January 21, 2020

Facebook, Google, and Twitter all maintain archives of political advertisements run on their platforms since mid-2018. Their shared stated goal is to provide transparency into political advertising.

We compare these ad transparency databases side-by-side. We seek to help journalists understand the varied comprehensiveness and functions of these ad transparency databases and help them navigate them. We hope that platforms can use this to design their databases better according to the needs of non-technical stakeholders. For researchers, we hope to provide a detailed look at what these ad databases compile so they can more efficiently direct their research, as well as provide a record of what these ad libraries looked like and their functionality at this snapshot in time. Finally, we hope to provide some recommendations for what the standards governing ad transparency should be, with an eye towards US federal regulation in the future.

This page is organized into five sections. First, we present brief descriptions of each ad transparency archive. Next, we lay out our recommendations for platform ad transparency databases based on what would make advertisements run in digital, personalized media environments as transparent as traditional mass media advertisements. Third, we analyze what types of accountability and coverage Facebook and Google’s ad libraries enable and what they constrain. After these take-aways, we lay out the data that allowed us to draw our conclusions beginning with what the platforms seem to mean by “transparency” and followed by in-depth descriptions of the functionality of each ad archive.

The information presented here was compiled from October-December 2019 by a class of undergraduate students at the Hussman School of Media and Journalism at the University of North Carolina at Chapel Hill and analyzed by CITAP researchers. To best capture the ways that most journalists, political practitioners, regulators, and the public will encounter these databases, we examined the databases from the perspective of an average, non-technical user, and as such did not utilize any advanced APIs.

---

Part I: About the Databases

Facebook’s Ad Library contains all active advertisements run across Facebook and its subsidiaries, not just political ads. The non-political ads are no longer visible in the library once they are inactive. Political ads run since May of 2018, both active and inactive, will be retained and visible in the library for seven years. Facebook’s definition of “political” is quite broad, including social issues in any geographic area an ad is run. Since Facebook currently prohibits political advertising on Messenger and Facebook Audience Network, Facebook and Instagram are the primary conduits of political advertising in this library (some ads run on Messenger and Facebook Audience Network before being flagged as “political.” When this happens, they are placed in the library with a note that they ran without a disclaimer). The company also has a downloadable Ad Library Report which is updated daily and includes basic data aggregated by last day, week, month, 90 day period, and all-time.

Google’s Transparency Report includes advertisements from verified political advertisers running their campaigns through Google Ads or Google Display & Video 360. This includes display, search, and YouTube ads, as well as video ads run outside of YouTube. The current ad library sits within Google’s larger Transparency Report which includes data on government requests for users’ information, among other initiatives. Google’s definition of what constitutes political advertising is narrower than Facebook’s, limiting “political” ads to ads that reference state and national candidates, government office holders, political parties, and any issue on a state ballot. Like Facebook, users are able to view all political ads, both active and inactive, run since May of 2018.

Prior to its ban on political advertising, Twitter’s Ad Transparency Center included all promoted tweets run on the platform in the last seven days, a list of certified campaign advertisers, and a list of certified issue advertisers in the United States. The advertisements in the library could be more than a week old, but they must have been promoted within the last seven days in order to be viewable in the library. Twitter also made a distinction between promoted tweets (which are visible on an advertiser’s timeline) and promoted-only tweets (which are not visible on an advertiser’s timeline or in searches), both of which were included in the library. Promoted tweets that were suspended were in the library as well, along with the reason for suspension. Twitter’s political advertising library is now a static database of political advertising run prior to its ban.

(See our report on advertising on platforms for more details on how these platform companies define political advertising and what the implications are.)

---

Part II. Recommendations

Our recommendations are directed toward solving a specific problem that we believe platforms created ad libraries to address: countering the personalized information environments fostered by digital media and micro-targeted advertising. To solve this problem, we believe that platforms should provide the same level of transparency into political advertisements that journalists and the public had into older forms of media, specifically television and radio. This would allow the public to have greater visibility into the communications of campaigns, rivals for the same office and contending parties would be better able to contest one another’s claims, and journalists could hold political advertisers accountable for false or inflammatory appeals as effectively as on television.

