Enterprise reputation evaluation

Abstract

Enterprise reputation evaluation techniques are disclosed. Reputation data extracted from each of a plurality of data sources is received. The reputation data is associated with a plurality of physical locations of an enterprise entity. An aggregate reputation score is determined for the entity using at least a portion of the received data. The aggregate reputation score is provided as output.

Description

BACKGROUND OF THE INVENTION

Businesses are increasingly concerned with their online reputations. For example, both positive and negative reviews posted to a review website can impact revenue. As more review websites are created, and as more users post more content to those sites, it is becoming increasingly difficult for businesses to monitor such sites. In the case of multi-location businesses, such as enterprises, the problem is compounded.

BRIEF DESCRIPTION OF THE DRAWINGS

Various embodiments of the invention are disclosed in the following detailed description and the accompanying drawings.

FIG. 1 illustrates an embodiment of an environment in which business reputation information is collected, analyzed, and presented.

FIG. 2 illustrates an example of components included in embodiments of a reputation platform.

FIG. 3 illustrates an embodiment of a process for enrolling a business with a reputation platform.

FIG. 4 illustrates an example of components included in embodiments of a reputation platform.

FIG. 5 illustrates an embodiment of a process for refreshing reputation data.

FIG. 6 illustrates an example of an interface as rendered in a browser.

FIG. 7 illustrates an example of components included in an embodiment of a reputation platform.

FIG. 8 illustrates an embodiment of a process for generating a reputation score.

FIG. 9 illustrates an example of an interface as rendered in a browser.

FIG. 10 illustrates an example of an interface as rendered in a browser.

FIG. 11 illustrates an example of an interface as rendered in a browser.

FIG. 12 illustrates a portion of an interface as rendered in a browser.

FIG. 13 illustrates a portion of an interface as rendered in a browser.

FIG. 14 illustrates an example of an interface as rendered in a browser.

FIG. 15 illustrates a portion of an interface as rendered in a browser.

FIG. 16 illustrates a portion of an interface as rendered in a browser.

FIG. 17 illustrates an example of an interface as rendered in a browser.

FIG. 18 illustrates a portion of an interface as rendered in a browser.

FIG. 19 illustrates a portion of an interface as rendered in a browser.

DETAILED DESCRIPTION

The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and/or a processor, such as a processor configured to execute instructions stored on and/or provided by a memory coupled to the processor. In this specification, these implementations, or any other form that the invention may take, may be referred to as techniques. In general, the order of the steps of disclosed processes may be altered within the scope of the invention. Unless stated otherwise, a component such as a processor or a memory described as being configured to perform a task may be implemented as a general component that is temporarily configured to perform the task at a given time or a specific component that is manufactured to perform the task. As used herein, the term ‘processor’ refers to one or more devices, circuits, and/or processing cores configured to process data, such as computer program instructions.

A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims and the invention encompasses numerous alternatives, modifications and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example and the invention may be practiced according to the claims without some or all of these specific details. For the purpose of clarity, technical material that is known in the technical fields related to the invention has not been described in detail so that the invention is not unnecessarily obscured.

FIG. 1 illustrates an embodiment of an environment in which business reputation information is collected, analyzed, and presented. In the example shown, the user of client device 106 (hereinafter referred to as “Bob”) owns a single location juice bar (“Bob's Juice Company”). The user of client device 108 (hereinafter referred to as “Alice”) is employed by a national chain of convenience stores (“ACME Convenience Stores”). As will be described in more detail below, Bob and Alice can each access the services of reputation platform 102 (via network 104) to track the reputations of their respective businesses online. The techniques described herein can work with a variety of client devices 106-108 including, but not limited to personal computers, tablet computers, and smartphones.

Reputation platform 102 is configured to collect reputation and other data from a variety of sources, including review websites 110-114, social networking websites 120-122, and other websites 132-134. In some embodiments, users of platform 102, such as Alice and Bob, can also provide offline survey data to platform 102. In the examples described herein, review site 110 is a general purpose review site that allows users to post reviews regarding all types of businesses. Examples of such review sites include Google Places, Yahoo! Local, and Citysearch. Review site 112 is a travel-oriented review site that allows users to post reviews of hotels, restaurants, and attractions. One example of a travel-oriented review site is TripAdvisor. Review site 114 is specific to a particular type of business (e.g., car dealers). Examples of social networking sites 120 and 122 include Twitter and Foursquare. Social networking sites 120-122 allow users to take actions such as “checking in” to locations. Finally, personal blog 134 and online forum 132 are examples of other types of websites “on the open Web” that can contain business reputation information.

Platform 102 is illustrated as a single logical device in FIG. 1. In various embodiments, platform 102 is a scalable, elastic architecture and may comprise several distributed components, including components provided by one or more third parties. Further, when platform 102 is referred to as performing a task, such as storing data or processing data, it is to be understood that a sub-component or multiple sub-components of platform 102 (whether individually or in cooperation with third party components) may cooperate to perform that task.

Account/Business Setup

FIG. 2 illustrates an example of components included in embodiments of a reputation platform. In particular, FIG. 2 illustrates components of platform 102 that are used in conjunction with a business setup process.

In order to access the services provided by reputation platform 102, Bob first registers for an account with the platform. At the outset of the process, he accesses interface 202 (e.g., a web-based interface) and provides information such as a desired username and password. He also provides payment information (if applicable). If Bob has created accounts for his business on social networking sites such as sites 120 and 122, Bob can identify those accounts to platform 102 as well.

Next, Bob is prompted by platform 102 to provide the name of his business (e.g., “Bob's Juice Company”), a physical address of the juice bar (e.g., “123 N. Main St.; Cupertino, Calif. 95014), and the type of business that he owns (e.g., “restaurant” or “juice bar”). The business information entered by Bob is provided to auto find engine 204, which is configured to locate, across sites 110-114, the respective profiles on those sites pertaining to Bob's business (e.g., “www.examplereviewsite.com/CA/Cupertino/BobsJuiceCo.html”), if present. Since Bob has indicated that his business is a juice bar, reputation platform 102 will not attempt to locate it on site 114 (a car dealer review site), but will attempt to locate it within sites 110 and 112.

