One article to quickly unlock: the entire process of the label system

When should the labeling system be built in the annual work of data analysis? How can it be done well? How can it be done accurately? In this article, the author starts with the pain points of building a labeling system and common mistakes in the labeling system, and gradually explains to you the methodology of breaking through the labeling system: understanding needs, sorting, classification, and exploration. Recommended for those who are interested in data analysis and labeling systems.

The label system is definitely the most worthwhile thing to do at the beginning of the year in the annual data analysis work. This is because it is related to all work, including business analysis, delivery analysis, user portraits, recommendation strategies, product operations, etc. All of them rely on labels.

If the labeling system is well designed, there will be enough materials for subsequent analysis and experience can be accumulated. If the labeling system is poorly designed, not only will the effort be wasted, but there will also be nothing to rely on when doing in-depth analysis later.

So how do we do it? Let me share with you briefly today.

1. Common mistakes in labeling system

The most common mistake is to put all kinds of things into a label.

• When uploading articles, just post a few: news, hot topics, products...
• When the product is put on the shelf, just post a few: function, packaging, price...
• When an event is released, just post a few: name, gifts, discounts...

Not to mention, when labeling users, "high value", "potential", "like XX" are randomly attached. Even similar names such as "high value", "high quality" and "high quality" exist at the same time.

These bad habits can be seen everywhere when doing user portrait projects. People often boast to me: "Teacher Chen, we are so awesome, we have added more than 3,000 user tags..."

At this time, you just need to ask him:

• Of the 3,000 labels, how many are actually used for business purposes?
• How much value do 3,000 tags generate?

He was dejected and said, "I'm still exploring how to apply it..." and then ran away.

Why? Because these tags are just a bunch of dimensions lying in the database. If you want to use them for business, you must first consider: what are the business needs and why they need to use tags.

2. Breakthrough, start with understanding needs

When building tags, there are at least 3 completely different types of requirements.

Management: Rapidly identify the need for value. Management is most afraid of seeing hundreds of pages of PPT reports. Tags can effectively refine business meaning and identify the most critical factors.

for example:

• Label users and distinguish between ABC types of users
• Label channels to distinguish between stable and unstable channels
• Label products to distinguish between before, during and after their life cycle

In this way, when performance fluctuates, management can see at a glance: Oh, it’s the problem in XX place. This saves a lot of time.

Operations department: Find the need for planning inspiration. The operations department likes to ask:

1. What do users like?
2. Where are the users?
3. Will users buy it?
4. If you don't buy it, can I send you a coupon?
5. If coupons don’t work, how about giving gifts?

These questions, in the final analysis, all revolve around the “five elements of planning”. It would be great if the answers to business questions could be clearly told through labels (rather than fragmented data) (as shown in the figure below).

Frontline employees: The need for clear answers to questions. Frontline employees face customers directly, and they know user behavior and needs much better than data analysts thousands of miles away. What frontline employees need is not: "Teach me how to do it" (actually they can't teach me), but:

• When customers ask me questions, I can answer them clearly.
• When headquarters came to inspect, I was able to hand in my work accurately.
• When there is a sales opportunity, I can find the corresponding tools

for example:

• When customers inquire about products, I can quickly find information
• There is a reward policy, I can quickly check how much I have achieved
• When an event is online, I can quickly check which guests can participate

Such clear guidance is the best tool.

After carefully understanding the business needs, you will find that: there is no need to label a large area. It is even more unnecessary to use a large area of ​​labels for business use! Providing a small number of labels, cultivating business usage habits, and gradually building a complete system are the keys to project success.

So, where do you start?

3. Sorting, start with the simplest

Note that the difficulty of implementing the above three types of requirements is different.

The easiest thing to achieve is the needs of frontline staff. In theory, you only need to label the activities, products, and articles that the frontline staff frequently searches for in a standard format and put them into the library (as shown below).

But! This cannot meet the needs of the front line.