These recommendations are aimed at creating a set of transparency standards for political advertising on platforms. We believe that Facebook and Google should, as much as possible, report the same categories of information about political advertising in the same ways. Other experts and researchers have put forth guidelines for ad archives based on what is needed for high-quality research and developed frameworks for platform transparency. We support these recommendations as well, but here focus on what forms of digital ad transparency are needed for non-technical users. Since Twitter no longer permits political ads these recommendations focus primarily on Facebook and Google.

To reach this standard of transparency in digital campaigning, there should be:

Visibility into local, state, and national politics. Ad watch coverage by journalists and the ability to hold advertisers accountable for their claims has never been limited to national politics. State and local races need the same level of scrutiny as national elections.

More transparency in political targeting. Traditional media was targeted primarily by where it was placed, such as the market, time, and television station or specific show the ad aired on. This gave journalists, rival campaigns, and viewers an implicit transparency into the strategies of political advertisers and provided for the default publicity of political advertising content. To reach this same level of transparency as traditional media, the targeted audience for digital political ads must be made explicit given that geographic location, time, page, or website rarely reveal how an ad was targeted or which actual audiences it was displayed to.

The grouping of multiple ads into the campaigns and ad groups they were run in. Unlike older forms of media, digital advertisements can be run with hundreds of versions of the same message. Advertisers organize and budget their ads in Facebook Ad Manager and Google Ads user interfaces at the Campaign and Ad Group level. Just as additional information on targeting must be disclosed to reach the same level of transparency as older media, so too must digital ads be clearly grouped together in order to provide the same level of interpretability. This can be achieved through transparency into what Ad Groups and Campaigns ads ran under and by grouping ads with the same visuals that were targeted to different audiences.

Given these standards for transparency, our recommendations are as follows.

To accomplish visibility into local politics in addition to state and national politics:

• Platform databases should reveal the actual geographic locales that political advertisers targeted and users encountered ads within on Facebook and Google, such as zip code, city, or congressional district.
• Platform ad databases should include an “ads running near me” filter, so users can see all the ads targeted to their communities.

To provide more information on targeting:

• All platforms should include more types of demographic and additional data advertisers use to target political advertising, as well as that of the audience that actually views the advertising. This includes contextual and interest-based targeting.
• All platforms should be clearer about which data refers to targeting and which is actual impression data.
• Facebook and Google should make data on the language ads use available similar to how Twitter does.
• Facebook should, at a minimum, declare whether or not political advertisers used their own data to create targeting universes and require advertisers to provide a public rationale explaining on what basis they were created. Facebook allows advertisers to use their own or outside data to create custom and ‘look-alike’ audiences. Because of this, there is little transparency around the actual ways that advertisers target political advertising. Meanwhile, Facebook chooses to categorize impression data in ways (gender, state, and age) that often say little meaningful about how advertisers target ads or the demographic, interest, attitudinal, or behavioral characteristics of those who saw them.

In order to make online advertisements as interpretable as traditional media:

• Platforms should organize ads by the Campaign and Ad Groups the advertiser ran them in.
• Like Facebook, Google should group variations of the same ad that are targeted or tested on various audiences. 
• Because the terminology around digital advertising is new, Facebook and Google should clarify their definitions of terms within their ad transparency databases.

To provide more information with which to identify bad actors:

• Google should include the click through URLs used in ads (the websites that ads drive to) as Facebook and Twitter do.
• As much as possible, information contained in the Google CSV files should be available in the online portal where it is more accessible to users with few technical skills.
• Google should work to ensure that all ads, even commercial ads, are displayed in its ad library, like Facebook. This is especially important given that Google has a more limited definition of ‘political’ ads than Facebook.
• Twitter should continue to maintain its library of political and cause-based ads and supplement it by including ads that it rejects as political or as inappropriate caused-based content.
• All platforms should include and label ads that have been taken down due to policy violations, and provide a description as to why

Finally, the recommendations below are more generally about functionality and improving the information that these platforms currently include:

