain; and (1) a cannibalization score for the specified paid advertisements, the cannibalization score indicating that the presence of the specified paid advertisements in the SERP reduces the likelihood that a user will click on the corresponding unpaid listing, and The present invention is directed to a computer-implemented method including: collecting (1) a cannibalization score for a specified paid advertisement, where the specified paid advertisement has a linked web page in the same domain as the linked web page of the specified paid advertisement; and (2) an actual revenue attributable to clicks on the specified paid advertisement and any corresponding unpaid listings in the SERP during a period in which a non-cannibalization action was not taken, where a non-cannibalization action occurs when the paid advertisement is not served to a search engine due to the cannibalization score; estimating a recovered advertising cost for the specified paid advertisement as the difference between an estimate of the expected cost to purchase the specified advertisement for a period in time and the actual cost reported during a comparable period in which a non-cannibalization action was taken; and reporting the estimated recovered advertising cost.

The invention will be more fully understood and appreciated by reading the following detailed description in conjunction with the drawings.

FIG. 1 is a simplified block diagram of a cannibalistic ad system (CAS) that automatically identifies cannibalistic ads and calculates the potential improvements that may be achieved by not purchasing such ads. FIG. 1 shows an example of cannibalizing advertising. 1 is a simplified flow diagram of an overall method for identifying cannibalized advertisements. FIG. 1 illustrates a general method for determining an effectiveness measure for an advertising campaign based on an estimate of potential savings from reducing spending on cannibalized advertising. FIG. 2 is a block diagram illustrating software modules included in the cannibalization advertising system.

The drawings depict embodiments of the present invention for purposes of illustration only. Those skilled in the art will readily recognize from the following description that alternative embodiments and combinations of the structures and methods shown herein may be employed without departing from the principles of the present invention as described herein.

The present invention will now be described more fully hereinafter with reference to the accompanying drawings, which form a part of this specification, and which show, by way of illustration, specific exemplary embodiments in which the present invention may be practiced. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, but rather these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Among other things, the present invention may be embodied as a method, process, system, business method, or device. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Thus, the following detailed description is not to be construed in a limiting sense.

The following terms used in this specification have the meanings given below.

User - means an individual who uses a mobile device, PC, or other electronic device to access the services provided by the present invention over a network.

Advertiser - refers to an individual, company, or other organization that places, or causes online advertising to be placed through a search engine, for a product or service that the advertiser is advertising, selling, or promoting.

Keyword or Search Term - refers to a word, multiple words, phrase, or sentence entered by a user into a search field in a web page, also called a keyword query, and sent to a search engine that performs the requested search and returns results. Advertisers can purchase or bid on advertisements corresponding to keywords, in which case the SERPs returned by the search engine in response to a user entering a keyword will contain paid advertisements placed by advertisers that correspond to the keyword.

Search Engine or Web Search Engine - refers to a computer server or Internet service that receives keywords, usually as the result of a keyword query, uses the keywords to search for web pages that correspond to the keywords, and returns one or more search engine results pages (SERPs) that contain paid advertisements and free or organic listings.

Listing - a result from a keyword search that appears in the SERP. Each listing contains a link to a corresponding web page. A listing can be a paid advertisement generated by a search engine, i.e., a paid listing, or a free or organic listing. Listings in the SERP are ranked, and each listing has a numerical position starting from the top position, or the first position, or the highest position. Unless otherwise specified, a listing's position in the SERP refers to the numerical position of the listing from the top of the free search results. Thus, the first position is the highest position, the second position is the next highest position, and so on. Paid listings have paid listing positions and free listings have free listing positions.

Search Engine Results Page (SERP) - refers to a list of web pages returned by a search engine in response to a keyword query. Each element in the list, i.e., each listing, typically includes a title, a URL or link to the web page, and a brief description indicating where the content within the page matches the keywords. A SERP may refer to a single web page, including a set of paid and unpaid listings, or the set of all links returned for a search query, which may span multiple web pages.

Landing Page - refers to a web page whose URL corresponds to a listing in the SERP. When a user clicks on a listing in the SERP, the web browser requests and displays the corresponding landing page.

Cannibalized Ads - refers to paid ads served in a web page by a search engine in response to a user search, which diverts clicks away from nearby free listings. In general, if a cannibalized ad appears adjacent or close to a corresponding free listing, the cannibalized ad is less valuable than if the cannibalized ad appears in the SERP and there is no corresponding free listing. In this context, value is usually measured in terms of the number of visitors, revenue from sales of the advertised product, or a similar metric. Furthermore, the corresponding free listing is a free listing that references the same product or service advertised by the paid ad. As explained below, the links in the paid ad and the free listing may reference the same landing page or different landing pages.

Generalized Operation The operation of several aspects of the present invention is described below with respect to FIGS.

