This article is inspired by the Zhihu Fengchenqike Zhihu column "Edison Business Laboratory" article "How do apps with JOIN product internal reference market values ??of 1 million to 1 trillion design the filtering function?" 》Inspired by

How do people perceive things? It is by summarizing and abstracting the characteristics of things. By giving different labels to various things in life, people have learned this way of quickly identifying and grabbing the target things among the numerous and complex things. Classifying and labeling things is one of the common and extensive means for people to recognize and organize things in life.

Information taxonomy often has two information classification methods: line classification and surface classification. In the APP framework, line classification is generally used for the functional architecture of the product (as shown in the Maoyan movie product architecture), while surface classification is used for category subdivision specific to a certain function. Line classification guides users to the corresponding functional categories, while surface classification helps users quickly locate the corresponding goals in complex categories. The focus of this article is the filtering tags that belong to face classification.

One of the characteristics of the Internet is that it brings a large amount of information to people, but it also makes it difficult for people to accurately capture the information they need. We know that online platforms are oriented to a large number of users in different regions, with different hobbies and different lifestyles. So when all the information is gathered in one place, how can we help different users find information that meets their own needs? There is no doubt that that is the filter tag. So when it comes to the filtering function in the mobile terminal, we have to first understand the classification of mobile content. The biggest difference between the mobile terminal and the web page is that the mobile terminal is mostly about instant and improvised operations performed by consumers in fragmented time and scenarios. Therefore, consumers have higher requirements for the speed and accuracy of search compared to the PC terminal. Users on mobile Most people in the scene don’t have the patience to browse slowly (except for recreational information). This puts forward higher requirements for the classification and screening of information. In my opinion, the label-based content classification is actually part of the product architecture, but the product architecture is an organic combination of the functional modules of the system, while the label-based content classification is actually to classify the target information according to user needs and enterprise appeals. Reorganizing, or in other words, a purposeful reorganization of the content in order to cooperate with the realization of the functional goals in the product architecture.

So, how to extract tags?

First, we should consider the demands of users, then consider the demands of the platform, and then balance the three parties by fully weighing the first two and combining the attributes of the content itself.

1) What are "user demands"? User behavior is often purposeful. For products, there is often two-way selectivity between products and users. In other words, while the product selects target users by deliberately creating a certain tone, the users are also choosing the product. Since users choose a product, they must have certain expectations, and this expectation is the user's appeal.

2) What is "platform appeal"? The platform's appeal lies in what purpose the platform hopes to achieve through this design when planning a certain function or a certain page.

3) What is "content itself attribute"? The properties of the content itself are the characteristics of the content. For example, when talking about clothes, it must involve fabric styles, and when talking about community content construction, it must involve the best posts, the latest posts, etc. Another example is the Helijia manicure service mentioned below, which must involve time attributes and geographical attributes.

So how to extract it specifically? Take keep fitness as an example:

The user’s appeal is to find fitness projects. I have just started fitness, so I can choose the primary difficulty through all difficulties. If I want to train the abdomen, I can look for training on related parts. Fitness exercises involve training parts, intensity, whether equipment is involved, venue, time, training category, etc. These are all things that need to be considered. This is reflected in the filtering tags under the entire training course label. This is the user's appeal.

But if we take a closer look at the picture below, under the training course page there will be classification methods such as latest launches, recommendations for you, etc. Obviously, from the perspective of the platform, the platform hopes that users can keep abreast of new content and new training projects on the platform, so that users can feel that the content of the platform is constantly enriched. In other words, it is to import traffic for new projects, from From a platform perspective, the classification method of adding fitness items can add up the exposure of each fitness item and give users the illusion that the platform has a large amount of training content.

From the perspective of content attributes, this content classification method spanning two pages itself is based on the fact that fitness involves more complex factors that need to be considered.

When we extract content tags, we should note that the fewer tags, the rougher the target description and the lower the accuracy, that is, rough classification. The more labels there are, the more precise the final goal will be, that is, refined classification. However, this does not mean that the more detailed the classification, the better, but that reasonable choices must be made based on the demands of users and enterprises. The reason is that excessive classification is not only equivalent to users quickly obtaining target items, but also increases the user's screening burden and consumes unnecessary system resources. The middle scale is a standard to measure the design level.

After the designer uses filter tags to organize the content according to the above standards, it needs to be visually presented in the form of pages.

