A brief discussion on user link analysis from the perspective of traffic

User link analysis from the perspective of traffic plays an indispensable role in Internet industries such as short videos, live broadcasts, and e-commerce. This article takes live broadcast e-commerce as an example to sort out a relatively complete link diagnosis framework from the perspective of its revenue, and conducts attribution, problem drilling, and final positioning.

01 What is a traffic link

In industries with "strong recommendations" and "strong content orientation", the traffic chain is generally recommended by the recommendation system or manually directed purchases, followed by the back-end system to complete the entire chain of content/product distribution, exposure to users, and user completion of consumption/conversion. The traffic funnel mainly refers to the process under the traffic chain, starting with system distribution and exposure, each stage in the middle contains some behavioral characteristics of users on the platform, and finally completing consumption/conversion on the platform. That is, this link represents funnel analysis from the perspective of traffic, which plays an indispensable role in Internet industries such as short videos, live broadcasts, and e-commerce.

This article mainly takes the live streaming e-commerce industry as an example, and from the perspective of live streaming e-commerce revenue, it sorts out a relatively complete link diagnosis framework, and attributes problems, drills down to problems, and finally locates them under this framework.

02 How to disassemble the traffic link for beginners

If the ultimate goal is to achieve the GMV of live streaming (i.e. the total amount of live streaming transactions under live streaming e-commerce), the traffic chain can be broken down into the following core logic:

Live broadcast GMV = Number of people broadcasting * Average broadcasting time per person * Exposure per broadcasting time * Click-through rate * Average viewing time per time * Number of orders per hour * Average order price

In this decomposition method, the revenue of the entire live e-commerce can be attributed to four templates: content supply, streaming efficiency, user viewing efficiency, and user purchasing efficiency.

Content supply:

This module mainly attributes GMV from the supply perspective. For example, in the past week, the anchor may not broadcast due to holidays (such as the Spring Festival), so there is no live broadcast income.

Streaming efficiency:

This module mainly explains from the perspective of streaming efficiency, that is, it indicates how much exposure the platform will give to the live broadcast under a certain broadcast duration (under natural traffic, commercial traffic, etc.)

User viewing efficiency:

This module mainly starts from the user viewing efficiency. If the CTR is low, it means that users will not choose to click on the live broadcast that has been exposed, indicating that the live broadcast room has poor ability to attract people. At this time, the seller can consider changing to a more attractive cover, or live broadcast preheating and other operational strategies; and if the average viewing time is low, it means that after the user clicks into the live broadcast room, he chooses to exit after watching for a short time, indicating that the live broadcast room has poor ability to retain people. To address this phenomenon, the seller can be pushed to further explain his products in detail, design more attractive live broadcast room activities, meet the audience's expectations, etc.

User transaction efficiency:

This module is mainly based on user transaction efficiency. If OPH is low, it means that users place fewer orders for live broadcasts and have low purchasing desire; while the average price per unit measures the price of the seller's inventory. If in the abnormal analysis, if OPH is low but the average price per unit has increased significantly, it can generally be inferred that the seller is raising the price of his inventory and using the higher gross profit to make a profit.

Generally speaking, merchants with ideal growth and profitability may see an increase in both OPH and average unit price in the short term; most merchants with normal profits will have a trade-off between OPH and average unit price; and some merchants with problems may see a decrease in both OPH and average unit price, that is, the number of user orders decreases, and although the pallet price has been lowered, it still fails to stimulate user consumption. In this case, further drilling down is required.

03 How to locate and drill down problems in the traffic link funnel

In the previous part, we learned in detail the meaning of different parts in the traffic chain. In this part, we will focus on how to locate and drill down the problem. If in part 2, we locate that CTR (click-through rate) and OPH (orders per hour) have a significant decline, then can we conclude that the live cover is not attractive enough, users are less willing to place orders, and purchasing activity is low? In most cases, such a conclusion is not solid enough. In most cases, any analysis should adhere to the following principles:

Eliminate external factors: When drawing conclusions from any analysis, you should first exclude external factors such as cyclical influences (such as off-season and peak season), external influences (such as competitor actions), etc. After confirming that it is an internal problem, continue with internal analysis.