In the example shown in FIG. 2, sites 110 and 114 make available respective application programming interfaces (APIs) 206 and 208 that are usable by auto find engine 204 to locate business profiles on their sites. Site 112 does not have a profile finder API. In order to locate a business profile there, auto find engine 204 is configured to perform a site-specific search using a script that accesses a search engine (e.g., through search interface 210). As one example, a query of: “site:www.examplereviewsite.com ‘Bob's Juice Company’ Cupertino” could be submitted to the Google search engine using interface 210.

Results obtained by auto find engine 204 are provided to verification engine 212, which confirms that information, such as the physical address and company name provided by Bob are present in the located profiles. Verification engine 212 can be configured to verify all results (including any obtained from site 110 and 114), and can also be configured to verify (or otherwise process) just those results obtained via interface 210. As one example, for a given query, the first ten results obtained from search interface 210 can be examined. The result that has the best match score and also includes the expected business name and physical address is designated as the business's profile at the queried site.

In some embodiments, verification engine 212 presents results to Bob for verification that the located profiles correspond to his business. As one example, Bob may be shown (via interface 202) a set of URLs corresponding to profiles on each of the sites 110-114 where his business has been located and asked to verify that the profiles are indeed for his business. Once confirmed by Bob, the URLs of the profiles (also referred to herein as “subscriptions”) and any other appropriate data are stored in database 214. Examples of such other data include overview information appearing on the business's profile page (such as a description of the business) and any social data (e.g., obtained from sites 120-122).

In various embodiments, users are given the option by platform 102 to enter the specific URLs corresponding to their business profiles on review sites. For example, if Bob knows the URL of the Google Places page corresponding to his business, he can provide it to platform 102 and use of auto find engine 204 is omitted (or reduced) as applicable.

FIG. 3 illustrates an embodiment of a process for enrolling a business with a reputation platform. In some embodiments process 300 is performed by platform 102. The process begins at 302 when a physical address of a business is received. As one example, when Bob provides the address of his business to platform 102 via interface 202, that address is received at 302. At 304, the received address is used as a query. As one example of the processing performed at 304, the received address is provided to site 110 using API 206. As another example, a site-specific query (e.g., of site 112) is submitted to a search engine via search interface 210.

At 306, results of the query (or queries) performed at 304 are verified. As one example of the processing performed at 304, verification engine 212 performs checks such as confirming that the physical address received at 302 is present in a given result. As another example, a user can be asked to confirm that results are correct, and if so, that confirmation is received as a verification at 306. Finally, at 308, verified results are stored. As one example, URLs for each of the verified profiles is stored in database 214. Although pictured as a single database in FIG. 2, in various embodiments, platform 102 makes use of multiple storage modules, such as multiple databases. Such storage modules may be of different types. For example, user account and payment information may be stored in a MySQL database, while extracted reputation information (described in more detail below) may be stored using MongoDB.

Where a business has multiple locations, the business owner (or a representative of the business, such as Alice) can be prompted to loop through process 300 for each of the business locations. Physical addresses and/or the URLs of the corresponding profiles on sites such as sites 110-114 can also be provided to platform 102 in a batch, rather than by manually entering in information via interface 202. As one example, suppose ACME Convenience Stores has 2,000 locations throughout the United States. Instead of manually entering in the physical location of each of the stores, Alice may instead elect to upload to platform 102 a spreadsheet or other file (or set of files) that includes the applicable information.

Tags associated with each location can also be provided to platform 102 (e.g., as name-value pairs). For example, Alice can tag each of the 2,000 locations with a respective store name (Store #1234), manager name (Tom Smith), region designation (West Coast), brand (ACME-Quick vs. Super-ACME), etc. As needed, tags can be edited and deleted, and new tags can be added. For example, Alice can manually edit a given location's tags (e.g., via interface 202) and can also upload a spreadsheet of current tags for all locations that supersede whatever tags are already present for her locations in platform 102. As will be described in more detail below, the tags can be used to segment the business to create custom reports and for other purposes.

Ongoing Data Collection and Processing

Once a business (e.g., Bob's Juice Company) has an account on reputation platform 102, and once the various subscriptions (i.e., the URLs of the business's profiles on the various review sites) have been identified and stored in database 214, collecting and processing of review and other data is performed. FIG. 4 illustrates an example of components included in embodiments of a reputation platform. In particular, FIG. 4 illustrates components of platform 102 that are used in conjunction with the ongoing collection and processing of data.

Reputation platform 102 includes a scheduler 402 that periodically instructs collection engine 404 to obtain data from sources such as sites 110-114. In some embodiments, data from sites 120-122, and/or 132-134 is also collected by collection engine 404. Scheduler 402 can be configured to initiate data collection based on a variety of rules. For example, it can cause data collection to occur once a day for all businesses across all applicable sites. It can also cause collection to occur with greater frequency for certain businesses (e.g., which pay for premium services) than others (e.g., which have free accounts). Further, collection can be performed across all sites (e.g., sites 110-114) with the same frequency or can be performed at different intervals (e.g., with collection performed on site 110 once per day and collection performed on site 112 once per week).

In addition to or instead of the scheduled collection of data, data collection can also be initiated based on the occurrence of an arbitrary triggering event. For example, collection can be triggered based on a login event by a user such as Bob (e.g., based on a permanent cookie or password being supplied). Collection can also be triggered based on an on-demand refresh request by the user (e.g., where Bob clicks on a “refresh my data” button in interface 202). Other elements depicted in FIG. 4 will be described in conjunction with process 500 shown in FIG. 5.