So, how does the designer present it? In my opinion, there are two forms:

One: One tag corresponds to one page (rough classification).

This tag is the end point of the line classification, and below it is a large amount of category information. Obviously, this single tag corresponds to a category page. However, this page does not introduce auxiliary tags to help users classify information in a refined manner. This makes information classification stay at the rough classification stage. So, does information have to be classified in a refined way? In fact, sometimes rough classification is more conducive to quick browsing of information for users.

As shown above, Toutiao, 36Kr, and enjoy all use a content classification method where a single tag corresponds to a single page. What is the reason?

Toutiao provides news content. Then the nature of the news content itself determines that its classification is divided into several major categories in various fields and industries. Moreover, news has obvious timeliness. Therefore, the classification of large lines can basically cover the important news happening in society. There is no need to subdivide it again; from the perspective of corporate demands, Toutiao targets users nationwide, so overly detailed classification is not conducive to meeting the needs of different users. From the perspective of user demands, users often browse headlines with a recreational mentality and have no desire to filter professional news in related subdivisions.

Thirty Six Krypton’s slogn is “Provide entrepreneurs with the best products and services. Its own corporate attributes have already labeled the content provided by the platform, and because of the timeliness of news, The content produced by its own platform is not so complicated that it needs to be classified.

From the perspective of its corporate attributes, Enjoy is positioned as a "selected gourmet e-commerce", so its content is bound to be small but refined. , so when users filter content, the company's own attributes allow the platform to fine-tune the content before presenting it, so there is not much content under the "Localization" tag. The most important thing is that this page belongs to the right side of the homepage. If you compare the displayed content after searching for keywords on the search page on the side, you will find that the local selection page will be intelligently arranged when you compare the same keyword search in the local selection. The reason is that the former function focuses on highlighting local services and nationwide delivery to guide users to understand. We know what services are available under the platform for this keyword, and the latter focuses on refined screening to help users quickly find target items. We know that products and users choose products in a two-way manner. Users are also choosing products. I chose this platform because of the "selected food". What is selected food? If the user enters the word "hot pot" on the platform, and a large number of merchants and filtering tags appear like Dianping, I think this will definitely make people happy. The user platform has raised questions. Therefore, from the perspective of user demands, users do not expect to see too much content under the "hot pot" label, so the filtering label will not have too many segmented filtering labels like Dianping. It only provides two general filtering dimensions: distance and price (see the figure below)

In fact, strictly speaking, most of our mobile pages use single-tab pages. The content of the tab page is generally a functional page, such as the page corresponding to the tab in the mobile app tabbar; or it is quick browsing content, such as the news APP in the above example; or the content is of high quality but not enough. If you have three attributes, you should consider adding corresponding tags to quickly filter target information.

Two: Multiple tags correspond to one page (refined filtering). ).

The premise that multiple tags correspond to one page is that the page contains a lot of content. In order to help users filter out the target information among the numerous contents, through multi-dimensional characterization of the target items, This allows the system to quickly filter out target items. Multi-dimensional filtering does not mean that the more dimensions, the better, but the optimal solution after weighing the attributes of the user, platform, and content itself. Of course, all filtering must be done in order to help users first. Rough filtering, and then helping users perform refined filtering, which can be divided into:

1. One main tag + multiple filtering tags (randomly) based on the degree of satisfaction the user needs for each tag. Filter combination).

In this case, the main tag is the user's rigid need, and the filter tag can help the user quickly target the target.

Obviously, the main label is the item of manicure. When the user clicks on the main label of nail art (the end point of the line classification), he will enter the corresponding content page. This page helps users quickly narrow down the search scope of target services through tags of different dimensions (surface classification).

1) General dimensions (rough screening).

Different platforms and services will have general filtering dimensions related to their fields that can meet the needs of most users.

As a service item, manicure should not only meet the needs of users from the perspective of the platform, but also enable the platform to retain merchants as much as possible. The default comprehensive sorting should be to give all merchants an opportunity to obtain traffic entrance. As a consumer service, users generally focus on price, popularity, praise, etc. Therefore, we can see that the general dimension tags provided by the platform are two categories of rough filtering tags: comprehensive sorting and price.

2) Personalized dimension (first-level fine screening).

Since different platforms and different services will have general screening dimensions in this field, then there will also be personalized screening dimensions related to the platform and the service. These screening dimensions tend to better reflect the characteristics of the platform and services.