Figure 1 is a simplified block diagram of a cannibalized advertising system (CAS) 100 that automatically identifies cannibalized ads and calculates the potential improvements achieved by not purchasing such ads.

A user 110 uses a web browser, referred to herein as browser 118, such as GOOGLE CHROME or Mozilla Firefox or other client application, to visit a website that allows the user to perform keyword searches. The browser 118 sends the keywords to a search engine 120, which performs the requested search and returns a (SERP), which is then displayed by the browser 118. A SERP typically includes one or more paid listings or advertisements and one or more free listings. Each advertisement or free listing corresponds to a web page, also called a landing page, that is determined by the search engine 120 to match the keyword. In the case of paid advertising, an advertiser "purchases" a keyword and provides the corresponding advertisement to an ad server 130, which communicates with the search engine 120. Thus, when a user 110 enters a keyword into a search box, the search engine returns the corresponding paid advertisements in a SERP to the browser 118.

The landing page belongs to a domain or website 145 hosted by a web server or web service simply called web server 140. Web server 140 may host multiple domains. The web pages may be static, i.e., existing as computer files in HTML format or another format, or may be dynamically generated. Additionally, web server 140 may provide electronic commerce, allowing user 110 to purchase goods or perform revenue generating transactions from website 145.

Ad server 130 provides paid advertisements to search engine 120 for inclusion in the SERPS. Cannibalized advertisement (CA) analyzer 135 analyzes the SERPS and identifies cannibalized advertisements. Generally, CA analyzer 135 generates a list of keywords that, if purchased, may result in cannibalized advertisements being placed. The operation of CA analyzer 135 is described in more detail below with reference to Figures 2-4.

It should be understood that the CA analyzer 135 may operate on a different server or computer system than the ad server 130. Additionally, the ad server 130 may be implemented as two or more physical server computers or by a cloud service such as Amazon AWS. Additionally, the CA analyzer 135 may be implemented as two or more physical server computers or by a cloud service such as Amazon AWS.

Network 150 allows the various computers, servers, and services identified in CAS 100 to exchange data. Network 150 typically refers to the public Internet, but may also refer to a private network or any combination of private and public networks.

Figure 2 is an example of a search engine results page (SERP) 200 that includes cannibalized advertising. In response to a user 110 entering the search term "Surebuy Grocery" 215 into a browser 118, a search engine 120 returns the SERP 200. The SERP 200 includes paid listings 210 and unpaid listings 220 that are advertisements for a grocery store or chain of grocery stores named "Surebuy Grocery."

Listing 210 is the first and only paid advertisement contained in SERP 200. Listing 220 is a free advertisement. Free advertisement 220 is in the position of the first free listing in SERP 200.

Advertisement 210 is a cannibalized advertisement because it is directly above free listing 220. If advertisement 210 had not appeared above free listing 220, a greater percentage of users would have clicked on free listing 220. Thus, the advertiser paid for advertisement 210 even though users would have clicked on free listing 220 if advertisement 210 had not been present in SERP 200. Thus, by placing advertisement 210 directly above free listing 220, the cost of advertising increased accordingly.

2, where a paid advertisement has a free listing for the same product or service appearing directly below it, with the paid advertisement in the first position, is considered a prime example of cannibalizing advertising. In this example, there is only one paid advertisement, the cannibalizing advertisement. However, there are other examples where an advertiser may consider an advertisement to be cannibalizing.

Several factors may be considered in determining whether paid advertising is cannibalistic. These factors include: (1) location, (2) distance, (3) occurrence rate or frequency, (4) same or different landing pages, (5) "friendly" or "competitive" (or neither) ads, and (6) the incremental or net revenue generated from the ads placed by the advertiser compared to the revenue generated by organic search results. We first describe factors 1-5 and present a rules-based approach to assess whether advertising is cannibalistic based on these factors.

Position refers to the position in the search results of a paid ad or a corresponding unpaid listing. As mentioned above, the primary case is when the paid ad and the unpaid listing are each in the first position, i.e., the paid ad is in the first position among the paid ads and the unpaid listing is in the first position among the unpaid listings. However, in some cases, a paid ad in the second or third position and a unpaid listing in the first position may be considered cannibalization.

Distance refers to the number of paid and free listings between a paid advertisement and a free listing. Distance can be determined from the location of the paid advertisement, the total number of paid advertisements, and the location of the free listing as follows:
Distance = (#PA - Position PA) + Position UL Formula 1
where #PA is the number of paid ads, position PA is the position of the paid ad being analyzed among all paid ads, and position UL is the position of the corresponding free listing among the free listings. In this example, it is understood that the paid ads appear above the free listings in a sequential order. However, a similar distance measure can be formulated when the paid ads appear at a side position or another geometric location on the web page.