FIG. 5 illustrates an embodiment of a process for refreshing reputation data. In some embodiments process 500 is performed by platform 102. The process begins at 502 when a determination is made that a data refresh should be performed. As one example, such a determination is made at 502 by scheduler 402 based on an applicable schedule. As another example, such a determination is made at 502 when a triggering event (such as a login event by Bob) is received by platform 102.

At 504, a determination is made as to which sites should be accessed. As one example, in some embodiments collection engine 404 reviews the set of subscriptions stored in database 214 for Bob's Juice Company. The set of subscriptions associated with Bob's company are the ones that will be used by collection engine 404 during the refresh operation. As previously mentioned, a refresh can be performed on behalf of multiple (or all) businesses, instead of an individual one such as Bob's Juice Company. In such a scenario, portion 504 of the process can be omitted as applicable.

At 506, information is obtained from the sites determined at 504. As shown in FIG. 4, collection engine 404 makes use of several different types of helpers 420-428. Each helper (e.g., helper 420) is configured with instructions to fetch data from a particular type of source. As one example, although site 110 provides an API for locating business profiles, it does not make review data available via an API. Such data is instead scraped by platform 102 accordingly. In particular, when a determination is made that reviews associated with Bob's Juice Company on site 110 should be refreshed by platform 102, an instance 430 of helper 420 is executed on platform 102. Instance 430 is able to extract, for a given entry on site 110, various components such as: the reviewer's name, profile picture, review title, review text, and rating. Helper 424 is configured with instructions for scraping reviews from site 114. It is similarly able to extract the various components of an entry as posted to site 114. Site 112 has made available an API for obtaining review information and helper 422 is configured to use that API.

Other types of helpers can extract other types of data. As one example, helper 426 is configured to extract check-in data from social site 120 using an API provided by site 120. As yet another example, when an instance of helper 428 is executed on platform 102, a search is performed across the World Wide Web for blog, forum, or other pages that discuss Bob's Juice Company. In some embodiments, additional processing is performed on any results of such a search, such as sentiment analysis.

In various embodiments, information, obtained on behalf of a given business, is retrieved from different types of sites in accordance with different schedules. For example, while review site data might be collected hourly, or on demand, social data (collected from sites 120-122) may be collected once a day. Data may be collected from sites on the open Web (e.g., editorials, blogs, forums, and/or other sites not classified as review sites or social sites) once a week.

At 508, any new results (i.e., those not already present in database 214) are stored in database 214. As needed, the results are processed (e.g., by converting reviews into a single, canonical format) prior to being included in database 214. In various embodiments, database 214 supports heterogeneous records and such processing is omitted or modified as applicable. For example, suppose reviews posted to site 110 must include a score on a scale from one to ten, while reviews posted to site 112 must include a score on a scale from one to five. Database 214 can be configured to store both types of reviews. In some embodiments, the raw score of a review is stored in database 214, as is a converted score (e.g., in which all scores are converted to a scale of one to ten). As previously mentioned, in some embodiments, database 214 is implemented using MongoDB, which supports such heterogeneous record formats.

Prior to the first time process 500 is executed with respect to Bob's Juice Company, no review data is present in database 214. Portion 506 of the process is performed for each of the data sources applicable to Bob's business (via instances of the applicable helpers), and the collected data is stored at 508. On subsequent refreshes of data pertinent to Bob's company, only new/changed information is added to database 214. In various embodiments, alerter 432 is configured to alert Bob (e.g., via an email message) whenever process 500 (or a particular portion thereof) is performed with respect to his business. In some cases, alerts are only sent when new information is observed, and/or when reputation scores associated with Bob's business (described in more detail below) change, or change by more than a threshold amount.

Reputation Scoring

Platform 102 is configured to determine a variety of reputation scores on behalf of businesses such as Bob's Juice Company. In the case of multiple-location businesses, such as ACME, individual reputation scores are determined for each of the locations, and the scores of individual businesses can be aggregated in a variety of ways. As will be described in more detail below, the scores provide users with perspective on how their businesses are perceived online. Also as will be described in more detail below, users are able to explore the factors that contribute to their businesses' reputation scores by manipulating various interface controls, and they can also learn how to improve their scores. In the case of multi-location businesses, such as ACME, users can segment the locations in a variety of ways to gain additional insight.

FIG. 6 illustrates an example of an interface as rendered in a browser. In particular, Bob is presented with interface 600 after logging in to his account on platform 102 using a browser application on client device 106 and clicking on tab option 602.

In region 604 of interface 600, a composite reputation score (728 points) is depicted on a scale 606. Example ways of computing a composite score are described in conjunction with FIG. 7. The composite reputation score provides Bob with a quick perspective on how Bob's Juice Company is perceived online. A variety of factors can be considered in determining a composite score. Six example factors are shown in region 608, each of which is discussed below. For each factor, Bob can see tips on how to improve his score with respect to that factor by clicking on the appropriate box (e.g., box 622 for tips on improving score 610). In the example shown in FIG. 6, a recommendation box is present for each score presented in region 608. In some embodiments, such boxes are only displayed for scores that can/should be improved. For example, given that score 614 is already very high, in some embodiments, box 626 is omitted from the interface as displayed to Bob, or an alternate message is displayed, such as a general encouragement to “keep up the good work.”

Overall Score (610): This value reflects the average review score (e.g., star rating) across all reviews on all review sites. As shown, Bob's business has an average rating of 0.50 across all sites. If Bob clicks on box 622, he will be presented with a suggestion, such as the following: “Overall score is the most influential metric. It can appear in both the review site search results and in your general search engine results. Generating a larger volume of positive reviews is the best way to improve the overall score. Typically, volume is the best approach as your average, happy customer will not write a review without being asked.” Additional, personalized advice may also be provided, such as telling Bob he should click on tab 634 and request five reviews.