The personalized dimension of Helijia lies in the following "available appointments today", "available appointments at the current location", and "nearby craftsmen". These three filtering items are based on the attributes of the mobile platform and the service itself, and the three characterization dimensions of the service that are in line with consumer demands are extracted. Of course, these three labels also belong to refined classification. If it is JD.com, then JD Daojia must be included in this dimension.

3) Refinement dimension (secondary fine screening).

The filtering on the right is to meet the user's demands for further refined screening. Therefore, the price is more subdivided, the star rating is more subdivided, and the time is more subdivided. The refined dimensions are actually based on the general dimensions. further. The refined dimension takes into account the user's frequency of use and the page space required and generally adopts hidden forms, such as side sliding bars, pop-up boxes, etc.

2. One main tag + multiple filter tags (must be satisfied at the same time) + multiple tags (random filter combination).

When the user clicks on swimsuit and beach to enter the page, we find that the filter items account for 1/3 of the page. When the user clicks on the hotel and enters the page, we find that the middle page is actually a super filter tag page. After the Oxygen APP first enters the page, the effect is as shown on the left, and when the user slides up and down, the effect is as shown on the right. The reason why users are presented with filtering labels such as size is that when women filter underwear, they must first consider filtering such as size, and this type of filtering is often fixed and necessary for users. By default, the first operation a user performs after entering the page is to select filter tags such as size. Regardless of whether the user performs operations or not, the system will automatically hide them after sliding the page. In fact, filter tags such as size are equivalent to the separate filter tag page on Qunar.com (middle). It's just that Oxygen integrated it into a page in a clever interactive form, and naturally hidden it in conjunction with the user's scrolling action.

Therefore, the filtering dimension of this type of tag page is actually an additional dimension---rigidity---based on the three dimensions included in a main tag + multiple filter tags (random filter combination) page demand dimension. The filtering tags under this dimension help users filter out content that is of no value to users.

So why do these applications use precious page resources for the placement of filter tags? If you read the content of this article carefully, you will find:

1) One tag corresponds to one page In this form, in fact, the content does not need to be filtered to meet the needs of users in this scenario of casual browsing. Users do not have specific needs for content, or users have relatively arbitrary standards for content needs.

2) When it comes to the form of one main tag + multiple filter tags (random filter combination), although most of the content can meet the needs of users, due to the user's certain purpose , so a low matching degree often means that users need to spend greater costs.

3) Finally, there is a main tag + multiple filter tags (must be satisfied at the same time) + multiple filter tags (random filter combination). If the content in this form cannot meet any tags, the content may be invalid for users. Therefore, we found that whether it is Oxygen or Qunar.com, the filtering tag will be placed in such an important position. The reason is that users are not interested in the content. The standards are relatively high, so information that does not meet user demand standards is invalid information for users. Therefore, it is more important to present filtering labels than to display content.

It can be seen from this that users have extremely high requirements for accommodation information. If the browsing content of a single-tab page like 36 Krypton may have the possibility of arousing users’ interest in browsing, then in many The filtering tags must satisfy the content on this page at the same time. Irrelevant content not only has no possibility of arousing the user's desire to browse, but may even cause the user's irritation. It is not difficult to understand that Qunar.com will separate the filter tags into an independent page to first help users filter out a large amount of irrelevant information, and then help users obtain target information through a multi-filter tag page. Designers do not need to worry about this Disturb the user, because accurate filtering is far more important to the user than browsing the content.

In summary, how to reasonably present the extracted filter tags?

First of all, we need to study clearly, the position and role of the main tag in the entire product architecture? Use this to decide whether to add filter tags.

Secondly, we need to study clearly, what are the user demand standards for the content under the main tag? What are the scenario-based needs of users browsing this content? Based on this decision, add the category of the label.

Once again, we need to study clearly. If we need to add tags, how should the form of adding tags be presented? Whether the form is presented in a hidden form or displayed on the page should be determined according to the intensity of the user's need for content filtering. The stronger the user's need for filtering, the filtering will be in a form that is convenient for the user to operate. As for the page, it will not disturb the user. It will please users, and when the user's demand for filtering is not high, hiding the filter may better arouse the user's favor. Of course, factors such as the appeal of the platform and the attributes of the content itself should also be considered. A balance should be made after considering all factors. The value of a designer lies in the ability to balance factors to maximize results.