Advertising strategies for the Internet automobile industry

In many advertising courses, the targeting techniques summarized over many years of advertising are even considered advertising strategies. This also leads to the fact that when people search for content related to advertising strategies on the Internet, most of them are teaching us how to make various settings in the advertising background. But in fact, advertising strategies are far more than just targeting, and for a long time many people have had a large deviation in their understanding of targeting strategies.

Ad targeting is not a strategy in essence but a product feature used to screen advertising audiences. This function can filter the target population for different delivery purposes through label combinations, so the essence of the targeting strategy is to design different label combinations, but it does not involve the core decision-making of advertising delivery and the reach of the population. Therefore, a commercial product manager who unilaterally regards targeting strategy as advertising strategy is obviously unqualified.

From the perspective of advertising optimizers, because they are not involved in the underlying logic of the advertising system, their delivery strategies are often implemented based on account, targeting and budget settings. However, from the perspective of a commercial product manager, it is necessary not only to clearly understand the position of targeting in the entire system and the implementation principles, but also to consider what delivery strategies optimizers can implement based on this on the basis of functional implementation, and to support various intelligent delivery strategies (machine automatic delivery).

This part requires a high level of personal experience from the business product manager and an understanding of the entire advertising system.

1. The principles of advertising targeting and optimization direction

Many articles make the principles of ad targeting sound mysterious, such as recommending ads of interest to users in real time based on their current browsing behavior . As soon as this kind of rhetoric comes out, you know it is trying to "fool" laymen. The reason is very simple, because currently the targeting function of most advertising platforms is achieved through a label system. Although the user's browsing behavior can be obtained in real time, those who have actually done user behavior analysis will know how chaotic and disordered user behavior is.

Since we cannot exhaustively know all user behaviors, we naturally cannot make different advertising decisions for each user behavior. Therefore, it is a more economical and reasonable solution to classify user behaviors into limited labels and then make advertising decisions based on different label combinations.

Next, let’s talk about tags. There are many types of tags, the most common ones are: region, user attributes, context, behavior, preferences, etc. Without considering the accuracy of the tags, these are standard features of all major platforms. But multiple types of labels will also bring another problem. When there are too many labels to place, you don’t know how to choose?

At this time, DMP (data management platform) is needed to provide user portraits of their target users for advertisements. For example, if the target group is men born after 1995, then the targeting should be set based on this conclusion. In theory, the more accurate the label, the better the advertising effect will be.

The examples we gave above are the simplest and most intuitive cases, but in real life, user behaviors and decision-making cycles are different in various industries and types of products. This requires us to deal with it in different categories.

Back to the automobile industry we are familiar with, the user's car purchase decision cycle is much longer than that of other e-commerce products, so the user behavior during this period will also become extremely complex. How to unravel these behaviors to find the patterns and form a label system for automobile industry users, and then form different targeting strategies in advertising has become a problem that we, the commercial product managers in the Internet automobile industry, have to think about.

2. UVN-BI label system

Here we share a user labeling system for car purchases that we have developed after years of practice - the UVB-BI labeling system.

The design concept of the entire system draws on the famous RFM model. The RFM model uses the three dimensions of the most recent consumption (Recency), consumption frequency (Frequency), and consumption amount (Monetary) to establish a coordinate system that can divide customers into up to 125 categories.

Inspired by RFM, we apply this method to the segmentation of car purchasing users. UVN-BI user clustering is to characterize and cluster multiple basic attributes of users and map these attributes with the user's car purchase stage and the user's interest in cars to form a user clustering model based on car series. Before presenting the model, we first define the dimensions involved in detail:

• U (User) User basic attributes: mainly using two characteristics: city level and generation. City level (first to sixth tier cities) Generation (post-00s, post-90s, post-80s, post-70s, pre-70s)
• V (Value) user annual income level: less than 100,000, 100,000-200,000, 200,000-300,000, 300,000-500,000 or less, more than 500,000;
• N (Need) user group preferences: models below 80,000 yuan, SUVs between 80,000 yuan and 100,000 yuan, sedans between 80,000 yuan and 100,000 yuan, SUVs between 100,000 yuan and 150,000 yuan, sedans between 100,000 yuan and 150,000 yuan, SUVs above 500,000 yuan, MP
• B (Behavior) User car purchase stage: attention, rough selection, interest, preference, intention;
• I (Interest) User interest points: space, power, handling, fuel consumption, comfort, appearance, interior, and cost-effectiveness.

The construction of user group preferences is mainly abstracted upward based on the car series that users are interested in. By crossing the price range of a car series with the type of car series, a user may pay attention to multiple cars at the same time, but in most cases the cars he pays attention to will be of the same type and price range. The feature of ethnic preference can well describe this situation, and such features have better generalization ability than car series features.

In addition, since there is an intersection of at least three dimensional features, the three features U, V, and N all appear in the form of intervals to avoid too many dimensions after intersection. There are 480 features in total after various combinations and intersections of the three types of features, forming 480 blocks in three-dimensional space. By counting the number of people in a certain car according to these 480 blocks, we can get the user distribution. Users who follow a car series can be divided into 1738 categories at most. After observation, it is found that users of a car series are mainly concentrated in 160-210 categories. As shown in the following figure:

The entire 3D coordinate system can be rotated 360 degrees, and each subcategory can be clicked individually, and the specific information of the category will be displayed on the right side of the page. In addition to the three indicators of UVN, each subcategory will be given a corresponding B label and I label, which means the car purchase stage and buying point preference. This display method is more novel and has more analytical depth than the traditional DMP that shows advertisers common charts such as users' age, gender, and regional distribution. It is a step away from the primary stage of only directly displaying indicators.

Author: Everything needs Jingsheng

Source: Everything needs Jingsheng