Occurrence or frequency refers to the fact that the position of ads and free listings may vary from search to search. Thus, in some embodiments, search terms may be "sampled" over a period of time or over a number of iterations to determine the average position of paid ads or free listings in the SERP that are provided in response to receiving a particular keyword. For example, a search may be repeated once per minute or once per hour over the course of a day or week to obtain the position of paid ads or free listings in the received SERP. Alternatively, a search may be repeated 100 times per day. Of course, other sampling methods may be used.

In some cases, the landing page of a cannibalized ad is different from the landing page of its corresponding free listing. In some cases, such ads are considered cannibalized. In other cases, an advertiser may be testing a landing page or may simply prefer that users see their ads rather than the free listing generated by the search engine.

"Friendly" and "Competitive" Advertisements Advertisements may be categorized as "friendly" or "competitive" based on the particular advertisement placed by the advertiser. The category assigned to an advertisement may then be used as part of a determination as to whether the advertiser's advertisement is cannibalized.

Example 1: In the first example, if a first company is a business partner of a second company, the first company may agree to treat ads by the second company as "friendly" and not advertise in a competitive manner with ads placed by the second company. Thus, in this example, the presence of a friendly ad in any position among the paid ads means that the advertiser's own ads are cannibalized. Some other examples are given below.

Example 2: Automobile Dealer B sells automobiles manufactured by Automobile Manufacturer A. Manufacturer A may consider advertisements by Dealer B for Manufacturer A's products to be friendly and may decide not to advertise when advertisements by Dealer B appear.

Example 3: Alternatively, Manufacturer A may consider advertisements placed by Distributor B for its products (i.e., Manufacturer A's products) to be competitive and may want to advertise directly against these advertisements, i.e., when the advertisements placed by Distributor B are statistically more likely to appear in the SERP.

Example 4: In the absence of competing ads, advertisers will cannibalize their ads regardless of the position of the corresponding free listing.

More generally, advertisers may deem advertisements placed by particular companies or organizations to be friendly or competitive and enforce advertising rules based on such determinations.

Furthermore, advertisers can easily identify the domains of landing pages for their advertisements in the SERPs by scraping and then analyzing the SERPS. Thus, friendly advertisements may be considered as advertisements that have landing pages in friendly domains, and competitive advertisements may be considered as advertisements that have landing pages in competitive domains. Thus, in some embodiments, friendly and competitive advertisements may be determined based on a list of friendly domains and a list of competitive domains. In other embodiments, company and organization names, or even product names, may be used to determine whether an advertisement is friendly or competitive.

Thus, from the advertiser's perspective, each ad in the SERP can be categorized as: 1) its own ad, 2) friendly ad, 3) competitive ad, and 4) other, i.e., an ad from a company, organization, or domain that is neither the advertiser's own ad nor a friendly or competitive ad.

In some implementations, the CA analyzer 135 analyzes only the first SERP returned by a search for a search term. Typically, there is a maximum number of paid advertisements in the SERP. For example, the GOOGLE search engine returns a maximum of four advertisements in a SERP. Thus, the most frequently returned advertisements for each advertisement position may be identified. In the following description, it is assumed that the CA analyzer 135 identifies the most frequently returned advertisements for each SERP position, as well as the categories of advertisements.

In other embodiments, each advertisement may be categorized or classified in a more general manner, i.e., any number of categories other than friendly and competitive may be used.

Rules for Identifying Cannibalized Advertisements Paid advertisements may be evaluated using rules that evaluate a cannibalization score or measure. The rules may be formulated based on the factors discussed above: (1) the category assigned to the paid advertisement in the SERP, (2) the number of paid advertisements in the SERP, (3) the average distance from the advertiser's advertisement to its corresponding unpaid listing, and (4) the average location of the corresponding unpaid listing.

Table 1 shows an example approach for formulating rules that can be used to evaluate whether an ad is cannibalized using the factors mentioned above: ad location, ad category, distance, and location of the corresponding free listing.

In Table 1, each row represents one rule and the columns are as follows: A. is the number of the rule, B. is the category of the ad that appears in the first position of the first SERP, C. is the category of the ad that appears in the second position of the first SERP, D. is the average position of the free listings corresponding to the ads placed by the advertiser, E. is the average distance between the advertiser's ads and their corresponding free listings, F. indicates whether the advertiser's ads are considered cannibalistic (yes) or not (no) if the rule is met, and column G. indicates a brief description of the rule. Furthermore, in this example, the categories that can be assigned to the ads are A-advertiser, F-friendly, C-competitive, and O-other.