Timeliness (612): This score indicates how current a business's reviews are (irrespective of whether they are positive or negative). In the example shown, reviews older than two months have less of an impact than more recent reviews. Thus, if one entity has 200 reviews with an average rating of four stars, at least some of which were recently authored, and a second entity has the same volume and star rating but none of the reviews were written in the last two months, the first entity will have a higher timeliness score and thus a higher composite reputation score. If Bob clicks on box 624, he will be presented with a suggestion, such as the following: “Managing your online reviews is not a one-time exercise, but a continual investment into your business. Encourage a steady trickle of new reviews on a regular basis to ensure that your reviews don't become stale.” Other measures of Timeliness can also be used, such as a score that indicates the relative amount of new vs. old positive reviews and new vs. old negative reviews. (I.e., to see whether positive or negative reviews dominate in time.)

Length (614): This score indicates the average length of a business's reviews. Longer reviews add weight to the review's rating. If two reviews have the same star rating (e.g., one out of five stars), but the first review is ten words and the second review is 300 words, the second review will be weighted more when computing the composite score. If Bob clicks on box 626, he will be presented with a suggestion, such as the following: “Encourage your positive reviewers to write in-depth reviews. They should detail their experiences and highlight what they like about your business. This provides credibility and the guidance makes review writing easier for them.” Other measures of Length can also be used, such as a score that indicates the relative amount of long vs. short positive reviews and long vs. short negative reviews. (I.e., to see whether positive or negative reviews dominate in length.)

Social Factors (616): Reviews that have been marked with social indicators (e.g., they have been marked by other members of the review community as being “helpful” or “funny”) will have more bearing on the outcome of the composite score. By clicking on box 632, Bob will be presented with an appropriate suggestion for improvement.

Reviewer Authority (618): A review written by an established member of a community (e.g., who has authored numerous reviews) will have a greater impact on the outcome of the composite score than one written by a reviewer with little or no history on a particular review site. In some embodiments, the audience of the reviewer is also taken into consideration. For example, if the reviewer has a large Twitter following, his or her review will have a greater bearing on the outcome of the score. If Bob clicks on box 628, he will be presented with a suggestion, such as the following: “Established reviewers can be a major boon to your review page. Their reviews are rarely questioned and their opinions carry significant weight. If you know that one of your customers is an active reviewer on a review site, make a special effort to get him or her to review your business.”

Industry (620): Review sites that are directly related to the vertical in which the entity being reviewed resides are given more weight. For example, if the entity being reviewed is a car dealership and the review site caters specifically to reviews about car dealerships, the reviews in that specific site will have a greater impact on the outcome of the composite score than those on vertically ambiguous websites. If Bob clicks on box 630, he will be presented with a suggestion, such as the following: “The most important review sites for your business should have your best reviews. Monitor your website analytics to find the sites having the biggest impact on your business, and reinforce your presence on those sites.”

In various embodiments of interface 600, additional controls for interactions are made available. For example, a control can be provided that allows a user to see individual outlier reviews—reviews that contributed the most to/deviated the most from the overall score (and/or individual factors). As one example, a one-star review that is weighted heavily in the calculation of a score or scores can be surfaced to the user. The user could then attempt to resolve the negative feelings of the individual that wrote the one-star review by contacting the individual. As another example, a particularly important five-star review (e.g., due to being written by a person with a very high reviewer authority score) can be surfaced to the user, allowing the user to contact the reviewer and thank him or her. As yet another example, if an otherwise influential review is stale (and positive), the review can be surfaced to the user so that the user can ask the author to provide an update or otherwise refresh the review.

A variety of weights can be assigned to the above factors when generating the composite score shown in region 604. Further, the factors described above need not all be employed nor need they be employed in the manners described herein. Additional factors can also be used when generating a composite score. An example computation of a composite score is discussed in conjunction with FIG. 7.

Example Score Generation

FIG. 7 illustrates an example of components included in an embodiment of a reputation platform. In particular, FIG. 7 illustrates components of platform 102 that are used in conjunction with generating reputation scores.

In some embodiments, whenever Bob accesses platform 102 (and/or based on the elapsing of a certain amount of time), the composite score shown at 604 in FIG. 6 is refreshed. In particular, scoring engine 702 retrieves, from database 214, review and other data pertaining to Bob's business and generates the various scores shown in FIG. 6. Example ways of computing a composite reputation score are as follows.

(1) Base Score

First, scoring engine 702 computes a base score “B” that is a weighted average of all of the star ratings of all of the individual reviews on all of the sites deemed relevant to Bob's business:

$B=100·\frac{\sum _i^{N_r}{s}_i{w}_i}{\sum _i^{N_r}{w}_i}·\Theta \left(N_r-N_{\min }\right)$

where “N_r” is the total number of reviews, “s_i” is the number of “stars” for review “i” normalized to 10, “w_i” is the weight for review “i,” Θ is the Heaviside step function, and “N_min” is the minimum number of reviews needed to score (e.g., 4). The factor 100 is used to expand the score to a value from 0 to 1000.

One example of the function “w_i” is as follows:w_i=D_A·T_i·P_i·R_A·S_F·L_F

In the above, “D_A” is the domain authority, which reflects how important the domain is with respect to the business. As one example, a doctor-focused review site may be a better authority for reviews of doctors than a general purpose review site. One way to determine domain authority values is to use the domain's search engine results page placement using the business name as the keyword.

“R_A” is the reviewer authority. One way to determine reviewer authority is to take the logarithm of 1+the number of reviews written by the reviewer. As explained above, a review written by an individual who has authored many reviews is weighted more than one written by a less prolific user.

“S_F” is the social feedback factor. One way to determine the factor is to use the logarithm of 1+the number of pieces of social feedback a review has received.

“L_F” is the length factor. One way to specify this value is to use 1 for short reviews, 2 for medium reviews, and 4 for long reviews.