Because it is difficult to search for information on the front line. For example, if 30 activities are launched online at the same time this month, there may be only two or three that are of interest to the front line. And the front line staff and customers often give them nicknames, resulting in strange search keywords. If you directly open the tag library for query, the usage rate is often low and the search accuracy is low.

Therefore, the tools provided to the frontline can be further optimized:

1. Actively collect frontline opinions and optimize keyword searches
2. Common and popular tags are actively pushed to front-line users for understanding
3. Optimize tools such as sales assistant to highlight key tags

Only in this way can the frequency of label use be increased, and there will be an opportunity to improve front-line efficiency.

4. Classification and identification value are the most important

The second type that is easy to promote is labels that identify value.

• Firstly, the definition of value is relatively simple, clear and easy to do.
• Secondly, management often looks at the value label to increase its presence.
• Thirdly, it can be used in daily diagnostic index fluctuations and has a high appearance rate.

Even if you do nothing else, you must prioritize creating these valuable identifiers.

Common ones, such as:

1. Product value labels: hot items, traffic-generating items, profit items, matching items, defensive items
2. Channel value label: past input-output ratio, output volume + stability
3. User value label: Consumption already generated + consumption expected to be generated in the future

(As shown below)

1. The value of a product can be calculated simply by calculating the cost and pricing.
2. The value of a channel can be clearly seen by calculating the input and output of the channel and observing past trends.
3. User value, the actual value can be directly counted, and there may only be estimated output, which requires some work

(As shown below)

The only difficulty here is to popularize this concept among the management. It is very likely that the company has never done similar labeling before, and there is no consensus among the management on "what is a high-value user" and "what is a high-quality channel", so it may be difficult to mention it for the first time. However, as long as the company is not so stupid that it doesn't even know what its own products, channels, and users look like, it can gradually promote the application of labels. After all, reducing the pressure of reading reports and focusing on core issues are the common demands of everyone.

5. Explore and gradually achieve accuracy

Of the three types of needs, the most difficult to meet are those of the operations department. Labels such as “like” and “preference” are very difficult to create.

1. Insufficient data: It is not Touteng, and insufficient user decision paths can be collected
2. Difficult to define: What does it mean to like something? Does buying something count as liking something? Does just looking at something count as liking something?
3. Unstable: He didn’t like it at first, but he started to like it once you lowered the price…
4. Difficult to see results: The calculation shows that he likes it, but the promotion channel cannot contact the customer...

Not to mention, even if it works, what proportion does the advertising copy, promotional offers, and user needs each account for?

Therefore, if you want to do this clearly, you will need multiple iterations.

The iterative method is to start from the area with more data and move to the area with less data.

for example:

1. From a promotional perspective, separate those who like to get the most benefits first.
2. From a consumer perspective, first study the people who make the most frequent repeat purchases
3. From a behavioral perspective, first understand the people who interact most frequently

These extreme groups are usually the big contributors to performance, and they have a lot of data, so it is easy to summarize the rules. And when high consumption does not consume, and high activity does not convert, the business department will rush to find a solution, and can further verify the accuracy of the prediction by combining business actions.

As for users who have very little data, they can first fry fish according to a fixed recommended route (as shown below) and combine it with business actions to test user needs and gradually improve the accuracy of prediction.

VI. Summary

Labeling is extremely important. It is an important tool for quantifying qualitative factors and providing value judgments. It is a very basic construction. However, when doing labeling projects, it must be combined with business analysis (for management), activity support (for operations), and system tools (for the front line). You cannot just contribute in obscurity. Otherwise, everyone will think that you can quietly refine a furnace of elixirs, not participate in the process, and not use them. At the end, you will definitely be disappointed.

Here is another extended topic, how to start from scratch and iterate high-quality labels. I will continue to share it later.

Author: Down-to-earth Teacher Chen

WeChat public account: Down-to-earth Teacher Chen (ID: gh_abf29df6ada8)