In the example of Table 1, columns D and E are shown as having integer values, but it should be understood that in some embodiments they may be real or decimal numbers based on sample values obtained for D - the average location of the corresponding free listing, and E - the average distance from the advertiser's advertisement to the corresponding free listing. Additionally, columns B, C, and F may alternatively have percentage, decimal, or fractional values. For example, in column B, the category values for the listing in the first position may be A (.75), F (.1), C (.1), and O (.05), indicating the ratio or percentage when the advertisement in the first position is in categories A, F, C, or O.

It should be further appreciated that in some embodiments, friendly ads are considered to be those with landing pages within domains that are considered friendly, while in other embodiments, there may be specific rules for specific domains. For example, rule 5 is evaluated based on whether an ad with a link to a specified domain (Dom A) appears in the first or second paid ad position, and whether the corresponding unpaid ad is less than 3 distances away from the advertiser's ad.

Finally, if more than one rule applies to an ad, the cannibalization score represents the sum, average, or weighted average of the results of all the rules that apply to the ad. For example, if one rule rates it at 75% and another rule rates it at 25%, then in the simplest case, the average value of 50% is the ad's cannibalization score.

Rules Based on Incremental Advertising Value There may be a difference between the revenue generated for an advertiser from an advertisement and the revenue generated from the corresponding free listing. This difference may occur when the advertisement and the free listing each link to a different landing page, because different landing pages may have different levels of effectiveness. Thus, if an advertisement performs worse on average than the corresponding free listing, the number of advertisement clicks will, on average, reduce the revenue by cannibalizing the number of clicks to the free listing. Using data sources provided by search engines such as GOOGLE ANALYTICS, GOOGLE ADWORDS, and GOOGLE SEARCH CONSOLE, which are described in more detail in Table 2 below, it is possible to determine the average value or revenue attributable to the number of advertisement clicks and the average value or revenue attributable to the number of clicks to the corresponding free listing. The average value of a click refers to the revenue expected per click or visit by a visitor to the linked landing page, i.e., the average revenue.

One rule that may be defined by an advertiser is based on an incremental or comparative value between the number of clicks on the advertisement and the number of clicks on the free listing, as defined in Equation 2 below.
[(value of ad click – value of free click)] / cost of ad click Equation 2
Here, if the incremental value defined by Equation 2 is greater than 1, then there is a net positive revenue from purchasing the advertisement after considering the cost of purchasing the advertisement. Thus, a simple rule based on Equation 2 is that if the incremental value is less than 1, the advertisement is cannibalized. However, other rules are contemplated, and rules based on the incremental value (Equation 2) may be mixed with rules such as those illustrated in Table 1. Additionally, other formulas may be defined for defining the incremental value without departing from the scope of the present invention.

It will be understood that each block of the flow diagram shown in FIG. 3 and combinations of blocks in the flow diagram may be implemented by computer program instructions. These program instructions may be provided to a processor to create a machine such that the instructions, executed on the processor, create means for performing the actions specified in one or more blocks of the flow diagram. The computer program instructions may be executed by the processor to cause the processor to perform a series of operational steps that perform the actions specified in one or more blocks of the flow diagram to create a computer-implemented process or method. The computer program instructions may also cause at least some of the operational steps shown in the blocks of the flow diagram to be performed in parallel. Furthermore, some of the steps may be performed across two or more processors, as may occur in a multiprocessor computer system. Furthermore, one or more blocks or combinations of blocks in the flow diagram may be executed simultaneously with other blocks or combinations of blocks, or even in a different order than that shown, without departing from the scope or spirit of the present invention.

The blocks of the flow diagram thus support combinations of means for performing the specified actions, combinations of steps for performing the specified actions, and program instruction means for performing the specified actions. It will also be understood that each block of the flow diagram and combinations of blocks in the flow diagram may be implemented by a dedicated hardware-based system that performs the specified actions or steps, or a combination of dedicated hardware and computer instructions.

Figure 3 provides an overall method 300 for identifying cannibalized ads and generating a cannibalization score for the ads. The purpose of method 300 is to evaluate search engine ads being purchased by advertisers to identify which are cannibalized.

In step 305, the CA analyzer 135 collects a set of search terms to evaluate. These may be ads currently purchased by the advertiser or potential ads. Typically, each ad corresponds to a search term or keyword. This step may be performed in a variety of ways. For example, if the search engine marketing (SEM) program of a particular advertiser, such as Surebuy (see FIG. 2), is being evaluated, the advertiser may typically provide a list of search terms to be purchased and the corresponding paid ads that will appear in the resulting SERPS.

Alternatively, a search engine may provide a set of keywords. For example, GOOGLE SEARCH CONSOLE, provided by GOOGLE, INC., may provide listings of search terms and the average search position of ads placed in association with each search term. Thus, for example, if one is only interested in cases where paid ads appear in the first and second positions and unpaid listings are in the first or second positions, then only search terms with an average position less than 2 may be evaluated.