“T_i” is the age factor. One way to specify this factor is through the following: If the age is less than two months T_i=1, if the age “a,” (in months) >2 months, then the following value is used:T_i=max(e^{−ω·(ai−2)},0.5)

where ω is the time-based decay rate.

“P_i” is the position factor for review “i.” The position factor indicates where a given review is positioned among other reviews of the business (e.g., it is at the top on the first page of results, or it is on the tenth page). One way to compute the position factor is as follows:

$P_i=e^{-\frac{p_i}{\lambda }}$

where λ is the positional decay length.

In some cases, a given site (e.g., site 110) may have an overall rating given for the business on the main profile page for that business on the site. In some embodiments, the provided overall rating is treated as an additional review with age a=a₀ and position p=p₀ and given an additional weight factor of 2.

(2) Normalization

Once the base score has been computed, it is normalized (to generate “B_norm”). In some embodiments this is performed by linearly stretching out the range of scores from 8 to 10 to 5 to 10 and linearly squeezing the range of scores from 0 to 8 to 0 to 5.

Optional Correction Factors

In some embodiments, a correction factor “C” is used for the number of reviews in a given vertical and locale:

$C=a+b·\frac{2}{\pi }{\tan }^{-1}\left(\frac{2·N_r}{\stackrel{\_}{N_r}}\right)$

where “N_r” is the number of reviews for the business and the median number of reviews is taken for the business's vertical and locale. An example value for “a” is 0.3 and an example value for “b” is 0.7.

One alternate version of correction factor “C” is as follows:

$C=a+b·\frac{2}{\pi }{\tan }^{-1}\left(\frac{2·N_r}{\min \left(\max \stackrel{\_}{N_r},N_{\min }\right),N_{\max })}\right)$

where “N_min” and “N_max” are the limits put on the comparator “N_r” in the denominator of the argument of the arctan in the correction factor. An example value for “N_min” is 4 and an example value for “N_max” is 20.

A randomization correction “R” can also be used:

$R=\min \left(1000,C·B_{\mathrm{norm}}+\frac{\mathrm{mod}\left(\mathrm{uid},40\right)-20}{N_r}\right)$

where “C” is a correction factor (e.g., one of the two discussed above), “B_norm” is the normalized base score discussed above, and “uid” is a unique identifier assigned to the business by platform 102 and stored in database 214. The randomization correction can be used where only a small number of reviews are present for a given business.

Another example of “R” is as follows:R=max(0,C·B_norm−37.5·e^−0.6·α)

where “a” is the age of the most recent review.

Additional Examples of Scoring Embodiments

As explained above, a variety of techniques can be used by scoring engine 702 in determining reputation scores. In some embodiments, scores for all types of businesses are computed using the same sets of rules. In other embodiments, reputation score computation varies based on industry (e.g., reputation scores for car dealers using one approach and/or one set of factors, and reputation scores for doctors using a different approach and/or different set of factors). Scoring engine 702 can be configured to use a best in class entity when determining appropriate thresholds/values for entities within a given industry. The following are yet more examples of factors that can be used in generating reputation scores.

Review Volume:

The volume of reviews across all review sites can be used as a factor. For example, if the average star rating and the number of reviews are high, a conclusion can be reached that the average star rating is more accurate than where an entity has the same average star rating and a lower number of reviews. The star rating will carry more weight in the score if the volume is above a certain threshold. In some embodiments, thresholds vary by industry. Further, review volume can use more than just a threshold. For example, an asymptotic function of number of reviews, industry, and geolocation of the business can be used as an additional scoring factor.

Multimedia:

Reviews that have multimedia associated with them (e.g., a video review, or a photograph) can be weighted differently. In some embodiments, instead of using a separate multimedia factor, the length score of the review is increased (e.g., to the maximum value) when multimedia is present.

Review Distribution:

The population of reviews on different sites can be examined, and where a review distribution strays from the mean distribution, the score can be impacted. As one example, if the review distribution is sufficiently outside the expected distribution for a given industry, this may indicate that the business is engaged in gaming behavior. The score can be discounted (e.g., by 25%) accordingly. An example of advice for improving a score based on this factor would be to point out to the user that their distribution of reviews (e.g., 200 on site 110 and only 2 on site 112) deviates from what is expected in the user's industry, and suggest that the user encourage those who posted reviews to site 110 do so on site 112 as well.

Text Analysis:

Text analysis can be used to extract features used in the score. For example, reviews containing certain key terms (e.g., “visited” or “purchased”) can be weighted differently than those that do not.

FIG. 8 illustrates an embodiment of a process for generating a reputation score. In some embodiments, process 800 is performed by platform 102. The process begins at 802 when data obtained from each of a plurality of sites is received. As one example, process 800 begins at 802 when Bob logs into platform 102 and, in response, scoring engine 702 retrieves data associated with Bob's business from database 214. In addition to generating reputation scores on demand, scores can also be generated as part of a batch process. As one example, scores across an entire industry can be generated (e.g., for benchmark purposes) once a week. In such situations, the process begins at 802 when the designated time to perform the batch process occurs and data is received from database 214. In various embodiments, at least some of the data received at 802 is obtained on-demand directly from the source sites (instead of or in addition to being received from a storage, such as database 214).

At 804, a reputation score for an entity is generated. Various techniques for generating reputation scores are discussed above. Other approaches can also be used, such as by determining an average score for each of the plurality of sites and combining those average scores (e.g., by multiplying or adding them and normalizing the result). As mentioned above, in some embodiments the entity for which the score is generated is a single business (e.g., Bob's Juice Company). The score generated at 804 can also be determined as an aggregate across multiple locations (e.g., in the case of ACME Convenience Stores) and can also be generated across multiple businesses (e.g., reputation score for the airline industry), and/or across all reviews hosted by a site (e.g., reputation score for all businesses with profiles on site 110). One way to generate a score for multiple locations (and/or multiple businesses) is to apply scoring techniques described in conjunction with FIG. 7 using as input the pool of reviews that correspond to the multiple locations/businesses. Another way to generate a multi-location and/or multi-business reputation score is to determine reputation scores for each of the individual locations (and/or businesses) and then combine the individual scores (e.g., through addition, multiplication, or other appropriate combination function).

