What is a tag system? Simply put, it means how many categories you classify users into. Of course, each user can be divided into multiple categories. What these categories are and how they are related to each other constitute the label system. There are two common requirements for the design of label systems, one is to facilitate retrieval, and the other is to be effective. In different scenarios, the requirements for these two points are different. The author has seen many product managers who work on user portraits. They are often obsessed with designing a great, glorious, and correct label system. This is often a formalist tone.

Generally speaking, there are three ideas for designing a label system:

1. Structured label system

To put it simply, labels are organized into a relatively regular format Trees or forests have clear hierarchical divisions and parent-child relationships. The structured tag system looks neat and easy to explain, making it easier to use when targeting brand advertisers. Population attribute labels such as gender and age are the most typical structured systems. The picture below shows the structured tag system used by Yahoo!’s audience-targeted advertising platform.

However, in practice, even for brand advertisers, selling non-demographic audiences is very difficult. The reason goes back to the problem mentioned at the beginning of the article: these tags cannot be monitored in principle.

2. Semi-structured label system

When used for effect advertising, the flexibility of label design is greatly improved. Whether the labeling system is regular or not is not that important, as long as it is effective. Under this idea, user labels often present a certain parallel system in the industry, and the label design within each industry is based on the highest guiding principle of "catching mice is a good cat" and must not be rigidly rigid in form. The picture below is a semi-structured tag system formed by Bluekai aggregating multiple data.

Of course, if the tag system is too confusing, it will be more difficult to launch and operate. Therefore, in practice, it is often necessary to compromise on a certain degree of structuring, unless the entire delivery logic is determined by a machine (such as personalized redirection).

3. Unstructured tag system

Unstructured means that each tag is specific to the matter and reflects its own user interests. There is no hierarchical relationship between them and it is difficult to organize them into Regular tree structure. A typical example of unstructured tags are keywords used in search ads. There are also user interest words used by Facebook, which have the same meaning.

It is already very difficult to operate semi-structured tags. Why are unstructured keywords so popular in the market? This is mainly because the market position of search advertising is so important, and a mature methodology has been formed around its keyword selection and optimization.

For brand-oriented structured tag systems, the quality of design does not seem to be very important; and for completely unstructured tags, there is not much need for design. The difficulty product dogs often encounter is how to design a reasonable semi-structured label system to drive the effectiveness of advertising. The most critical trick here is to delve into the user decision-making process of a specific industry.

Currently, Chuanglue Technology is the first CDP customer data platform in China, with outstanding performance in tag basic data integration, tag management, and customer portraits.