Additional data may be obtained from search engines. For example, GOOGLE ADWORDS, offered by GOOGLE, INC., provides information including ad position, ad cost, click through rate, ad impression share, and similar metrics.

In step 310, rules for evaluating and scoring the advertisements are generated. These rules may be similar to the rules given in Table 1 or Equation 2.

In step 315, a keyword search is performed on a search engine for one of the collected search terms.

In step 320, the SERPs are received from the search engine. Note that while method 300 applies to a single search engine, it may be performed for additional search engines of interest. Thus, method 300 applies equally to all search engines.

In step 325, the listings in the received SERP (paid and free) are analyzed to identify the listing positions of the paid advertisements by the advertiser and the corresponding free listings, if any. In some embodiments, the search is performed iteratively, i.e., sampling is performed to obtain the average listing positions. In this case, in this step, the average listing positions of the paid advertisements and the corresponding free listings are updated.

In step 330, if a sampling of the average listing position has been performed, it may be determined whether further sampling needs to be performed. If so, processing returns to step 315; if not, processing proceeds to step 335.

In step 335, the rules formulated in step 310 are applied, resulting in a cannibalization score for the paid advertisement. The cannibalization score can have a variety of meanings. For example, in some implementations, the cannibalization score estimates the percentage or amount of sales or revenue lost due to the proximity of the paid advertisement to a corresponding listing in the SERP.

In another embodiment, the cannibalization score represents the probability that an ad is in fact cannibalizing and reducing the revenue generated by the free listing. In such an embodiment, a cannibalization score of 75 indicates that there is a 75% probability that an ad is cannibalizing.

In other implementations, the score may simply be a Boolean value (true, false) indicating that the ad is considered cannibalistic.

In step 340, if all keywords have not been processed, processing returns to step 315. If all keywords have been processed, processing continues at step 345.

In step 345, a report is generated that includes a cannibalization score for each keyword and paid advertisement combination. Such a report may include all collected keywords or only those keywords that are determined to be cannibalized. For example, only keyword and paid advertisement combinations that have a score higher than a given threshold may be included in the report. In some cases, the method ends at this step and the report is provided to the advertiser or a service company designated by the advertiser, such as an online advertising company.

In another embodiment, in step 350, the advertiser or the advertiser's web service company may modify their search engine advertising ad buys based on the report generated in step 345.

Measure of Effectiveness: Recovered Advertising Costs The concept of cannibalizing ads and improving the effectiveness of advertising campaigns by automatically analyzing whether paid ads are likely to be cannibalized is novel. By using this measure, the cost of an advertising campaign can be significantly reduced and its effectiveness improved. It is therefore important to be able to measure the cost reduction and effectiveness improvement of not purchasing cannibalized ads. Such a measure of effectiveness can be provided to advertisers to validate the effectiveness of employing an automated method for identifying cannibalized ads and taking appropriate actions, called non-cannibalizing actions.

FIG. 4 provides an overall method 400 for determining an effectiveness measure for an advertising campaign based on an estimate of potential savings from not purchasing some or all of the cannibalized ads identified by method 300. Method 400 calculates a measure, referred to herein as "recovered advertising cost," that estimates savings from not purchasing one or more paid ads determined to be cannibalized. Method 400 uses the cannibalization scores determined by method 300 to determine whether to purchase individual ads or refrain from purchasing individual ads.

In some implementations, method 400 calculates the reclaimed ad spend (RAS) for a set of keywords for a day (or another suitable time period), where the reclaimed ad spend for a day's search terms may be defined as follows:
RAS = Expected Advertising Cost – Optimized Advertising Cost Equation 3

In step 430, the expected ad spend (EAS) for one day may be calculated as follows:
EAS = average daily ad click cost (CPC) * average daily ad click-through rate (CTR) * daily search term impressions (i.e., the number of times users searched for that keyword) Equation 4

While a one-day period is used in some embodiments, it is understood that other periods may be used in other embodiments of the invention.

Optimized ad spend (OAS) is the actual measured advertising spend or cost, typically reported by a search engine during the period during which non-cannibalizing actions are taking place, i.e., the period during which the cannibalizing method 300 is operating, and the decision to purchase a keyword to implement paid advertising is based at least in part on the cannibalizing score for the keyword. Operation of the method 300 that includes the decision not to purchase cannibalizing advertising is also referred to as taking a non-cannibalizing action. An example of a non-cannibalizing action is not to purchase paid advertising whose cannibalizing score exceeds a threshold.

As an example, if method 400 is performed for ads purchased from the GOOGLE search engine, the measure of advertising cost is obtained from a service provided by GOOGLE named GOOGLE ADS.

As an example of expected advertising cost (Equation 4) for a search term in a day, if the CPC is $1, the CTR is 10%, and the expected number of impressions per day is 2,000, then the expected advertising cost (EAS) for that day = $1 * 0.1 * 2,000 = $200.

