Online behavioral advertising

Online behavioral advertising (OBA) is a configuration of online advertising that allows advertisers to target consumers with advertisements personalized based on their behavioral data

Keywords: online advertising, digital platforms, behavioral personalization, targeted advertising, artificial intelligence, digital market

Read about customer RFM segmentation

OBA paradigm, a pattern or model

OBA is the online phenomenon that entails someone, ex company to show to consumers advertisements that are personalized based on their behavioral data. By OBA definition
the three premises form the OBA paradigm:

(i) targeting individual consumers with ads is beneficial for advertisers and it is also
possibly beneficial for consumers,
(ii) consumer’s observed behavior reveals (discover) what consumer reacts to better than surveying, = facts vs declarations
(iii) the Internet (or store circumstances) can be used to observe and influence consumer behavior.

In the search to define consumer audiences in more granular ways, the marketing industry not only collected data through voluntary self-disclosure, e.g., surveys but increasingly adopted
the logic of behaviorism.

Behaviorism

Behaviorism is a branch of psychology that understands a human experience as measurable,
observable behavior that can be studied, predicted, and influenced without the subject’s awareness (John B. Watson 1924). Nowadays, widely utilized in online behavioral advertising.

Targeting – online behavioral advertising

Supermarkets pioneered using behavioral information for targeting campaigns. A recent example of a supermarket relying on consumer behavioral data to target them with a
marketing communication is when Target Inc., a United States(US) store. They made headlines in 2012 for its data-driven targeting practices. By analyzing the shopping behavior of their consumers who disclosed that they were pregnant, Target constructed a “pregnancy prediction” score. When new consumers exhibited similar purchasing behavior, Target automatically predicted that they were pregnant and targeted them with appropriate marketing communications e.g., sending booklets about diapers to the home address of their consumers.

In early 2000s, Google Search emerged as the superior online search engine that relied on the
PageRank algorithm, which accomplished unprecedented relevance and efficiency in delivering search results. Google Search’s technological superiority stemmed from its behaviorist logic – it observed cues of consumers’ online behavior, such as the pattern of searched terms, spelling, punctuation, dwell times = time of activity, and locations that were ignored by other search engines. It used these cues = factors, often called “data exhaust” or “digital breadcrumbs,”
to turn the search engine into a recursive algorithmic system that continuously learned and improved the search results.

The ban on commercial use of online activities

The ban on commercial use of online activities was lifted in 1994. At that time, internet users were primarily members of a homogenous group of middle-to-upper-income college-educated men, and advertisers were slow to show interest.

By the 2000s, as a more significant part of human society moved online, search engines became a new venue for marketers to reach audiences that now disclosed their interests by typing keywords.

Bidding = the offering of particular prices for something, especially at an auction

Some search or catalogue websites ex. GoTo.com, allowed marketers to bid for their websites to be prioritized in the search results: the highest bidder was listed first, the runnerup was listed second, and so forth

In contrast, Google Search faced bankruptcy, as its founders, committed to retaining its
technological superiority and high standards of search relevance, refused to rely on advertising.

Google Search adds

In response to the continuous pressure from investors to find a profitable business model, Google Search adopted several forms of online targeted advertising that were claimed to provide
the users with an advertisement that they found relevant, which could be demonstrated by increased conversion rates.

Conversion rate is the number of times consumers clicked the ads.

One configuration of Google advertising was OBA that, similar to when improving search results, relied on observing consumer behavior and targeting advertisements based on “digital breadcrumbs” Google picked up about the consumers. It demonstrated the highest conversion rates compared to other configurations, becoming most popular amongst advertisers and
thus becoming Google’s primary revenue stream.

Read about call to action here

OBA configurations

Online Targeted Advertising

Online targeted advertising (OTA) refers to an online advertising practice that delivers an advertisement tailored to a particular context or an individual consumer. Therefore, two
major types of OTA are online contextual advertising (OCA) and online personalized advertising (OPA).

Online contextual advertising (OCA)

In OCA, advertisers target consumers based on the interaction context. Online behavioral advertising.

This may include the digital content on the publisher’s web page or app that the consumer is accessing, the language content is presented in, the time of the day content is accessed, the general geographic location (e.g., country, state) of the content is accessed from, as well as the weather on that location.

This contextual information allows advertisers to present ads in the correct language, in the correct market, with the awareness of the elements of the day, and achieve relevance
by analyzing the content consumers access instead of analyzing information about the consumers themselves.

Online personalized advertising (OPA).

In contrast, OPA targets individual consumers based on consumer identity or using the data about consumers themselves.

OPA can be based on data that consumers provide voluntarily. Online segmented advertising (OSA) is a stipulatory term used in policy documents to describe OPA that relies on broad demographic information that the consumers voluntarily disclose by, for example, signing up for digital services or content.

Such information usually includes gender, age, country of residence, and in some instances, the parental status of the consumer. OPA can rely on more detailed demographic information,
such as the consumer’s education (e.g., high-school graduate), finances (e.g., household income top 10%), relationship status (e.g., married), employment (e.g., tech industry), or other dociodemographic categories. Online behavioral advertising.

Profilling

Advertisers can build such a consumer profile based on the data voluntarily disclosed by the
consumer (i.e., “explicit profile”) or based on the data about consumer online behavior that they observed (“predictive profile”). Developing predictive profiles by algorithmically inferring
attributes based on the observed online behavioral data about the consumer is commonly called “profiling”.

Online behavioral advertising is an advertising practice that relies on profiling to target individual consumers. Observed online behavioral data about the consumer may include social media data (e.g., posts and likes),
search data (e.g., history), web browsing data (e.g., media consumption data), mouse cursors movement, keyboard strokes, and location data.