Internal analysis locates the target group: It can be broken down from multiple angles, such as the user angle (demand), seller angle (supply), platform angle (management, product, strategy, etc.) for breakdown and positioning. Because many times the changes are due to changes in a specific group, so general internal analysis, first of all, needs to make certain direction guesses based on demand, and secondly, needs to verify its own guesses.

If there is no effective business perception and speculation, we can follow the following framework, carry out some expansion (of course, the workload is relatively large), and then find the main problem groups from the process.

1. Horizontally, distinguish traffic links of different traffic types

Live streaming GMV = organic traffic GMV + commercial traffic GMV + company/department adjustable traffic GMV

First of all, why do we need to distinguish between traffic types?

Traffic generally refers to the recommended units controlled by the system or manually. If the recommended system or manual purchase is large, the work/live broadcast room will generally be recommended to more users and get more exposure. The types of traffic can generally be divided into three categories: natural traffic, commercial traffic, and company/department controllable traffic.

The purpose of distinguishing traffic types is to monitor traffic links under different traffic types, and to locate the target stage for adjustable traffic types and make targeted improvements. Natural traffic generally refers to traffic calculated by recommendation algorithms and automatically promoted by the system. Its recommendation effect is closely related to the author's own characteristics. It is relatively black box and difficult to analyze and push its traffic strategy. Commercial traffic refers to users (merchants) purchasing traffic themselves, and the platform completes its promotion and exposure in the form of certain advertising products.

Under this traffic, the more you put in, the more commercial exposure you get. Therefore, the traffic entrance of this department is purchased manually, and the subsequent links can also be disassembled and analyzed to analyze the conversion process of users under the advertising (commercial) traffic of this department.

The company/department's adjustable traffic generally means that the company/department has a certain proportion of traffic and can adjust this part of traffic to specific users or give it to users in a specific way. It is also a department that can intervene and adjust, so it is also important to monitor its link effect.

2. Vertically, drill down from the perspective of users/sellers/platforms, etc.

Live broadcast GMV = Buyer group 1_GMV + Buyer group 2_GMV + … + Buyer group N_GMV

or

Live broadcast GMV = Seller group 1_GMV + Seller group 2_GMV + … + Seller group N_GMV

or

Live broadcast GMV = Product type 1_GMV + Product type 2_GMV + … + Product type N_GMV

In the previous part, we broke down GMV by traffic type from a horizontal perspective. This part will mainly break it down from a drill-down perspective.

Imagine a scenario like this: if you see that there is a certain decline in each stage of the traffic chain under natural traffic, what should the operators do? Should they make certain improvements to all merchants? Or for all users of the platform? This is obviously unrealistic, so this part is mainly to further locate the traffic chain under different traffic types from a vertical perspective.

Drill down from the user's perspective:

Generally speaking, users can be broken down into: user basic profile, user active behavior, user transaction behavior, and user-host relationship behavior. This can be broken down based on the normality of the business, such as: fans and non-fans, new customers and old customers, the range of monthly transaction orders, etc. Here, combined with traffic type and traffic link, we can get the following conclusion:

Under XXX traffic (natural? user-invested commercialization? platform self-control?), XXX users (by distribution) have a significant decrease in XXX stage (is it insufficient supply? or low streaming efficiency? or low viewing efficiency? or low purchasing efficiency?);

Suppose our positioning here is: natural traffic, non-fans, and low viewing efficiency.

Then the suggestions can be: for users who have not followed (solving the problem of users), the anchor can promote some new fan benefits in the live broadcast room to increase the attraction and influence on new fans, attract them to enter the live broadcast room, and enhance their ability to retain people (solving the problem of traffic links)

other:

From the perspective of sellers and platform product types, the disassembly method is similar to that of users, so I will not go into details here.

04 Conclusion

Above we have completed a gradual understanding and simple diagnosis of the traffic link. In summary, there are three major modules: locating the traffic stage; locating the traffic type; locating the target group type.

In actual applications, you can first locate the traffic stage, and then drill down to the traffic type and target group under the target stage; you can also locate the target group first, and then drill down to the traffic stage and traffic type. The specific analysis can be reasonably laid out according to business needs, and there is no strict order requirement.

Author: kk, member of the "Data Creator Alliance".

WeChat public account: a data person’s private land