Case Study: LinkedIn

LinkedIn is a professional social network geared toward promoting business and employment-focused services. The company was founded in 2002¹ and it ranks 58th for global internet engagement on Alexa rankings.² In December 2016, Microsoft acquired LinkedIn for $26.2 billion.

An examination of LinkedIn offers an interesting perspective on how automated tools can be used for ad targeting and delivery, because prior to LinkedIn’s acquisition, advertising was not its primary source of revenue. The company also relied on other sources of revenue such as premium membership subscriptions and its learning solutions, which enable users to learn new skills and subject matter.

The company’s ad operations transformed significantly after its acquisition. Before the deal, marketing and advertising solutions accounted for approximately 18 percent of LinkedIn’s total revenue, at $175 million. In comparison, talent solutions for recruiters accounted for approximately 65 percent of LinkedIn’s total revenue at $623 million and premium subscriptions accounted for approximately 17 percent of LinkedIn’s total revenue at $162 million.³ Under Microsoft, little data has been available regarding the size and scope of LinkedIn’s advertising business in comparison to other revenue streams. However, LinkedIn shared that it expected to bring in approximately $2 billion from its media business by the end of 2018, with the majority of this revenue coming from advertising.⁴

Since its acquisition by Microsoft, LinkedIn has expanded its data collection practices. In May 2019, the company acquired Drawbridge, a machine-learning identity management company, with the aim of enhancing ad personalization⁵ through greater cross-device targeting.⁶ Prior to this, LinkedIn primarily relied on methods such as cookie-based retargeting. However, as browsers (such as Google Chrome) have limited this functionality, this has decreased the effectiveness of these methods.⁷ LinkedIn’s expansion of data collection practices and advertising operations suggests that they are both vital components of growing the company’s ad operations, and thus its revenue.

LinkedIn offers advertisers a range of methods through which they can engage with users: These include:⁸

1. Text ads: Text ads are text-based ads that link to an external website, such as a company’s website. The ads appear in the right panel or top banner of numerous pages on LinkedIn.com, and they can include images (such as an advertiser’s logo). Text ads can appear on both the desktop website and the mobile app.
2. Sponsored Content: Sponsored Content is a post that is promoted by a company that can appear on both the desktop website and the mobile app. Sponsored Content appears in a user’s LinkedIn feed and it can feature text, an image or a video, and links to a website. Sponsored Content is always labeled as “Promoted” or “Sponsored.”
3. Sponsored InMail: Sponsored InMail is a sponsored message that a user receives from an advertiser in their LinkedIn messaging inbox. The message is labeled as “Sponsored” in the subject line. These ads appear on both desktop and mobile devices.
4. Dynamic ads: Dynamic ads are personalized ads that are tailored based on a user’s profile image, name, and job function. Dynamics ads can cover advertisers, jobs or professional content that a user may be interested in. Dynamic ads appear in the right panel of LinkedIn.com pages on desktop devices when a member is signed into their LinkedIn account.
5. Display ads: Display ads are image-based ads that appear in the right panel of LinkedIn.com pages on desktop devices.
6. LinkedIn Audience Network: Most of the ads that LinkedIn places appear on LinkedIn.com. However, an advertiser can also opt to join the LinkedIn Audience Network, which enables ads to appear on third-party publisher websites that LinkedIn characterizes as “brand-safe premium publisher apps and sites.” These are apps and sites that LinkedIn believes its members would benefit from viewing and that LinkedIn has deemed are in compliance with their brand safety protocols and requirements.⁹ According to LinkedIn, during this process, no data that can be used to identify users is transferred to the third-party websites.¹⁰

Just as on Facebook and Google, when an advertiser creates an ad campaign on LinkedIn, they have to select the type of ad they want to run, define their business goals, select their audience based on targeting parameters, and set a budget for their campaign. Advertisers can also choose to use LinkedIn’s Matched Audiences feature to engage with users who have shown an interest in their company before by visiting their website or sharing their email address.¹¹ Once LinkedIn users see or interact with an ad, the advertiser receives an aggregated report with information on user impressions. LinkedIn asserts that it does not provide advertisers with data on specific users, and it does not provide data that could be used to identify specific users.¹² Rather, the company states that it provides advertisers with aggregated “general demographic information” on all of the users who have seen an ad.¹³ This information could include, for example, the cities that users who engaged with the ads live in.

LinkedIn’s advertising auction system operates similarly to those of Google and Facebook. Advertiser’s can bid per click (CPC) or per thousand impressions (CPM), depending on the advertiser’s campaign objectives. Bidding on LinkedIn’s advertising products operates in a second-price auction manner. This means that if an advertiser wins with a bid they placed, they only need to pay the minimum amount that is needed to beat the second-place bidder, no matter what their maximum bid was. LinkedIn also calculates and considers a relevance score when determining which ads a user should see. This score measures how engaging a given update is for the targeted audience, and it is calculated using a combination of click-throughs, comments, likes, shares, and other data points. By factoring in such a relevance score, LinkedIn aims to deliver relevant advertisements to its users, thus incentivizing them to spend more time on the platform.¹⁴

Since its acquisition in 2016, the company has introduced a host of new advertising features that aim to broaden targeting and delivery opportunities. For example, over the course of 2019, the company expanded its targeting options to include categories such as interest-based and Lookalike Audiences (similar to Facebook’s Lookalike Audiences feature). This is in addition to existing parameters such as company affiliation, demographics, education, job experience, and the advertiser’s own contact lists.¹⁵

Currently, the only targeting parameter LinkedIn requires advertisers to choose is location. Permanent and long-term location data is collected from the locations users specify in their profile. This information is used in advertising campaigns, such as Sponsored InMail campaigns. Short-term location data is inferred from a user’s IP address.¹⁶ While continuous and granular location tracking can result in the collection of highly personal and sensitive data, collecting a user’s single, static location raises fewer privacy concerns. Moreover, since a user’s location in their profile is likely relevant to which jobs may interest them, this information has clear relevance to LinkedIn’s primary function of connecting users with job opportunities, and it permits the platform to deliver specific employment-related results.