Finally, at 806 the reputation score is provided as output. As one example, a reputation score is provided as output in region 604 of interface 600. As another example, scoring engine 702 can be configured to send reputation scores to users via email (e.g., via alerter 432).

Enterprise Reputation Information

As explained above, in addition to providing reputation information for single location businesses, such as Bob's Juice Company, platform 102 can also provide reputation information for multi-location businesses (also referred to herein as “enterprises”). Examples of enterprises include franchises, chain stores, and any other type of multi-location business. The following section describes various ways that enterprise reputation information is made available by platform 102 to users, such as Alice, who represent such enterprises.

FIG. 9 illustrates an example of an interface as rendered in a browser. In particular, Alice is presented with interface 900 after logging in to her account on platform 102 using a browser application on client 108. Alice can also reach interface 900 by clicking on tab option 902. By default, Alice is presented in region 912 with a map of the United States that highlights the average performance of all ACME locations within all states. In various embodiments, other maps are used. For example, if an enterprise only has stores in a particular state or particular county, a map of that state or county can be used as the default map. As another example, a multi-country map can be shown as the default for global enterprises. Legend 914 indicates the relationship between state color and the aggregate performance of locations in that states. Controls 928 allow Alice to take actions such as specifying a distribution list, printing the map, and exporting a CSV file that includes the ratings/reviews that power the display.

Presented in region 916 is the average reputation score across all 2,000 ACME stores. Region 918 indicates that ACME stores in Alaska have the highest average reputation score, while region 920 indicates that ACME stores in Nevada have the lowest average reputation score. A list of the six states in which ACME has the lowest average reputation scores is presented in region 922, along with the respective reputation scores of ACME in those states. The reputation scores depicted in interface 900 can be determined in a variety of ways, including by using the techniques described above.

The data that powers the map can be filtered using the dropdown boxes shown in region 904. The view depicted in region 906 will change based on the filters applied. And, the scores and other information presented in regions 916-922 will refresh to correspond to the filtered locations/time ranges. As shown, Alice is electing to view a summary of all review data (authored in the last year), across all ACME locations. Alice can refine the data presented by selecting one or more additional filters (e.g., limiting the data shown to just those locations in California, or to just those reviews obtained from site 110 that pertain to Nevada locations). The filter options presented are driven by the data, meaning that only valid values will be shown. For example, if ACME does not have any stores in Wyoming, Wyoming will not be shown in dropdown 910. As another example, once Alice selects “California” from dropdown 910, only Californian cities will be available in dropdown 930. To revert back to the default view, Alice can click on “Reset Filters” (926).

Some of the filters available to Alice (e.g., 908) make use of the tags that she previously uploaded (e.g., during account setup). Other filters (e.g., 910) are automatically provided by platform 102. In various embodiments, which filters are shown in region 904 are customizable. For example, suppose ACME organizes its stores in accordance with “Regions” and “Zones” and that Alice labeled each ACME location with its appropriate Region/Zone information during account setup. Through an administrative interface, Alice can specify that dropdowns for selecting “Region” and “Zone” should be included in region 904. As another example, Alice can opt to have store manager or other manager designations available as a dropdown filter. Optionally, Alice could also choose to hide certain dropdowns using the administrative interface.

Suppose Alice would like to learn more about the reputation of ACME's California stores. She hovers (or clicks) her mouse on region 924 of the map and interface 900 updates into interface 1000 as illustrated in FIG. 10, which includes a more detailed view for the state. In particular, pop-up 1002 is presented and indicates that across all of ACME's California stores, the average reputation score is 3. Further, out of the 24 California cities in which ACME has stores, the stores in Toluca Lake, Studio City, and Alhambra have the highest average reputation scores, while the stores in South Pasadena, Redwood City, and North Hollywood have the lowest average reputation scores. Alice can segment the data shown in interface 1000 by selecting California from dropdown 1006 and one or more individual cities from dropdown 1004 (e.g., to show just the data associated with stores in Redwood City).

Alice can view more detailed information pertaining to reviews and ratings by clicking tab 932. Interface 1100 makes available, in region 1102, the individual reviews collected by platform 102 with respect to the filter selections made in region 1104. Alice can further refine which reviews are shown in region 1102 by interacting with checkboxes 1112. Summary score information is provided in region 1106, and the number of reviews implicated by the filter selections is presented in region 1108. Alice can select one of three different graphs to be shown in region 1110. As shown in FIG. 11, the first graph shows how the average rating across the filtered set of reviews has changed over the selected time period. If Alice clicks on region 1114, she will be presented with the second graph. As shown in FIG. 12, the second graph shows the review volume over the time period. Finally, if Alice clicks on region 1116, she will be presented with the third graph. As shown in FIG. 13, the third graph shows a breakdown of reviews by type (e.g., portion of positive, negative, and neutral reviews).

If Alice clicks on tab 934, she will be presented with interface 1400 of FIG. 14, which allows her to view a variety of standard reports by selecting them from regions 1402 and 1406. Alice can also create and save custom reports. One example report is shown in region 1404. In particular, the report indicates, for a given date range, the average rating on a normalized (to 5) scale. A second example report is shown in FIG. 15. Report 1500 depicts the locations in the selected data range that are declining in reputation most rapidly. In particular, what is depicted is the set of locations that have the largest negative delta in their respective normalized rating between two dates. A third example report is shown in FIG. 16. Report 1600 provides a summary of ACME locations in a list format. Column 1602 shows each location's average review score, normalized to a 5 point scale. Column 1604 shows the location's composite reputation score (e.g., computed using the techniques described in conjunction with FIG. 7). If desired, Alice can instruct platform 102 to email reports such as those listed in region 1402. In particular, if Alice clicks on tab 940, she will be presented with an interface that allows her to select which reports to send, to which email addresses, and on what schedule. As one example, Alice can set up a distribution list that includes the email addresses of all ACME board members and can further specify that the board members should receive a copy of the “Location vs. Competitors” report once per week.