So if the actual optimized ad cost reported by the search engine is $50, then the recovered ad cost is $200 - $50 = $150.

In some embodiments, method 400 operates with only a single search engine, for example the GOOGLE search engine. In such embodiments, method 400 may operate using the data presented in Table 2 below. In other embodiments, comparable data is obtained from other search engines, or from search engines in addition to a single major search engine. Results may be aggregated across multiple search engines.

Some of the terms used in Table 2 are defined as follows:

Traffic source refers to how a visitor got to your site, such as by clicking on a paid ad, by clicking on an organic listing, etc.

Ad clicks refer to the number of times a user clicks on an advertisement corresponding to the search term being evaluated that appears within the SERP returned as a result of a keyword search using the Google search engine.

Advertising Cost: Dollars spent by advertisers on paid advertising that appears within SERPS.

Organic clicks refer to the number of times a user clicks on a free, "organic" Google search result listing.

An organic impression refers to the number of times a free listing appears on a SERP in response to a Google search performed by a user.

A baseline period refers to a period during which no non-cannibalizing actions are performed for a search term, i.e., the cannibalizing score for a search term is not considered in the purchase decision. A baseline period is typically measured in days. These days can be the number of days within a longer time interval, e.g., 4 days out of the last 8 days, or a number of consecutive days.

Baseline data refers to data captured during a baseline period.

Further Measures of Effectiveness In addition to recovered ad spend, additional or alternative measures can be used to show the effectiveness of reducing spending on cannibalized advertising. Each of the following measures can be compared between a period in which no non-cannibalized action is taking place and a similar period in which non-cannibalized action is taking place: ad clicks, ad cost, organic clicks, impressions, and revenue. Data can be aggregated across all search terms and across various time intervals (days, weeks, months, etc.).

Returning to method 400, in step 405, keywords to be evaluated are received or collected. Then, in step 410, for each received keyword, the corresponding cannibalization score generated by method 300 and the performance data described above in Table 2 are collected.

In step 420, a keyword is selected for processing. In step 430, an expected advertising cost for the keyword is calculated according to Equation 4. Then, in step 440, an optimized advertising cost is obtained from the data collected in step 410. As previously mentioned, the optimized advertising cost is typically collected from a search engine.

In step 450, the recovered advertising cost is calculated according to Equation 3. In step 450, other measures may also be calculated, as described above, including ad clicks, ad cost, organic clicks, impressions, and revenue, among others. As a basis for comparison, these measures may be calculated or collected for comparable time periods when non-cannibalizing actions were taken and when they were not taken.

In step 460, a determination is made as to whether all keywords have been processed. If not, processing returns to step 420; if so, processing proceeds to step 470.

In step 470, optionally, the results including the recovered advertising spend are aggregated across all keywords, possibly across multiple time intervals.

In step 480, the results are provided to the customer or client. The results may be in the form of a file, such as a MICROSOFT EXCEL table, among others, or may be provided as a presentation. Furthermore, such data may constitute intermediate results and may be further analyzed and used for reporting or decision-making purposes.

Figure 5 is a block diagram illustrating the software modules of the cannibalized advertising system (CAS) 100. Figure 5 describes the relevant software elements of the CAS 100, including a client computer 115, a search engine 120, a web server 140, and an ad server 130.

The client computer 115 interacts with the user 110 and enables the user 110 to perform web searches using a web browser 118.

Browser 118 is typically a standard commercially available browser such as MOZILLA FIREFOX or MICROSOFT INTERNET EXPLORER. Alternatively, browser 118 may also be a client application configured to receive and display graphics, text, multimedia, etc. over a network.

The browser 118 issues HTTP requests to Internet-connected computers such as the search engine 120, the web server 140, and the client computer 115, and receives HTTP responses. The application server 420 receives the HTTP requests and invokes the appropriate ad server 130 software modules to process the requests. The application server 520 may be any commercially available application server, including a web server that accepts and processes HTTP requests and returns HTTP responses with optional data content, which may be web pages such as HTML documents and linked objects (e.g., images).

The application server 520 establishes and manages sessions with the search engine 120 and the web server 140. In addition, the application server 520 may interact with the client computer 115.

The software modules of the search engine 120 are generally outside the scope of the present invention. However, as explained above and detailed in Table 2, the search engine is expected to provide various results data related to advertisements purchased by advertisers, which the search engine provides in the SERP in response to searches. The web server 140 manages one or more websites 145, each website including one or more domains.