In addition, because LinkedIn is an employment-focused website, the company has only narrowly permitted targeting based on sensitive and protected characteristics. For example, the platform only permits gender-based targeting in specific instances, such as for women-oriented events and conferences. In addition, all ads that are based on gender-related targeting parameters go through a specialized review process, which can be automated or manual depending on the type of ad and the advertiser’s history on LinkedIn’s ad network.¹⁷

One component of this review process requires advertisers to self-certify on LinkedIn’s Campaign Manager platform that, if an ad is related to employment, housing, education, or credit, it will not use LinkedIn to discriminate based on gender or other protected characteristics. Once an advertiser self-affirms that they are not intending to discriminate based on gender, the option to target their ads by gender becomes available to them.¹⁸ While this self-certification process aims to highlight the importance of—and ensure equitable procedures during—ad targeting, it does not provide strong safeguards against discrimination based on gender and other protected characteristics. An advertiser could easily self-certify based on false information or intentions, and continue to implement discriminatory targeting and delivery outcomes on the platform.

Since its acquisition, LinkedIn has adopted a more aggressive and streamlined targeted advertising model. These advertising practices have yielded some concerning outcomes. For example, a 2017 investigation by ProPublica and the New York Times found that major companies such as Amazon, T-Mobile, and Goldman Sachs were using platforms such as Facebook, Google, and LinkedIn to run ad campaigns that explicitly excluded audiences over the age of 40 years old from viewing and responding to employment-related ads.¹⁹ Such pratices violate the Age Discrimination in Employment Act (ADEA), which prohibits employment discrimination against people over the age of 40.²⁰

In response to the investigation, LinkedIn changed its ad targeting offerings so that age-related targeting for employment ads is only available to advertisers who affirm through a self-certification process that they are not running age-discriminatory ads.²¹ If the campaign's objective is to generate talents leads and recruitment, LinkedIn does not offer targeting based on age at all.²² Additionally, the self-certification process advertisers need to go through is similar to the self-certification process advertisers need to go through when targeting based on gender. Therefore, while it demonstrates an attempt to provide greater accountability around LinkedIn’s ad targeting and delivery practices, it does not establish sufficient and concrete safeguards.

Another troubling example emerged in April 2019. Cynet Systems, a technology recruiting firm based in Virginia, posted an employment ad on LinkedIn and other platforms for an account manager. The ad listed a preference for a “Caucasian who has good technical background.” The company also posted an ad for an account manager, soliciting “female candidates only.”²³ According to a LinkedIn spokesperson, the company removed these ads once they were discovered. LinkedIn also stated that such postings were not frequent.²⁴ The platform’s Ad Targeting Discrimination guidelines outline that ads that are flagged as discriminatory are instantly removed.²⁵ However, there is little transparency around the scope and volume of ads that are removed for violating LinkedIn’s advertising policies, as well as little transparency around why ads are removed.

Compared to Google and Facebook, LinkedIn played a smaller role in hosting campaigns to influence voter behavior during the 2016 U.S. presidential elections. However, Russian operatives were found to have used the platform to create fake accounts, aiming to connect with individuals who were part of U.S.-based political organizations.²⁶

In 2018, LinkedIn banned political advertising on its platform entirely. Although this is not an insignificant decision, the ramifications of this decision for the platform’s overall functionality and advertising revenue stream are small compared to the impact that platforms such as Google and Facebook would face if they made a similar decision. This is because LinkedIn’s services focus on professional—rather than personal—social networking, and as a result political advertising is primarily useful for politically-driven groups looking to use the platform to recruit members rather than expressly spread their message. In addition, in May 2019, the platform also introduced an Ad Transparency tool for LinkedIn company pages that enables LinkedIn users to view all of the Sponsored Content that advertisers have run on LinkedIn in the past six months when visiting a company’s LinkedIn page²⁷

The company does not currently have an ad transparency database for any type of ads. In addition, although the company does publish a transparency report that highlights the scope and scale of its content moderation efforts to an extent, it does not disclose any data on how the platform enforces its ad content and targeting policies. Nor does it disclose how many ads and accounts are subsequently removed or suspended. As previously discussed, this is a vital mechanism for ensuring transparency and accountability around the enforcement of a platform’s ad-related policies.

Like Google and Facebook, LinkedIn offers its users a limited set of ad controls. These include whether a user receives targeted ads based on the websites they have visited, whether a user sees promoted jobs and ads on websites and applications that are separate from LinkedIn, how certain ads are targeted and delivered based on a user’s profile data, and whether a user views promoted jobs and ads based on their activity and similar members’ activities on both LinkedIn and Bing (Microsoft’s search engine).

Users also have the ability to control what data LinkedIn collects on them, to an extent. For example, users can opt out of having the company collect and subsequently target and deliver ads to them based on factors such as their connections, location, demographics, companies they follow, groups, education, and employer.

Users can also opt out of seeing targeted and promoted ads based on information provided to third parties and based on prior engagements by the user with ads. However, there is no single universal control for opting out of receiving advertising on the platform. In addition, LinkedIn states that even after a user opts out of certain forms of advertising, they “may continue to see ads that feature an option to manage your advertising preferences, including to opt out of interest-based advertising.”²⁸