If Alice clicks on tab 936, she will be presented with interface 1700, depicted in FIG. 17. Interface 1700 shows data obtained from platform 102 by social sites such as sites 120-122. As with the review data, Alice can apply filters to the social data by interacting with the controls in region 1702 and can view various reports by interacting with region 1704.

Requesting Reviews

If Alice clicks on tab 938, she will be presented with the interface shown in FIG. 18, which allows her to send an email request for a review. Once an email has been sent, the location is tracked and available in interface 1900, shown in FIG. 19.

Although the foregoing embodiments have been described in some detail for purposes of clarity of understanding, the invention is not limited to the details provided. There are many alternative ways of implementing the invention. The disclosed embodiments are illustrative and not restrictive.

Claims

1. A method, comprising: extracting, over a network, reputation data from a plurality of external review websites, wherein the reputation data is associated with a plurality of physical locations of an enterprise entity, wherein the reputation data represents one or more reviews authored by one or more individual users regarding the plurality of physical locations, and wherein the one or more reviews include raw scores;wherein the reputation data is extracted at least in part by a plurality of instances of different types of helpers that are executed to obtain information, over the network, from the plurality of external review websites, wherein each type of helper is configured with instructions to fetch reputation data from a particular external review website, wherein, for a first external review website for which reputation data is available via an API, an instance of a first type of helper configured with instructions to obtain review data from the first external review website using the API is executed, and wherein, for a second external review website for which reputation data is not available via an API, an instance of a second type of helper configured with instructions to scrape reputation data from the second external review website is executed;storing at least some of the extracted reputation data to a data store configured to store heterogeneous data records, wherein reputation data for at least some of the external review websites includes reviews of different types that include scores on different scales, and wherein, for a given review, a raw score of the given review and a converted score are stored in the data store;generating individual online reputation scores for each individual location in the plurality of physical locations of the enterprise entity, wherein an online reputation score for a given physical location is generated at least in part by: obtaining a portion of the reputation data that pertains to the given physical location, wherein the portion of the reputation data that pertains to the given physical location includes at least one of a raw score and a converted score associated with a review for the given physical location; andevaluating the portion of the reputation data that pertains to the given physical location according to a plurality of factors, wherein the online reputation score for the given physical location comprises a composite online reputation score that comprises a composite of one or more scores associated with the plurality of factors according to which the portion of the reputation data that pertains to the given physical location was evaluated;receiving, via a user interface, a user interaction with a region of a map, and in response to receiving the user interaction, identifying, based at least in part on tags associated with the plurality of physical locations of the enterprise entity, at least two physical locations of the enterprise entity corresponding to the region of the map;subsequent to identifying, in response to the user interaction with the region of the map and based at least in part on the tags associated with the plurality of physical locations of the enterprise entity, the at least two physical locations of the enterprise entity corresponding to the region of the map, determining an aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity, and wherein determining the aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity includes obtaining individual online reputation scores that were previously generated for the identified at least two physical locations based at least in part on the extracted reputation data representing the one or more user-authored reviews, and combining the obtained individual online reputation scores; andrefreshing the user interface with a summary view associated with the region of the map, wherein the summary view comprises: the aggregate online reputation score determined for the identified at least two physical locations of the enterprise entity identified as corresponding to the region of the map; andan indication of a sub-region in the region that includes one or more physical locations of the enterprise entity with one of a lowest or highest average online reputation score.

2. The method of claim 1 wherein at least some of the tags are provided by an administrator associated with the enterprise entity.

3. The method of claim 1, further comprising accessing information associated with the plurality of physical locations of the enterprise entity including, for a particular physical location in the plurality of physical locations, at least one of the following: a name, a physical address, a type of business, and one or more URLs corresponding to profiles associated with a location on one or more external review websites.

4. The method of claim 1, further comprising accessing information associated with the plurality of physical locations of the enterprise entity that includes an uploaded file.

5. The method of claim 1, wherein the plurality of factors includes at least two of the following: an authority of a domain on which a review appears, an authority of a reviewer who authored the review, social feedback associated with the review, length of the review, an age of the review, and a physical position of the review on a corresponding external review website.