The ad server 130 includes a keyword collector 530, a rule definer 532, a cannibalized advertising (CA) analyzer 135, a cannibalized advertising (CA) report generator 534, and in some embodiments an ad buyer 536, a recovery ad analyzer, a keyword database 550, a rule database 552, and an ad database 554. It should be appreciated that each of the above databases may be implemented as one or more computer files across one or more physical storage mechanisms. In one embodiment, each of the above databases is implemented as one or more relational databases and accessed using structured query language (SQL).

The keyword collector 530 obtains keywords from the search engine 120 and possibly from other sources. The keyword collector 530 may also obtain keywords from advertisers, for example in computer files supplied by the advertisers. The keyword collector 530 stores the keywords in the keyword database 450. The keyword collector 530 performs the processing associated with step 305 of the method 300.

The keyword collector 530 further collects data from the search engine 120 required by the method 400 to calculate the recovered advertising cost including revenue, average cost per click (CPC), average click-through rate (CTR), and impressions attributed to the keywords.

The rule definer 532 defines rules that determine a score for an advertisement in the SERP. The rule definer 532 stores the rules in a rule database 552. The rule definer 532 performs step 310 of method 300. The rule definer 532 may be implemented in a variety of ways, for example, in some embodiments the rule definer 532 simply receives a text file that defines the rules, while in other embodiments the rule definer 532 provides the client computer 115 with a graphical interface that allows a user to interactively define the rules. In general, the method for defining the rules is outside the scope of the present invention.

CA analyzer 135 performs the processing associated with steps 315-340 of method 300. CA analyzer 135 uses the collected keywords stored in keyword database 550 to obtain SERPS and locate paid advertisements and corresponding listings, and stores the results in keyword database 550. CA analyzer 135 further evaluates the SERPS to generate a cannibalization ad score, and stores the results in ad database 554.

The CA report generator 534 generates a report that provides the cannibalization ad scores for the ads. The CA report generator 534 performs step 345 of method 300.

In some implementations, the ad purchaser 536 purchases advertisements from the search engine 120 that take into account the results of the cannibalization advertising analysis represented by the cannibalization advertising report stored in the ad database 554. Specifically, the ad purchaser 536 decides on a case-by-case basis whether to purchase an advertisement based on its cannibalization score. In other implementations, the ad purchaser 536 is not part of the CAS 100. For example, the functions of the ad purchaser 536 may be performed by a third-party advertising agency.

The recovered advertising spend analyzer (RAS analyzer) 538 analyzes the results, i.e., effectiveness, obtained by not purchasing the ads that are considered cannibalizing. In general, the RAS analyzer 538 executes the method 400 to generate reports or other results that can be provided to the advertiser. The RAS analyzer 538 may also calculate other measures of effectiveness besides recovered advertising spend, as described above. Such measures include, among others, ad clicks, ad costs, organic clicks, impressions, and revenue for comparable periods when the non-cannibalizing action was taken and when it was not taken. The RAS analyzer 538 stores the results in the advertising database 554.

The keyword database 550 stores collected keywords. The keyword database 550 also stores the position or average position of the advertisements and of the corresponding free listings, where the advertisements are placed by the advertiser, or by the ad server 130, or by another party acting on behalf of the advertiser, and where the advertisements appear within the SERPS returned as a result of the keyword search.

The rules database 552 stores the rules used to generate a cannibalization score for an advertisement.

The advertisement database 554 stores advertiser-supplied advertisements that correspond to the collected keywords. Typically, each keyword of interest to an advertiser has a corresponding paid advertisement that may be placed by a search engine. The advertisement database 554 also stores the CA reports generated by the CA report generator 534 and the advertising spend recovered and other results data generated by the RAS analyzer 538.

The above specification, examples and data provide a complete description of the manufacture and use of the composition of the invention. Many embodiments of the invention can be made without departing from the spirit and scope of the invention.

Claims (13)