• Search bars should clearly specify what a user can search for and, if they cannot search for it, where to find other data.
• Facebook should create unique dedicated URLs for its detailed views of individual ads to enable stakeholders to easily share them. Currently, copying a URL will simply revert a user back to a search page.
• All platforms should provide a comprehensive set of ad filters, including: active and inactive ads, date ranges of when ads were active, more fine-grained ranges of number of impressions and spending, type of targeting used (ex. geographic, contextual, behavioral, custom), and format of ad (ex. video, display, search, Instagram).
• Google should create smaller ranges for impression and spending data to allow users to sort ads more precisely and identify trends and strategies.
• Users should be able to access ad spend over any custom range of time on each platform, whether the user searches by campaign, geographic location, candidate, issue, or any other criteria available on each platform.
• All platforms should update their FAQ pages to provide more detailed information.
• Facebook and Twitter should follow Google’s lead by prioritizing Federal Election Commission (FEC) IDs and providing easy links to FEC filings.
• Google should provide more information in its FAQ page about how ads are targeted and where users can change their advertising targeting preferences.

---

Part III. What stories about elections do each platform enable to be told?

Through the data they make available, their visualizations, the search, filtering, sorting mechanisms, and general design of their ads transparency databases, each platform makes some stories about elections easier to tell than others.

Based on the functionalities outlined in detail in parts IV and V, here are the types of stories that are enabled and constrained by Facebook and Google:

The Possibilities and Limits of Facebook

Enables:

National politics: By organizing data by top spenders nation-wide and having limited filters for location, Facebook’s ad archive is easy to use for reporting aggregated national spend and analyzing top-dollar campaigns.

National political spending on Facebook has been covered extensively by The New York Times, ACRONYM’s weekly newsletter, and others.

State politics: Particularly with the recent addition of state-level targeting included in a recent ad library update (not covered comprehensively in the rest of this page), Facebook now enables the public to see what political campaigns specific to a state are active on Facebook.

Finding bad actors: By including all advertisements in the archive and providing the landing pages that ads drive to, Facebook provides a set of transparency tools for journalists and members of the public.

Reporting by ProPublica, Popular Information, Wired, and others have all used Facebook’s ad library to investigate misinformation in advertising, scams using political issues to draw victims in, and cases of Facebook failing to enforce its own rules.

Constrains:

Local politics: Facebook does not provide zip code, city, congressional district, or “near me” filters, making reporting on local and smaller-scale geographically targeted political ad campaigns difficult.

Data and privacy: By limiting the information about targeting, Facebook constrains the public’s ability to understand and report on issues of data and privacy in digital political advertising.

Effects: Facebook’s library does not include data on likes, shares, or interactions. This is notable because Facebook and Instagram are social networks and these social interactions are visible to the users who see the ads. In addition, Facebook does not include information on the goals of the advertisers and outcomes in relation to ads, such as donations to a campaign. Such detail is not necessarily expected, but since the only metrics provided are spend and impressions, it is important to note that neither are representative of the effects that campaigns seek to elicit.

The Possibilities and Limits of Google

Enables:

National politics: Like Facebook, Google’s ad library biases towards aggregated, national spend statistics.

Reporting from CBS News, ACRONYM, and others have used Google’s data to report on national campaign’s digital advertising.

State politics. Google’s CSV files include state-level geographic targeting allowing for reporting on state-level political campaigns in addition to national.

Systematic comparison of mid-level strategies: Like Facebook, due to the broad ranges of spend and impressions and the inability to filter or sort by groups of similar ads, analyzing messaging strategies systematically is possible, but difficult and time-consuming.

Constrains:

Local politics: Google does not provide zip code, city, congressional district, or “near me” filters in the user interface of the its transparency report, but does include information about geographic targeting in the CSV files. However, the difficulties presented by finding individual ads in the CSV report and then searching for them in the web interface ads significant friction to reporting on ads about local issues.

Data and privacy: By limiting the information about targeting, Google constrains the public’s ability to understand and report on issues of data and privacy in digital political advertising. Google has recently limited what data political advertisers are allowed to use, making this less relevant. However, including more information on advertiser’s targeting strategies would allow the public more insight into the strategic communication attempts of campaigns.

Finding bad actors: Google is the only ad library that does not include access to the websites that ads drive to. Meaning, users can see an advertisement in the Google archive but they cannot click on it. In addition, Google’s inclusion of only narrowly defined political ads prevents the public from checking if ads not flagged as political should be. Altogether, Google severely limits the public’s ability to hold political advertisers (or Google itself) accountable.

