collection and liberation of advertisement data
The lab runs an infrastructure for the collection and liberation of political advertising data from five companies and platforms, consisting of a dozen advertising products and services. The archive created with the operation of this infrastructure is meant for accessing
insights within the datasets, as well as for devising methods to study the nature and form of the political advertising transparency paradigm itself. This infrastructure is a gnarly, growing thing that is the result of a year-long iterative process to examine influence on
social media. The lab deployed a spatial enunciation of the infrastructure first at esc medien kunst labor in March 2020, and the infrastructure has since moved to the servers of network culture initiative mur.at.
WHAT'S IN THE ARCHIVE
DATA COLLECTION SCHEDULE
| SCRIPT | QUERY PARTICULARS | FREQUENCY | DATASETS |
|---|---|---|---|
| EARTHWORM An investigative script that digs into data on subjective topics, currently the COVID-19 pandemic and the climate crisis. |
(All available fields, of ads containing (All available fields, of ads containing |
1. Historical take 2. Every 24 hours |
COVID-19 |
| NEMESIS A slow-crawling script that pieces together all ads from 2013 until the present day, of all the platforms of FACEBOOK Inc. Not without multiple hidden obstacles. |
(All available fields, for all active and inactive ads in the category of Political and Issue Ads, in all publisher platforms that have reached all 101 countries where we can access this data.) |
1. Historical take 2. Every 24 hours |
FACEBOOK MESSENGER AUDIENCE NETWORK |
| SENTINEL An expected visitor, this script knocks on the doors of identified political candidates and political parties in 39 countries in the world. It observes all the platforms of FACEBOOK Inc. |
(All available fields, of all active and inactive ads in the category of Political and Issue Ads that are in all publisher platforms that have reached the respective country from the page ID of the specific political actors.) |
1. Historical take | ELECTIONS |
| MINUS Through some slick manuevers, this conniving script scoops out all political ads data of countries with a second degree of transparency across all of FACEBOOK Inc. platforms. |
(All available fields, of all active and inactive ads in the category of Political and Issue Ads that are in all publisher platforms that have reached 62 countries where political advertising transparency is not fully rolled out) |
1. Historical take | NEGATIVE SPACE |
| WOODPECKER A script insistently revisiting weblinks for the data dump of Snapchat, Twitter, Google LLC (Google, Youtube, AdWords, Ad Network) and Reddit. |
(all (all (all data on promoted (all data on promoted |
1. Every 24 hours |
GOOGLE LLC |
notes from an infrastructural method
A systematic collection across platforms, or an infrastructural approach, allows for making sense of the data, but also for testing the integrity and evolution of the datasets themselves across time, regions and platforms. For example, documented instances of disappearing ads in the Ad Library also bring to light the need for a parallel archive, creating possibilities of testing these datasets for integrity.
However, comparison of political advertising across platforms can be tricky because of the differences in their advertising business models, choice of data structures etc. For example, data about regions is available both within Facebook data as well as Snapchat data. The former refers to the regions that users were in at the time of receiving the ad, even when the ad was delivered for reasons unrelated to their location. Snapchat on the other hand, seems to be providing information about the regions that were specified in the advertiser's ad buys.
Despite these differences, there also seems to be a consensus and ceiling emerging between platforms about the expectations from so-called transparency measures. For example, even platforms with seemingly oppositional approaches like Twitter (which has banned political ads) and Facebook (which provides special privileges for political ads) seem to proceed on the assumption that the category of political advertising is relevant and useful. This distinction between political ads and the categorisation-defying non-political ads rests on questionable determinations about what the contents of the political might be. More importantly, nominal transparency for a small category of political ads provides a cover for non-political or residuary ads to escape any scrutiny, insulating majority profit flows from any legibility or accountability.
| SNAPCHAT | ||||
|---|---|---|---|---|
| ad_snapshot_url ad_creation_time ad_creative_body ad_creative_link_title ad_creative_link_caption ad_delivery_start_time ad_delivery_stop_time currency demographic_distribution funding_entity impressions page_id page_name region_distribution spend potential_reach id publisher_platform |
Advertiser_ID Spend_Range_Min Spend_Range_Max Impressions Num_of_Days Ad_URL Date_Range_End Ad_Type Campaign_ID Age_Targeting Gender_Targeting Geo_Targeting_Included Geo_Targeting_Excluded Start_Date End_Date Ads_List Advertiser_ID Advertiser_Name Country Country_Subdivision_Primary Country_Subdivision_Secondary |
AD ID Creative Url Currency Code Spend Impressions Start Date End Date Organization Name Billing Address Candidate Ballot Information Paying Advertiser Name Gender Age Bracket Country Code Regions (Included) Regions (Excluded) Electoral Districts (Included) Electoral Districts (Excluded) Radius Targeting (Included) Radius Targeting (Excluded) Metros (Included) Metros (Excluded) Postal Codes (Included) Postal Codes (Excluded) Location Categories (Included) Location Categories (Excluded) Interests Os Type Segments Language Advanced Demographics Targeting Connection Type Targeting Carrier (ISP) Creative Properties |
account_name user_name bio_url city credit_card_full_name postal_code region advertising_agency_name company_name spend billing_state billing_city sold_to billing_postal_code target_type ad_campaign_spend ad_campaign_start_date ad_campaign_end_date ad_impressions target target_type number_tweets tweet_text tweet_url |
link_to_creative start_date end_date impressions spend subreddit_targeting negative_subreddit_targetting interest_targeting true_region_targeting true_country_targeting excluded_region_targeting |
FREQUENTLY ASKED QUESTIONS
1. What are political ads? Who classifies them as such?
Political ads are advertisements that are classified as "political" by different advertising platforms. The exact terminology of classification differs across platforms. For example, FACEBOOK Inc. refers to such ads as related to "social issues, elections or politics", Snapchat refers to such ads as "Political and Advocacy Ads" and so on.
Generally, advertisements created by politicians and political parties for election campaigning are considered political ads. However, the category had to be broadened to include burning political issues even when the ads about them were created by actors other than candidates and political parties.
Going by the logic that only certain ads are political, the question of which ads should be considered so can get very complicated very quickly, as this involves differences across countries and regions, the need for issues having attained critical mass etc. Determinations about which ads are political are therefore extremely subjective. Facebook notes how it makes these decisions through this simplistic and ever-changing chart.
In recent years, events around election interference has created public opinion that has cornered platforms into revealing some information about ads. Platforms have formalised this category and only reveal information about such ads.
2. Are the data collected only of political ads?
Yes.
3. How do you actually collect the ads?
We collect ads from FACEBOOK Inc. through the Ad Library API. In the time that the API has existed, it has been highly unusable for large amounts of data collection. We detail the process we went through to collect the data in this piece for Exposing The Invisible kit.
We collect ads from Snapchat, Reddit and GOOGLE LLC through their weblinks where they provide this data.
Twitter stopped delivering political ads over its platform in late 2019. However, we regularly collect the archived data to check for consistency of its maintenance and other questions that might emerge even with a static dataset.