More on this book
Kindle Notes & Highlights
Read between
January 9 - January 15, 2023
Second-party data is obtained through partnerships with other entities, and is basically their first-party data. To illustrate, a price comparison site might share their first-party customer data with an e-commerce site. To the e-commerce site, this is second-party data – and is indeed very valuable. Often, partners would mutually share their first-party data in order to gain a better user understanding, or to extend their cross-device ID recognition. Second-party data sharing can also take the form of data co-ops, where multiple partners (typically publishers) pool their data together.
Finally, third-party data is data obtained from external providers, with no direct partnership with the buyer. For example, a publisher data sharing co-op can segment visitors based on their browsing behavior, and sell these segments to advertisers for campaign targeting. To the advertiser, these segments (say car purchase intenders or photography buffs) constitute third-party data. Such data can be very useful in efficiently extending advertising campaigns to wider audiences of potential customers.
Finally, third-party data use frequently shows high ROI, especially when a campaign is based around the right segments of sufficient quality.
Many third-party data providers are available, from publishers and publisher co-ops to specialist vendors such as Experian, Acxiom, Lotame, Oracle (BlueKai) and others.
Depending on how data was collected, it can be classified as declared, inferred or modelled. Declared data is given by the users themselves, and might include things such as age, gender, interests and so on. Facebook is a treasure trove of declared data, from user profiles to things they like or share. Declared data is commonly collected from registration forms, or by running sweepstakes. Inferred data is not given by the users directly, but is deduced – usually from their behavior. Someone reading a lot of computer game reviews is likely to be interested in playing computer games. Finally,
...more
Conversion pixel from the vendor is placed on a post-conversion page, to track users who have already converted. This can be a "thank you for your purchase” page for instance. This pixel is there
With simple e-mail retargeting, ads are shown to users who have opened an advertiser’s e-mail.
Increasingly popular is retargeting using e-mail lists (sometimes referred to as CRM retargeting). An advertiser simply uploads an e-mail list from their database to a vendor, who matches the e-mail addresses to user cookies (with varying match rates depending on the vendor). These users can then be retargeted as if they had visited the advertiser site. E-mail list retargeting is common on Facebook and Twitter, where other user credentials beyond e-mail address (such as a phone number or social network ID) can be used
Google has recently introduced a very powerful and exciting e-mail list retargeting product as well, called AdWords Customer Match. Retargeting pools can be used directly within Google AdWords, bringing potential efficiency to a whole new level.
While desktop retargeting is based largely on cookies, these are not widely available on mobile devices. Other means of identification in addition to cookies are used to accomplish the task, such as Apple’s Identifier for Advertisers (IDFA) on iOS, Google’s Identifier for Advertisers (AAID) on Android or the Facebook ID in the Facebook ecosystem (including the site itself, as well as apps using Facebook SDK).
There is a large number of specialized retargeting vendors to choose from, including AdRoll, Retargeter, Criteo or Sociomantic.
Partly due to the prevalence of Google Analytics as advertisers' primary analytics platform for evaluating campaigns, last-click (or last-impression) attribution is often used to compare effectiveness of different campaigns, channels or tactics. Unfortunately,
With last-click or last-impression attribution model, prospecting will likely get no credit for this conversion, while retargeting gets all of it. The
media planner would select a portfolio of sites where the advertiser’s audience could be reached,
Behavioral targeting, also known as audience targeting, relies on individual user web-browsing behavioral data to increase advertising relevance.
Behavioral/audience segments are then defined based on recency and frequency of relevant signals.
Data management platforms (DMPs) are typically used to facilitate behavioral targeting. Segments are usually defined manually, and the resulting data quality therefore depends not only on the quality of signals, but also on choices made in segment definition. These choices include factors like signal recency, frequency, or strength. For some segments, data quality can be increased through shorter recency windows – this is typically the case with segments capturing purchase intent. Someone searching for “Ford dealership” yesterday is more likely to be in-market than if the search happened a
...more
Data providers typically aim for a good balance between data quality and scale, which involves a lot of trial and error to get right.
Advertisers or typically purchase third-party audience data for this purpose, either from specialized data vendors, or directly from publishers.
An advertiser usually chooses the audience they would like to extend – base user group. This could be for instance users who have converted on the advertiser’s site, site visitors, users belonging to a certain audience segment, or even an e-mail list from a CRM database.
lot of data used for targeting in programmatic advertising is collected and stored separately from an advertiser’s Demand-side platform. This is particularly the case of third-party behavioral data, sold by external data vendors.
When the DSP makes a bidding decision, it needs to know that the bid request is for a user (cookie) belonging to a particular audience segment. There are two common ways to achieve this – using cookie synching and via a Deal ID.
The key seems to be matching users to cookies
For simplicity, let’s assume for now that the DSP knows which cookies set by the DSP belong to a particular audience segment (through an integration with a data provider for example).
assumed in the beginning that the DSP already knows which cookies belong to a particular audience segment. This can be because the DSP itself collected the data (typical for retargeting), or because it has an integration with external data vendor. In the latter case, the data vendor and the DSP first need to synchronize their cookies – just like in the above example. Then the data related to individual cookies can be transferred to the DSP either in real-time using pixels, or in regular intervals using batch processing (also referred to as server-to-server integration).
Location data is usually obtained either from ad impressions, or from apps with access to location services on a user's device.
Caution when using mobile location data is advised, and vendors are already offering solutions to address location fraud.
Retailers can also use location and in-store transaction data in attribution modeling to understand path to purchase and capture offline conversions.
Additionally, location data can be used for profile linking (for example a TV or desktop computer profile in the place of residence with a mobile device). Naturally, using location data for behavioral profiling can easily violate user privacy. Such practices need to be highly regulated.
initiatives aim to overcome this problem, such as publisher co-ops created for the purpose of sharing deterministic IDs. The question with such projects is typically scale, accompanied by privacy and data protection concerns.
Another use case for cross-device integration is data consolidation. Previously isolated user profiles related to the same user can be merged, creating a richer, more accurate picture of reality. This increases data quality, applicability and value.
Only non-PII data should be used for behavioral targeting.
Additionally, advertising channels are gradually becoming automated, from TV to audio, out-of-home and print.
aspects are critical when we talk about programmatic TV – automation and audience data. The
Unlike some digital channels where programmatic is already thriving, television, especially linear, doesn’t yet offer much individual (or household) audience data. Limited offline and digital data is used during TV campaign planning and execution, but this is often a manual process. Such data might include credit/credit card data, car registration data, or set-top box data (commonly referred to as addressable TV). For
The possibility to target TV advertising to a very granular audience segment will increase campaign efficiency of many advertisers,
Our segment chammes.could be very helful.here
number of audio ad exchanges already exist, including Triton with the a2x audio ad exchange[xxxiv] in the United States and elsewhere, or the Digital Advertising