In OBA, consumers can be profiled beyond demographic traits and may include inferring psychographic traits such as affinities, interests, values, and lifestyles. For example, a consumer can be inferred to be a “surf enthusiast”, a “sci-fi fan”, a “dog lover”, someone who “is about to have a wedding anniversary,” or who “recently moved to Hawaii”.

Recommendations

Profiling can also be used for personalizing any digital content more broadly. For example, using behavioral data for personalizing search results by changing their order is often called “personalized ranking” – a practice that almost all websites engage in that allows search (e.g., search engines and online marketplaces). Online behavioral advertising.

Algorithms for personalizing digital content are often called “recommender systems.” Behavioral
personalization of content through such systems is often framed as the core practice of digital service providers. For example, Netflix claims to provide “personalized digital content service”
– referring to its movie recommendation system, and Facebook defines its primary service as the provision of “personalized experience” – referring to its News Feed. While behavioral personalization of content is not the same as OBA, the latter often involves the former. Sometimes, they are bundled together to justify data collection for advertising personalization

Paid ranking

In addition, some websites that use recommender systems for personalizing search results allow advertisers to pay prominence to their products (i.e., “paid ranking”). The paid
ranking is part of OBA to the extent to which behavioral personalization considers consumers’ predictive profiles.

Re-targeting

Another form of OBA is “re-targeting,” which relies exclusively on consumers’ observed shopping behavior and shows consumers ads for the products and services interest they revealed by, for example, adding them to the shopping cart of the online marketplace.

Re-targeting is particularly noticeable for consumers, as they experience being followed by advertisements across the Internet. Re-targeting is sometimes dubbed as “creepy marketing” because of the following nature of the advertisement.

OBA Market: ….

Online behavioral advertising infrastructure

Real-Time Bidding (RTB)

In online behavioral advertising, advertising placements are determined programmatically, that is, by algorithmic systems instead of human-mediated ways. In this programmatic process, advertisers bid on the Real-Time Bidding (RTB) auction to compete with other advertisers to target an adds to a specific consumer online. Online behavioral advertising.

In the OBA open exchange, the RTB auction is housed by the ad exchanges, where SSPs sell the advertising inventory of their publishers and DSPs place bids for their advertisers.

The consumer visiting publisher’s website initiates the programmatic process. Using the trackers placed on the website, the publisher’s SSP (or an ad server in case of multiple SSPs) generates an advertisement request (“bid request”) that contains a broad array of information about the consumer seeing the ad inventory. Further, bid requests are passed to ad exchanges and to the DSPs that evaluate advertising opportunities based on their campaign objectives and respond with their bids, the amount of money the advertiser is willing to pay per click.
The publishers (via SSP or an ad serve) rank the offers based on the price (and other priorities) and decide which advertisement will be served on the webpage

Cookies

The most prevalent way to track consumers has long been via trackers known as “cookies”.
Cookies are small blocks of encoded or encrypted data that the website’s server places on the
consumer’s computer (that visits the website) and later accesses and reads to identify the returning user.

In the early days of the internet, publishers could not tell the difference between
visitors. Cookies were introduced in 1994 by Netscape Navigator, primarily to “give Web a memory” or, in other words, to identify the re-visiting users on the website.

Today, cookies are used for various purposes:

They can be strictly necessary for enabling website features, for example, accessing secure areas of the website or adding items to a shopping cart.
They can also be used to improve performance, such as tracking errors or which website pages are most visited.
They can also enable other functionalities, for example, to keep users logged in or retain their preferences.

Such cookies are also called first-party cookies as they are placed by the server of the publisher’s website that the consumer visits (i.e., first-party). There are also third-party cookies placed by a party other than the publisher, such as an advertising network.

While third-party cookies can provide significant functionalities (e.g., showing a video from another source), they also allow tracking of the users across the internet and, therefore,
have been used to operationalize OBA.

Disable third-party cookies for advertising

In 2019, Mozilla’s Firefox adopted a default configuration to disable third-party cookies for advertising unless activated by the user, and in 2020, a similar feature was adopted by Apple’s Safari.

Cookieless OBA

Device fingerprinting

Device fingerprinting is one such method by which seemingly insignificant information about the features of the device, such as screen resolution and the list of installed fonts, are analyzed to give the device a unique “fingerprint”. This fingerprint can be used, for example, to combat fraud (e.g., identifying a person trying to log in to a site is likely an attacker who stole the credentials), but also to track a single consumer across different websites without their
knowledge and without a way of opting out. Device fingerprinting allows tracking users without cookies, but also it can be used to respawn deleted identifiers in case the consumer
deletes cookies. The research found fingerprinting evidence on at least 4.4%–5.5% of top websites.

Federated Learning of Cohorts

Alphabet has been developing allegedly a privacy-preserving alternative to build behavioral
profiles called Federated Learning of Cohorts (FLoC). Instead of assigning unique identifiers to the users, like in the case of cookies, using FLoC, a web browser will analyze users’ browsing behavior and assign them to “cohorts” – clusters of users with similar browsing behavior and presumedly similar habits and interests.

FLoC aims at replacing functionality served by cross-site tracking but maintains detailed lifestyle targeting of OBA. In essence, such Privacy Enhancing Technologies (PETs) increase the confidentiality of data, but they do not limit the data OBA consumers nor change the way
data is used within the practice.

Based on: CONSUMER MANIPULATION VIA ONLINE BEHAVIORAL ADVERTISING, Lex Zard, 30 B. U. J. SCI. & TECH. L. 2, (forthcoming, Summer 2024)