Talk about how to accurately push users.

Recently, I came into contact with the related problems of accurate push in my work. Now I'll sort it out and share it with you. Interested parties can discuss it together.

In operation or product work, the common precision push is divided into the following two types:

1) accurate push in the station: including advertisement in the station, information flow recommendation, product recommendation, etc.

2) Accurate push outside the station: including advertisement, short message, push, etc.

In order to avoid problems that are not focused and summarized, we only talk about off-site (mainly push) here.

For products, the purpose of accurate push is often as follows:

1) Improve user activity and retention.

? 2) improve the conversion rate (often related to business, such as pushing coupons to users)

The activity of some companies' products often fluctuates with the online push, and the push is well done, and the users are active and rising. Once the push is stopped, the product is like a stagnant pool. It may be a bit exaggerated here, but push does play a very important role in product activity and business transformation.

Accurate push is essentially a message to users. Whether a message can awaken the user's attention depends on whether the message meets the following two needs of the user:

1) useful

2) Interesting

In addition, the negative impact of message push should be considered: users should not be disturbed.

So how can it disturb users? First, push users with information they don't need (useless and boring information); Second, the number of pushes is too frequent, which is beyond the tolerance of users.

To sum up, accurate push is to push useful or interesting information to users without disturbing them, thus improving user activity and retention, or improving conversion rate.

In my opinion, there are several key steps to solve the above problems:

1. Determine the user category of the product.

2. classify each user into the category he belongs to.

3. Push relevant information to meet the needs of different types of users.

Users can be classified according to the positioning and function of products. For example, users in 58 cities can be roughly divided into looking for a job, renting a house and looking for local services, while renting a house can divide users into sharing, whole renting and short rent. For another example, Kaniu Credit Card Manager can be roughly divided into users who apply for credit cards, play credit cards and apply for loans, while playing credit cards can be further divided into credit card concessions, cash withdrawals and installments.

Knowing the general classification of users, how should each user be classified into the category to which he belongs? In other words, what kind of labels should users belonging to a specific category carry?

Then, taking Kaniu Credit Card Manager as an example, how to accurately find the dynamic label of users who belong to the loan category? First of all, sort out some behaviors that users with loan needs may have on products:

1. Click to enter the loan page.

Looking for loan products

3. Browse the details of loan products.

4. Click the "Loan" button.

5. Submit loan information

6. Check the loan progress

7. Check the positions and strategies related to loans.

8. Participate in loan-related topics.

……

Users who have the above behaviors may all be users with loan needs. Then, with these behavioral data, can we conclude that users must have loan needs? Not necessarily, such as being late, such as unconscious browsing behavior. Therefore, in order to improve the accuracy, several dimensions need to be considered comprehensively.

1. For example, how many times does the user click on the frequency dimension a day? How many times a week?

2. For example, the loan application status, which status is the user in before application, during application or after application?

(You can continue YY in more dimensions, and the goal is to be close to the bottomless precision. )

Then there was a user: Ta entered the loan page five times this week, and browsed more than four loan products, and each loan product details page stayed for more than 1 minute. Such users may be competitors who want to conduct competitive product analysis.

What kind of news is useful or interesting to users needs to be finally determined according to the positioning of the product. For example, for users who have a clear loan demand on Ka Niu, information such as loan strategy and loan activity is far more attractive than information such as credit card application and credit card withdrawal. As for how to present the pushed landing page and what to push, this part involves too much knowledge, so I won't go into details here.

Summary: In order to accurately push users, we must first make clear the purpose of our push and how to solve users' problems, then clearly classify users on products, accurately label users by collecting useful data, and finally push the information we want to users.