1. A computer-implemented method for adjusting bids for internet advertisements in an internet advertising campaign, comprising:
a keyword collector collecting a plurality of keywords and a plurality of advertisements associated with a specified advertiser, each keyword corresponding to one of a plurality of advertisements that the advertiser intends to purchase as part of an Internet advertising campaign, and in response to receiving the keywords, a search engine returning a search engine results page (SERP), the SERP including: (1) at least one advertisement; (2) at least one free listing, the advertisement having a position in the SERP relative to other advertisements, the first advertisement position being the most valuable position and the higher the ordinal number of the advertisement position being less valuable; and (2) at least one free listing, each free listing including a link to a landing page, the free listing having a position in the SERP for the free listing, the first advertisement position being the most valuable position and the higher the ordinal number of the advertisement position being less valuable;
a rule definer maintaining at least one cannibalization rule that calculates a cannibalization score for an advertisement , the cannibalization score for an advertisement indicating that the advertisement is cannibalized if the advertisement generates reduced value due to the advertisement's position in a SERP and a listing position of a corresponding free listing in the same SERP, the corresponding free listing referencing the same product or service as referenced by the advertisement;
a cannibalization advertisement analyzer periodically determining whether to purchase a specified advertisement from the plurality of received advertisements, the periodic determination comprising :
performing a keyword search to obtain a SERP for the keyword corresponding to the specified advertisement ;
locating the specified advertisement within the SERP and the advertisement for the corresponding free listing within the SERP ;
generating a cannibalization score for the designated advertisement by applying the at least one cannibalization rule to the designated advertisement ;
an advertisement purchaser performing an advertisement purchase for the aggregated plurality of advertisements, where if the aggregated cannibalization scores for an advertisement indicate that the advertisement is cannibalized, the advertisement is not purchased;
4. A computer-implemented method comprising: 2. The method of claim 1 , wherein the cannibalization score represents a measure selected from the group consisting of an estimate of the revenue reduction to the advertiser due to the advertisement appearing in the same SERP as the corresponding listing, an estimate of the likelihood that the advertisement is cannibalized, or a BOOLEAN value indicating whether the advertisement is cannibalized. 3. The method of claim 1 or 2, wherein the cannibalization score indicates that the advertisement is a cannibalized advertisement if the advertisement appears in a first advertisement position in a SERP, there is no advertisement in a second advertisement position, and a corresponding free listing appears in the first free listing position in the SERP . 2. The method of claim 1 , wherein at least one cannibalization rule is additionally based on one or more factors selected from the group consisting of the number of advertisements in the SERP , the average revenue per click for the advertisements, and the average revenue per click for the free listings. The method of claim 1, wherein each advertisement in the SERP can be assigned a category selected from the group consisting of advertisements placed by the specified advertiser, friendly advertisements, competitive advertisements, or another advertisement, and at least one rule is based on the category of advertisements in the SERP. 2. The method of claim 1, wherein a distance between advertisements and free listings in a SERP can be calculated , and at least one cannibalization rule is based on the distance between the advertisements placed by the advertisers and the corresponding free listings. The method of claim 1 , further comprising generating a report including a cannibalization score for each advertisement that the advertiser intends to purchase . a keyword gatherer for gathering a plurality of keywords and a plurality of advertisements , each keyword corresponding to one of a plurality of advertisements that an advertiser intends to purchase as part of an Internet advertising campaign, in response to receiving the keywords, a search engine returns a search engine results page (SERP), the SERP including: (1 ) one or more advertisements; (2) at least one free listing, the advertisement having a position in the SERP relative to other advertisements, the first advertisement position being the most valuable position and the higher the ordinal number of the advertisement position being less valuable; and (2) at least one free listing, each free listing including a link to a landing page, the free listing having a position in the SERP relative to other advertisements, the first advertisement position being the most valuable position and the higher the ordinal number of the advertisement position being less valuable;
a keyword definer for maintaining at least one cannibalization rule that calculates a cannibalization score for an advertisement , the cannibalization score for an advertisement indicating that the advertisement is cannibalized if the advertisement generates reduced value due to the advertisement's position in a SERP and a corresponding listing position in the same SERP, the corresponding free listing referencing the same product or service as referenced by the advertisement;
1. A CA analyzer for periodically determining whether to purchase a specified advertisement from among a plurality of received advertisements , the periodic determination comprising:
performing a keyword search to obtain a SERP for the keyword corresponding to the specified advertisement ;
locating the specified advertisement within the SERP and the advertisement for the corresponding free listing within the SERP ;
generating a cannibalization score for the designated advertisement by applying the at least one cannibalization rule to the designated advertisement;
performing an advertisement buy for the aggregated plurality of advertisements, where if the aggregated cannibalization scores for an advertisement indicate that the advertisement is cannibalized, the advertisement is not purchased;
and a CA analyzer including: 9. The network computing device of claim 8, wherein at least one cannibalization rule is additionally based on one or more factors selected from the group consisting of: the number of advertisements in the SERP, the position of the corresponding free listing in the SERP, the average revenue per click for the advertisement, and the average revenue per click for the free listing. 9. The network computing device of claim 8, wherein each advertisement in a SERP can be assigned a category selected from the group consisting of advertisements placed by the specified advertiser, friendly advertisements, competitive advertisements, or another advertisement, and at least one rule is based on the category of advertisements in the SERP. 10. The network computing device of claim 8, wherein a distance between advertisements and free listings in a SERP can be calculated , and at least one cannibalization rule is based on the distance between the advertisements placed by the specified advertisers and the corresponding free listings. 10. The network computing device of claim 8 , further comprising a cannibalization advertisement report generator that generates a report including a cannibalization score for each advertisement that an advertiser intends to purchase. 13. The network computing device of claim 12 , further comprising an advertisement purchaser that purchases advertisements from the search engine based at least in part on the report of cannibalized advertisements.