Effects: Google does not include data on the goals advertisers have for political ads, such as donations, or engagement data. Since the only metrics provided are spend and impressions, it is important to note that neither spend nor impressions is representative of effects. In fact, “impressions” do not even mean the number of times an ad was viewed. For instance, display ads run on desktop computers through Google have an average viewability of 42%, meaning that 58% of desktop display ads purchased never met the standard definition of viewable (50% of the ad on screen for one second or more).

Systematic comparison of mid-level strategies: Like Facebook, due to the broad ranges of spend and impressions and the inability to filter or sort by groups of similar ads, analyzing messaging strategies systematically is possible, but difficult and time-consuming.

Ultimately, platforms thus far have, intentionally or not, enabled stories about national politics and constrained the public’s ability to hold state and local political advertisers accountable. While data privacy and the ethics of targeted advertisements are both matters of public interest, neither Facebook nor Google enable journalists or everyday users to understand how such issues are playing out in digital political advertising. Through the provision of outside links, Facebook has enabled the public to find bad actors such as scammers and purveyors of misinformation; Google has not.

---

In parts IV and V we provide detailed descriptions of Facebook, Google, and Twitter’s political ad archives from which we drew our recommendations and conclusions above. Part IV unpacks what types of information are currently available and made “transparent” in the ad archives while part V more broadly examines the usability of the ad library interfaces.

---

Part IV: What is transparency according to platforms?

Facebook, Google, and Twitter provide transparency into political advertising in three fundamental ways: 1) providing information about how advertising on their platform works, 2) providing information about political advertisers, and 3) providing visuals and details on the ads themselves, including who they reached or were supposed to reach depending on the platform.

How advertising works

One of the ways these companies aim to be more transparent about political advertising is by providing information about their policies and capabilities for political advertising. Such information is found on “about” pages, “frequently asked questions” (FAQ) pages, and links to further “help” centers.

In our estimation, Facebook prioritizes transparency into the advertising process the most out of the three platforms. Facebook has a page called “About Facebook Ads” which is accessible at the bottom of the library homepage. This page details what data advertisers utilize to show users ads and provides a link to managing users’ ad preferences. There is a step-by-step process outlining how advertisers reach users in accessible language: “We take the advertiser’s goal, desired audience and ad to show you ads that we think might be relevant to you, without advertisers knowing who you are – or selling your data to the advertiser.”

Google’s FAQ page drives to the political content policy which explains in detail the types of data advertisers can and cannot use to target users and where ads can and cannot appear.

The first link Twitter’s ad library page drives to a FAQ page on promoted-only tweets. This page briefly describes how ads (promoted tweets) are run on Twitter.

Information on political advertisers

In addition to providing information about how advertising is run on platforms, the ad libraries provide information about political advertisers.

Facebook provides Federal Election Commission (FEC) ID numbers when the advertiser used its FEC identification to verify its identity. Facebook includes other information if the advertisers choose to verify themselves through other means. When viewing an individual ad in Facebook’s ad library, users see a tab labeled “About the disclaimer” with an included link to a page titled  “How are ads about social issues, elections and politics identified on Facebook?” While this page does not explain that advertisers can use different means of verification, Facebook also has a specific page dedicated to “How disclaimers work for ads about social issues, elections or politics.” For each individual ad in the library beneath the “About the disclaimer” tab there is another labeled “Information from the advertiser.” This discloses the name of the group or person paying for the ad with their FEC ID number or other verified information.

When viewing ads by advertiser in Facebook’s ad library, on each advertiser’s page there is a large banner at the top showing a basic overview of the page with information such as when the page was created, how many likes it has, and the page owner’s country of origin. The header banner also includes a section on spending which outlines the “total spent,” as well as the money “recently spent” in the past seven days on advertisements.

Google similarly uses FEC numbers to verify advertisers and allows users to view advertising data by advertiser. Once Google verifies the advertiser and the ad runs, the ad library provides users with the advertiser’s name and their FEC ID number. The FEC ID number displayed is linked to the advertiser’s financial summary on the FEC’s website. Thus, a viewer could click this link to learn more about the campaign’s overall fundraising and spending. In addition to this disclaimer, Google provides users with the number of ads purchased, the amount of money spent on those ads, and the amount spent per week by the advertiser. 