6. A system, comprising: one or more processors configured to: extract, over a network, reputation data from a plurality of external review websites, wherein the reputation data is associated with a plurality of physical locations of an enterprise entity, wherein the reputation data represents one or more reviews authored by one or more individual users regarding the plurality of physical locations, and wherein the one or more reviews include raw scores;wherein the reputation data is extracted at least in part by a plurality of instances of different types of helpers that are executed to obtain information, over the network, from the plurality of external review websites, wherein each type of helper is configured with instructions to fetch reputation data from a particular external review website, wherein, for a first external review website for which reputation data is available via an API, an instance of a first type of helper configured with instructions to obtain review data from the first external review website using the API is executed, and wherein, for a second external review website for which reputation data is not available via an API, an instance of a second type of helper configured with instructions to scrape reputation data from the second external review website is executed;store at least some of the extracted reputation data to a data store configured to store heterogeneous data records, wherein reputation data for at least some of the external review websites includes reviews of different types that include scores on different scales, and wherein, for a given review, a raw score of the given review and a converted score are stored in the data store;generate individual online reputation scores for each individual location in the plurality of physical locations of the enterprise entity, wherein an online reputation score for a given physical location is generated at least in part by: obtaining a portion of the reputation data that pertains to the given physical location, wherein the portion of the reputation data that pertains to the given physical location includes at least one of a raw score and a converted score associated with a review for the given physical location; andevaluating the portion of the reputation data that pertains to the given physical location according to a plurality of factors, wherein the online reputation score for the given physical location comprises a composite online reputation score that comprises a composite of one or more scores associated with the plurality of factors according to which the portion of the reputation data that pertains to the given physical location was evaluated;receive, via a user interface, a user interaction with a region of a map, and in response to receiving the user interaction, identify, based at least in part on tags associated with the plurality of physical locations of the enterprise entity, at least two physical locations of the enterprise entity corresponding to the region of the map;subsequent to identifying, in response to the user interaction with the region of the map and based at least in part on the tags associated with the plurality of physical locations of the enterprise entity, the at least two physical locations of the enterprise entity corresponding to the region of the map, determine an aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity, and wherein determining the aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity includes obtaining individual online reputation scores that were previously generated for the identified at least two physical locations based at least in part on the extracted reputation data representing the one or more user-authored reviews, and combining the obtained individual online reputation scores; andrefresh the user interface with a summary view associated with the region of the map, wherein the summary view comprises: the aggregate online reputation score determined for the identified at least two physical locations of the enterprise entity identified as corresponding to the region of the map; andan indication of a sub-region in the region that includes one or more physical locations of the enterprise entity with one of a lowest or highest average online reputation score; anda memory coupled to the one or more processors and configured to provide the one or more processors with instructions.

7. The system of claim 6 wherein at least some of the tags are provided by an administrator associated with the enterprise entity.

8. The system of claim 1 wherein the one or more processors are configured to, in response to determining that the aggregate online reputation score has changed by more than a predefined threshold value, provide as output a report that indicates whether the aggregate online reputation score represents an increase or a decrease over a previously determined aggregate online reputation score.

9. The system of claim 1 wherein the one or more processors are further configured to make a recommendation to a user based at least in part on the aggregate online reputation score.

10. The system of claim 9 wherein the recommendation comprises a recommendation that the user solicit a review from a customer.

11. The system of claim 6, wherein the one or more processors are further configured to access information associated with the plurality of physical locations of the enterprise entity including, for a particular physical location in the plurality of physical locations, at least one of the following: a name, a physical address, a type of business, and one or more URLs corresponding to profiles associated with a location on one or more external review websites.

12. The system of claim 6, wherein the one or more processors are further configured to access information associated with the plurality of physical locations of the enterprise entity that includes an uploaded file.

13. The system of claim 6, wherein the plurality of factors includes at least two of the following: an authority of a domain on which a review appears, an authority of a reviewer who authored the review, social feedback associated with the review, length of the review, an age of the review, and a physical position of the review on a corresponding external review website.

14. The system of claim 6 wherein subsequent to the user interaction with the region of the map, the one or more processors are further configured to render in the user interface a graph generated using reputation data corresponding to the identified at least two physical locations of the enterprise entity.

15. A computer program product embodied in a tangible and non-transitory computer readable storage medium and comprising computer instructions for: extracting, over a network, reputation data from a plurality of external review websites, wherein the reputation data is associated with a plurality of physical locations of an enterprise entity, wherein the reputation data represents one or more reviews authored by one or more individual users regarding the plurality of physical locations, and wherein the one or more reviews include raw scores;wherein the reputation data is extracted at least in part by a plurality of instances of different types of helpers that are executed to obtain information, over the network, from the plurality of external review websites, wherein each type of helper is configured with instructions to fetch reputation data from a particular external review website, wherein, for a first external review website for which reputation data is available via an API, an instance of a first type of helper configured with instructions to obtain review data from the first external review website using the API is executed, and wherein, for a second external review website for which reputation data is not available via an API, an instance of a second type of helper configured with instructions to scrape reputation data from the second external review website is executed;storing at least some of the extracted reputation data to a data store configured to store heterogeneous data records, wherein reputation data for at least some of the external review websites includes reviews of different types that include scores on different scales, and wherein, for a given review, a raw score of the given review and a converted score are stored in the data store;generating individual online reputation scores for each individual location in the plurality of physical locations of the enterprise entity, wherein an online reputation score for a given physical location is generated at least in part by: obtaining a portion of the reputation data that pertains to the given physical location, wherein the portion of the reputation data that pertains to the given physical location includes at least one of a raw score and a converted score associated with a review for the given physical location; andevaluating the portion of the reputation data that pertains to the given physical location according to a plurality of factors, wherein the online reputation score for the given physical location comprises a composite online reputation score that comprises a composite of one or more scores associated with the plurality of factors according to which the portion of the reputation data that pertains to the given physical location was evaluated;receiving, via a user interface, a user interaction with a region of a map, and in response to receiving the user interaction, identifying, based at least in part on tags associated with the plurality of physical locations of the enterprise entity, at least two physical locations of the enterprise entity corresponding to the region of the map;subsequent to identifying, in response to the user interaction with the region of the map and based at least in part on the tags associated with the plurality of physical locations of the enterprise entity, the at least two physical locations of the enterprise entity corresponding to the region of the map, determining an aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity, and wherein determining the aggregate online reputation score associated with the identified at least two physical locations of the enterprise entity includes obtaining individual online reputation scores that were previously generated for the identified at least two physical locations based at least in part on the extracted reputation data representing the one or more user-authored reviews, and combining the obtained individual online reputation scores; andrefreshing the user interface with a summary view associated with the region of the map, wherein the summary view comprises: the aggregate online reputation score determined for the identified at least two physical locations of the enterprise entity identified as corresponding to the region of the map; andan indication of a sub-region in the region that includes one or more physical locations of the enterprise entity with one of a lowest or highest average online reputation score.