A brief analysis of precise audience targeting in Internet advertising!

With the development of Internet advertising technology and business, audience targeting has become the ultimate secret pursued by almost every advertising platform. How to conduct accurate audience targeting has naturally become a powerful means for Internet advertising platforms to establish core competitiveness. First, let’s talk about what precision targeting is.

In Internet advertising, precision targeting refers to the precise matching and delivery of advertisements by mining users' historical behavior data. For example, I just read a book on JD.com , and when I went to NetEase to read the news, I was surprised to find that NetEase happened to be showing an advertisement for this book. Is it really that magical? Let's take a look at the unknown process that happened in the middle:

I browsed products on JD.com, and JD.com recorded my user information: cookies, IP address, MAC address, and all other available information. Then when I opened NetEase News , NetEase learned that I was a JD.com user through cookie mapping, and immediately sent a request to JD.com:

"A user comes to my site and happens to be your user. Do you want to display an ad?"

JD immediately said:

"Okay, I want this user."

Then it quickly uses an algorithm to analyze my browsing history, shopping history, etc., matches an advertisement and sends it to NetEase, and NetEase immediately shows me the advertisement. This process takes less than one second. (We will not consider all third-party advertising platforms involved during this period)

This is a very important and widely used example of precise targeting. In fact, there are many other precise targeting methods. Next, follow the editor to look at the various precise targeting methods in computational advertising and its advantages and disadvantages.

Let’s first look at the more popular targeting methods on the market. According to their effectiveness and the stage in which they work in the advertising information reception process, we roughly qualitatively evaluate these targeting methods as shown in the following figure:

The horizontal direction represents the stage at which targeting technology roughly takes effect in the process of receiving advertising information, while the vertical direction represents the qualitative effect evaluation. Next, we will explain various qualitative techniques in turn:

1) Geo-targeting

This is a orientation method that has been widely used for a long time. Since many advertisers’ businesses have regional characteristics, geographic targeting plays a very important role and is also a targeting method that all online advertising systems must support. Regional targeting can also be considered a kind of contextual targeting, but its calculation is simple and can be completed with just a simple table lookup.

2) Demographic targeting

Although demographic targeting may not be particularly effective, it is particularly familiar to brand advertisers because such labels are widely used in traditional advertising discourse to express the audience. The main labels of demographic attributes include age, gender, education level, income level, etc. It should be noted that unless there are special specialized data sources , such as real-name SNS registration information or online shopping consumption records, it is generally not easy to carry out accurate demographic attribute targeting.

In the case of insufficient coverage of demographic attribute data, if we want to conduct CPM sales in this targeted manner, we can use users with known demographic attributes as the training set and construct a classifier to automatically label the demographic attributes. Generally speaking, the accuracy of using classifiers to determine population attributes is limited. In purely performance-based advertising campaigns, there is little need to predict demographic attributes, because the predicted demographic attributes are also obtained based on other behavioral characteristics of users and cannot provide additional information.

3) Channel targeting

Channel targeting is to divide the inventory by channel according to the supplier's content classification system, and deliver different advertisements to the traffic of each channel. This targeting method is more suitable for vertical media that are closer to conversion needs, such as automobiles , maternal and child care , shopping navigation, etc. For media with a wide range of content, the effect of this approach is limited. To give an extreme example, if we use a website’s military channel as a targeting tag, it would be difficult to find directly matching advertising needs.

4) Contextual targeting

Matching relevant ads based on the specific content of a web page is called contextual targeting. The granularity of contextual targeting can be keywords , topics, or categories determined based on advertiser needs. The effectiveness of contextual targeting varies greatly across different categories of content, but one big advantage of this approach is its high coverage. For most ad displays, no matter how much information is known about the current visiting user, it is often possible to infer the user's immediate interests based on the currently browsed page and push relevant ads.

5) Behavioral targeting

Behavioral targeting is a very important targeting method in display advertising. Its framework is to understand user interests based on their historical access behavior and deliver relevant advertisements. Behavioral targeting is important because it provides a general idea that allows user behavior data collected on the Internet to generate monetizable value. Therefore, the framework, algorithms and evaluation indicators of behavioral targeting have laid the foundation for the essential characteristics of data-driven online advertising, and have spawned derivative businesses of related data processing and trading.

If contextual targeting is considered as targeting based on a user's single visit behavior, then behavioral targeting can be considered as the fusion result of a series of contextual targeting. Therefore, contextual targeting is the basis of behavioral targeting, and there can be corresponding behavioral targeting methods for various types of contextual targeting. For example, geographic targeting determines the geographic area based on the IP address that the user is currently visiting. Correspondingly, it can also be targeted based on the geographic location that the user has most frequently visited in the past period of time. This method actually gets closer to the user's usual place of residence. Some people in the industry call it "where on earth" targeting.

6) Hyper-local targeting

When advertising on mobile devices, we have the potential to obtain very precise geographic location. For example, using cellular information or GPS, the accuracy of geographic positioning can reach the granularity of a block. If the indoor positioning technology of Wi-Fi, Bluetooth and other devices is further used, the accuracy can be further reached to several meters. This makes it possible to advertise based on precise geographic location, and also allows a large number of small advertisers with very strong regional characteristics (such as catering, beauty, etc.) to have the opportunity to place precisely targeted ads. This is a qualitative change from the traditional geographical targeting and has become one of the most important opportunities for mobile advertising .

7) Retargeting

This is the simplest customized tag. Its principle is to deliver advertisements to visitors of an advertiser over a period of time in order to improve the effect. Obviously, a visitor to a particular advertiser is unique to that advertiser, so this is a customized tag. Redirection is recognized as the most accurate and effective targeting method, but its coverage of the population is often small. This is because the reach of redirection is determined by the number of the advertiser’s inherent users and the overlap ratio with the media.

8) Look-alike targeting

Since the amount of redirection is too small and cannot meet the advertiser's need to reach potential users, it cannot be relied upon solely for advertising. The idea behind targeted new customer recommendations is to find potential customers with similar behaviors for advertisers based on the seed visitor information provided by advertisers and combined with richer data from the advertising platform. The purpose of this method is to achieve better results than some general interest tags with the same user coverage ratio, which essentially reflects the core value of advertiser data.

I believe that after reading the above introduction, you should have a preliminary understanding of precise targeting. In Internet advertising , there are many ways to accurately target the audience, each with its own advantages and disadvantages. However, the final targeting method to be adopted still depends on the technology and data resources owned by the platform. In-depth mining based on the data itself and finding meaningful signals for advertising are the key points of precise audience targeting. I hope sharing this here will inspire you.