Twitter provides billing information about political advertisers, but never made FEC ID numbers available to the public in its transparency center. Twitter did, however, include total spending by each advertiser.

Information about the advertiser from Twitter, Google, and Facebook’s advertising libraries.

The ads themselves

Finally, and most importantly, the ad libraries’ transparency includes visuals of the ads themselves and basic information about the targeting or reach of the ads. All three libraries display the visuals of the ads. However, not all display meaningful targeting information or even the click through links that the advertisements drove to.

Prior to its new policy banning political ads, Twitter pledged to be transparent by making its Ads Transparency Center easy to use and understand. In terms of transparency into the ads themselves and their targeting, in our estimation Twitter succeeded. (The significant limitations of the interface for other purposes will be discussed later.)

When a user clicked onto a specific ad campaign from a particular advertiser, there was a link to extensive targeting information in the sidebar. The linked page explained how advertisers were able to target their advertisements based on demographics, tailored audiences, and audience features such as conversation, keyword, and follower look-alike targeting. Additionally, there was a link to Twitter’s Ad Transparency FAQs on almost every page, which housed a specific political promoted tweets section with detailed information on the product. Notably, Twitter’s Ad Transparency Center offered detailed explanations for each category of information provided, including the campaign performance summary, the targeted audience, and the actual audience. Twitter clearly explained the difference between the targeted audience and the actual audience reached. The pages that each advertisement drove to were easily accessible.

Compared to Twitter, both Facebook and Google provided less information about individual ads themselves. Within its ad library, Facebook shows the actual number of impressions delivered to demographic groups the company itself specifies, not who the advertiser was attempting to reach. When looking at an advertiser’s page for a specific ad, users can see “Who was shown this ad” broken down by age ranges and gender – Facebook’s own categories on impressions that the company chooses to make public. Additionally, users can examine “Where this ad was shown” through an interactive map that shows users which states ad viewers live in. On Facebook, the pages that each advertisement drives to are also easily accessible by clicking on ads.

However, we wish to emphasize again that this data does not show targeting, nor does it provide a more detailed description of who advertisers were hoping to reach or who was actually reached beyond the categories – state, age range, and binary gender – that Facebook itself chooses reveal. As such, it is ultimately a limited tool for understanding the intent, purposes, and effects of political advertising campaigns.

In its user-interface, Google includes no targeting information beyond the date range and the type of advertisement (search, display, or video). There is no way to access the page that the advertisement drove to when clicked.

Users can also download additional data about political advertising from both Facebook and Google in the form of comma-separated values (CSV) files. Google’s CSV, updated weekly, contains cumulative data compiled according to the launch date of an ad campaign for a country or region. Upon downloading this CSV, users can see additional categories of ad targeting. Google’s public-facing data includes age, gender, and geography. Within each row, Google lists whether or not a specific ad was targeted based on these three criteria, and if so, how they were targeted. In contrast, Facebook’s downloadable CSV does not present any information on ad targeting. 

Google’s Downloadable CSV

Facebook’s Downloadable CSV

---

A closer look at demographics:

Facebook:

Twitter

Google:

---

Part V: Ad library functionality

In this section, we detail and review the three ad transparency databases based upon their usability. All references to Twitter within the section refer to the company’s platform prior to its ban on political advertising.

The ad libraries share similar features that hinder ease-of-use and make interpretation of the data presented difficult. A common difficulty is the lack of ease in being able to find all of the detailed information a user might desire. For example, in each library more in-depth data is either buried within a page about a single ad or within CSV files, or not made available on the platform at all. In addition, Facebook, Google, and Twitter all make it difficult to find a specific ad or see ads that are the same in terms of content but which are targeted to different audiences. And, overall, the differing ways that all three platforms report data make it difficult for users to compare political ad spending across platforms.

Facebook

The layout of Facebook’s Ad Library is accessible and relatively simple for an average non-technical user to navigate, including those who have never used an ad transparency database before. On the main page, there is a prominent search bar, a clear description of what the Ad Library is, an overview of which ads are available to be viewed, and a link to Facebook’s Ad Library Report, which is a “snapshot of data for ads about social issues, elections or politics.” The search bar is placed directly in the middle of the page, is large enough to draw the user’s attention, and allows the user to search ads by name, topic, or organization. The library is designed to sort ads primarily by advertiser or keyword.

