E-commerce operations: traffic source analysis

For e-commerce platforms, accurately identifying the effect of each traffic is one of the most important daily operational insights. The subsequent conversion of users at each traffic entrance is the core data indicator for evaluating a traffic pit. So how can we correctly attribute each traffic exposure, click and purchase conversion to the corresponding pitfall?

1. Introduction

2. Data underlying solution

The basis for traffic attribution is that the App’s tracking data has been basically perfected, and most users’ behaviors on the App have been stored through tracking data records.

2.1 Added PageId field to tracking log

First, you need to add the PageId field to the customer tracking log. The function is to assign a new PageId to a new page every time a user jumps to it; and when the user clicks back, no new PageId is generated. The closer the page browsing behavior is to the current time, the larger the PageId is, and there will be no duplication.

2.2 Analyze the tracking logs for link analysis

You need to first determine the end of all chains, that is, all the add-to-cart events.

Then determine the head of all chains. Since the entrance to e-commerce products may appear on multiple pages of the App, select all basic pages (enumerable) as the head.

Use the end to left-associate to get the nearest head end and get the interval range of each chain.

Use each interval to filter all corresponding click events and remove invalid events to obtain the complete path of each link: [1, 2, 9, 10, 11].

Diagram of the steps of a user's shopping process in the App

2.3 Implementation solutions for various special scenarios

1) Jumping between product detail pages

2) Jump across levels (jump back to the same page)

3) Jump across levels (jump back to a different page)

4) Jump back to the entry page directly

2.4 Attribution Scheme

Through the above tracking solution of embedded logs, we can obtain the complete and orderly link before the user adds the purchase in the App. Based on this link, we can perform attribution analysis. There are five basic attribution schemes commonly used in the industry, as follows:

First contact model: When multiple "unattributed events" contribute to the same "target conversion event", the first "unattributed event" is considered to have 100% of the credit.
Last touch attribution: When multiple "unattributed events" contribute to the same "target conversion event", the last "unattributed event" is considered to have 100% of the credit.
Linear attribution: When multiple "unattributed events" contribute to the same "target conversion event", it is considered that each "unattributed event" shares the credit evenly.
Position attribution: When multiple "unattributed events" contribute to the same "target conversion event", the first and last "unattributed events" are considered to have 40% of the credit each, and the remaining "unattributed events" share the remaining 20% of the credit.
Time decay attribution: When multiple "events to be attributed" contribute to the same "target conversion event", it is believed that the closer the "target conversion event" is, the greater the contribution.

Our company’s current size is not very large, so there is no need for complex attribution. Therefore, the last touch attribution solution is adopted to achieve this, that is, the last clicked entrance before adding to the cart is attributed to it. In this way, the underlying data implementation of e-commerce traffic attribution is completed.

3. Front-end page display

3.1 Entry Classification

First, classify all traffic entrances within the App and divide them into two levels based on location and size. I call it the primary and secondary traffic sources. Examples of some primary and secondary traffic sources:

3.2 Data Indicators

Module exposure number: The number of people exposed to the traffic entry module
Number of module clicks: Number of people exposed to the traffic entry module
Number of people who have been exposed to the product: Number of people who have been exposed to the product from this traffic portal
Product exposure ratio: product exposure from this traffic portal / total product exposure
Number of product clicks: Number of product clicks from this traffic portal
Number of payers: Number of people who paid for the product from this traffic entrance
Percentage of paying users: Number of users who paid for products from this traffic portal / Total number of users who paid for products
Payment amount: The payment amount of the product from this traffic entrance
Payment amount ratio: payment amount of goods from this traffic entrance / total payment amount of goods
Number of payments: The number of payments for products from this traffic entry
Product exposure click rate: the number of people who clicked on the product from this traffic entrance / the number of people who were exposed to the product from this traffic entrance
Product exposure payment rate: the number of people who paid for the product from this traffic entrance / the number of people who were exposed to the product from this traffic entrance
Product click-through payment rate: the number of people who paid for the product from this traffic entrance / the number of people who clicked on the product from this traffic entrance
UV value: the amount of money paid for products from this traffic entrance/the number of people who were exposed to the products from this traffic entrance
Average order value: the amount of money paid for goods from this traffic portal/the number of people who paid for goods from this traffic portal
Click value: the amount of money paid for the product from this traffic entrance / the number of clicks from this traffic entrance

3.3 Time dimension and population

Crowd: Overall, New Customers, Old Customers

Time dimension: today, yesterday, daily average in the past 7 days, daily average in the past 30 days

3.4 Traffic Source Platform UI

The homepage of the traffic source analysis platform can display specific data of different traffic sources and different data indicators for different groups of people at different times. The red and green data represent the current time's year-on-year data. The core data indicators also add percentage data.

You can analyze the source data of e-commerce as a whole, or you can look at the source data of each product separately.

The indicator visualization chart is shown below:

1) Data trends of different traffic sources

2) Data trends of a single traffic source

3) Hourly traffic source trends

4) Aggregate data trends by time interval

4. Summary

Traffic source analysis is an indispensable part of daily operational analysis of e-commerce companies. Building a traffic source analysis platform can greatly reduce the burden of data analysts in obtaining data, and also allow operational products to directly obtain corresponding data in a visual way for self-service analysis, greatly shortening the queue cycle for obtaining data, allowing the business to efficiently perform daily operational iterations.

Author: A data person's private land

Source: A data person’s private land

<<: Why is it that the bigger the brand, the harder it is to gain influence?

>>: The latest news on Zhengzhou lockdown in 2022: What are the specific lockdown areas? When will the blockade be lifted?