To access the FAQ page about the Facebook Ad Library, users can click a link on the Facebook Ad Library main page. This link, “ad library,” looks as though it drives to the library itself, but instead leads users to a summary page that describes what the library is and how users can search within it. From here there are four drop-down tabs with information, one of which is titled “Viewing social issues, elections or politics.” This page details ways to search for ads and what information is shown on each individual ad as well as provides links to the Ad Library Report.

When viewing ads by keyword, the search can be limited to political advertisers, but one must specifically select “Issue, Electoral, or Political” on the search bar or it will search all Facebook ads. Users can then sort the ads through six filters: country; active or inactive ads; impressions in the last one, seven, thirty, ninety days or all time; by name of page; by the disclaimer included on the ad; and by platform (Facebook, Instagram, Messenger, or Audience Network). As noted above, the filters do not include any information related to targeting, only impressions, and do not allow users to search for an exact date or group similar or repeated ads.

When viewing ads by advertiser, there are many repeats of similar-looking ads that may be confusing for users to look at or differentiate between. Facebook does not group similar ads together which creates more work for users and a more confusing layout.

Before clicking on an ad, users can see an overview of the ad which shows whether the ad is active or inactive, the date that the ad started running, the ad’s ID number, what Facebook platform the ad was shown on, and what the ad looks like. There is also the option to “See Ad Details” which opens onto the impression data detailed above. When viewing ads based on a keyword search, it is unclear how they are organized: by spend, recency, subject, or something else.

Google

Google’s ad transparency database provides a general overview of data about political advertising on the platform to give users a sense of its overall use by political advertisers. Like Facebook, its library is designed to sort ads primarily by advertiser or keyword. Like Facebook, after searching a keyword, the presentation of ads is difficult to navigate because results from so many advertisers are displayed. For example, searching Elizabeth Warren will result in ads paid for by the Warren campaign alongside the Democratic Senatorial Campaign Committee, National Rifle Association, and others. There is no option to filter by a specific advertiser. Meanwhile, results can be displayed a single advertiser, but there is no keyword search feature within the advertiser’s page. 

• 

Ads can be filtered by when they ran, how much the advertiser spent, how many impressions were delivered, and what format they were in (such as text ads in Google search, display ads, or videos). These results can be sorted by spend or impressions delivered. All of the filters narrow down information about ads and their reach. However, the ranges for impressions (10,000 – 100,000 impressions) and spending ($1,000 – $50,000) are quite large, making the filters less effective or helpful in sorting data into reasonably sized and logical categories that can be used to analyze data and draw conclusions. In addition, there are some ads that appear exactly the same in terms of content, but were targeted to different audiences; the lack of grouping these ads together leads to confusion.

Most ads can be viewed through the online portal, but some ads instead display the message “due to technical limitations, we are currently unable to display the content of the ad in the Transparency Report.” Ads removed due to policy violations are also not displayed, preventing users from determining which specific ads were in violation. These limitations decrease the usability of the ad transparency database.

• 

• 

The majority of Google’s data is contained in its CSV files. These files require, at minimum, a basic knowledge of Microsoft Excel. Google provides six main categories of data about ads in its CSV files: advertiser stats, advertisers’ weekly spend, campaign targeting, creative stats, geography, and top keywords history. Within each one there are detailed subheadings; therefore, filtering by column is effective in terms of having the data in separate categories, but it is very time-consuming to locate specific data for non-technical users. In the CSV files for just a single candidate there may be 1,000 or more rows in one spreadsheet, and each piece of information relates back to only an ad ID. As an example of the limitations of this, if you wanted to find targeting information about a specific ad, you have to first find the ad in the online database, copy its ad ID,  find the relevant spreadsheet, and then use the search and find feature to locate the ad ID within that spreadsheet. The difficulty in searching for more detailed data other than an advertiser or ad keyword restricts the Google database’s usability for the non-technical user.

Since Google’s Transparency Report contains much more than political advertising, when users click on the most visible “FAQ” link at the top of the page, they must then extend a drop down menu to find “Content removal and additional reports,” and then select “Political Advertising on Google FAQs.” This same FAQ page is available via direct link at the bottom of the page as well. Some questions answered here include: “Which advertisers are included in the Political Advertising Transparency Report?” and “Where did ads in the report appear?,” “Why can’t I see the content of some ads in the report?,” “How often is this report updated?,” and “What does ‘impressions’ mean?” The answers can be vague. For example, the question of “How is total spending determined for a country or region?” is answered with the brief: “Total spending is based on where ads were shown.”

Twitter

Twitter’s Ad Transparency Center was visually comprehensive and simple, but lacking in aggregated data. The homepage had three headers with simple descriptions and links that the user could click on. The first section was titled “A more transparent Twitter,” and had a description of what users were able to do within the Transparency Center. The second section was for political campaign advertisers, and the third section was for issue advertisers. Twitter provided an explanation of what these two groups of advertisers’ main focus is and provided the links mentioned earlier for users to “see a list of certified political campaigning advertisers” or “see a list of certified issue advertisers.”

There was also a search bar at the top right of the page, which allowed users to search for any Twitter user. This meant that users needed the exact name of the account they wanted to look at in order to search directly for data on political advertising. Otherwise, users could click through the section headings to the “list of certified political campaigning advertisers,” view advertisers from different regions of the world, and receive an alphabetized list of certified political campaign advertisers. While this helped users who did not know the exact name of the account they wished to access, it was a long list to sort through.

• 

Once users located the advertiser they were looking for on their page there was general information about them as well as specific data for each ad. General data was simple to view, but the link to find more detailed data was not obviously placed or identified. Twitter did not use specific filters besides general date ranges to allow a user to sort through ads, but did provide a detailed amount of data on each ad. Users could find under an advertiser’s Twitter header their name, picture, bio, when they joined the platform, how many people they were following, and how many followers they had. Each ad contained an “Ad performance summary,” which included the total spend and impressions for the ad in all of its campaigns. Following that was a section on targeting, where users could click on a specific campaign date range to view the targeted audience and the actual audience. 

A focus on top spenders and national politics

Below are screenshots from Facebook’s and Google’s ad libraries and reports, including the spending breakdowns by location, campaign page, and advertisers. These tools organize spending from greatest to least, with the location tracker showing states where more money was spent in darker shades. In doing so, both Facebook and Google direct attention to the highest-spending political advertisers, in the process (and perhaps unintentionally) focusing on national political spenders and making local political advertising information harder to find. (Twitter did not have an overview section that makes spending comparisons across different advertisers.)

A focus on high-level snapshots or fine-grained investigations, but no mid-level strategy

Facebook and Google encourage users to investigate political advertising at the highest-level (all spend by an advertiser) by giving exact numbers or at the lowest, most fine-grained level (an individual ad).

For example, on each of the platforms, spending is broken down by advertiser and individual advertisement but not by an advertisers ad group or targeting strategy. Both Facebook and Google show the exact total ad spend for each candidate, but when broken down by individual advertisements and campaigns, these platforms show much wider ranges. These ranges make it difficult for a user to determine the actual spend per advertisement or messaging strategy. For instance, there is no data available for how much candidates spend on individual campaign efforts, such as presidential candidate Elizabeth Warren’s “Grab A Beer” campaign. Users’ ability to understand the advertisements at any other level is incredibly limited due to the use of ranges. A user can’t accurately add up reach, spend, or geographic overlap to understand the impact of “Grab A Beer” or any other messaging strategy.

Facebook:

Google:

Unlike Facebook and Google, Twitter does offer individual ad and campaign spend information so users would know specific spend amounts. Twitter shows either exact numbers or closer ranges per individual advertisement and campaign. While Facebook and Google limit advertisers to either the highest-level of analysis or the lowest, Twitter biases towards the middle and lower.

• 

• 

---

---

Personnel:

Project Leads:

Bridget Barrett, Research Lead, CITAP Digital Political Ads Project

Daniel Kreiss, Principal Researcher, CITAP

Researchers:

Meredith Ammons

Maggie Brenan

Molly Brice

Ava Eucker

Amelia Fox

Jessica Hardison

Emily Kramer

Andrew Mason

Ariel Nissan

Carlyle Rickenmann

Noah Robertson

Vanessa Schoning

Alley Steele

Talia Wlcek

Chel Wock