How important is the attribution model in multi-channel promotion? The secret to quickly reducing conversion costs!

Today I would like to talk to you about the value of applying attribution models to the process of online promotion . At present, there are many classifications of attribution models abroad. Each model has its own advantages and disadvantages and applicable scenarios, such as the linear attribution model (average credit), the last interaction attribution model (which is what I called LAST TOUCH in the afternoon), the U-shaped attribution model, etc. Today I will tell you about the most common attribution models.

What is an attribution model? According to my own understanding, when the attribution model is applied to online marketing , the significance of the attribution model is to reasonably allocate the value contributed by each channel to the order based on the consumer's multi- channel access path.

For example:

An internet user first saw an advertisement for an English training class of an educational institution on a certain website and had a certain impression of this educational institution. After a while, he searched for English training classes on a search engine and saw the promotional information of this institution. He clicked on it, learned more about it and thought it was good, but he had not decided whether to place an order and closed the website.

A few days later, he searched for the brand name of the agency, entered the agency's website through natural search, made a consultation, and placed an order. From the beginning to the end, this netizen experienced display advertising, bidding advertising, and natural search. There are three paths, so to which channel should his conversion be attributed in the end?

The following content is extracted from the Internet :

I learned from my friends that more than 95% of domestic e-commerce companies calculate ROI based on 24-hour last- minute sales, such as JD.com , Vancl, Amazon , Vipshop , Suning , Dangdang, etc.

There are three mainstream statistical methods abroad:

1. Calculated based on the last touch of 30 days, direct/bookmarked traffic has the lowest priority, followed by CPS alliance, then SEO , then SEM , and then display advertising; among similar ads, attribution is calculated based on the last touch;

2. Calculated based on a 7-day cycle, the contribution factor of each type of traffic is calculated based on the distance from purchase = (count of touch points – rank of this touch point +1)* (time to purchase – time of closer touch point to purchase)/sum(time of touch points to purchase);

3. Calculate the contribution factor based on the average distribution of each touch point within 14 days.

Last-click P&L is commonplace these days and, while deceptive, is straightforward. The rule of the game for media performance ROI for most advertisers is = cost/number of single-click conversions within 24 hours. Other media that help with conversions are ignored in the formula for calculating ROI.

Through the application analysis of the attribution model, we can solve the following problems in future work:

1. Problems that attribution models can solve

1. Improve advertising effectiveness evaluation

Typically, evaluation of advertising effectiveness includes exposure, clicks, and direct purchase indicators, but the contribution of some marketing channels to purchases may be reflected in indirect contributions.

For example, users may be accustomed to obtaining advertising campaign consultation through Internet advertising channels, and then enter the website browsing activities and make purchases through SEM keywords. Under normal circumstances, the order contribution will be calculated in SEM, but Internet advertising will also generate advertising exposure and indirect support for the order. Therefore, this indirect contribution to orders will make the evaluation of advertising effectiveness more rational.

2. Reasonably arrange SEM keyword strategy

When SEM is launched, it is usually divided into brand words, competitor words , product words and general words, and users will show a certain inheritance on the SEM channel. For example, search engines are an important channel for users to obtain information. If some users want to buy long-sleeved bottoming shirts, they may first search for "long-sleeved bottoming shirts", and then click on the purchase keyword of the e-commerce website to enter the website. This time the users may not make a purchase.

Because users usually choose to compare prices and reputations on various websites, they will feel that a certain cat is good after confirming the information. Or if it’s cheaper on JD.com, just search directly, click on the brand word to enter the website and place an order.

During this process, the user completes the click trajectory from product words to brand words, and product words make indirect order contributions to brand words. By analyzing the user's trajectory, we can analyze the impact of product word placement on brand words, especially when controlling costs, and effectively evaluate changes in brand area traffic and target setting.

3. Reasonably arrange advertising strategies on the same channel

Similar to the different keyword influencing factors mentioned above, users will also have click dependence on the same channel. For example, a user sees Tmall 's promotional advertisement on the NetEase portal and does not complete the purchase after the first click. However, the second time the user may also get used to clicking from the NetEase portal to enter the website and place an order.

In this case, we can evaluate the user's dependence on a certain advertising channel by analyzing the user's clicks on different advertising forms or media in a channel, and then guide advertising scheduling and execution.

4. Effectively arrange the marketing rhythm among multiple channels

Normally, users will not make a purchase by just clicking on one channel to enter the website, so multi-channel coverage or integrated marketing communications will become the main marketing strategy. The key is how to evaluate the user's click trajectory in an activity, grasp the advertising sources at different purchase stages of users, and then effectively arrange the marketing rhythm. Multi-channel path analysis can capture the multi-channel browsing tracks of different users and discover the path trends of users through aggregation, providing a reference for knowing the marketing rhythm arrangement.

2. Analysis of channel value by attribution model

If a conversion path contains multiple channels, the assisted conversion value will be calculated for each assisting channel, so the sum of the assisted conversion values ​​of a multi-channel path will be greater than the final value of the path conversion. If a channel appears multiple times in the path conversion process, the assisted conversion value will be calculated multiple times. In other words, GA counts every assisted conversion and final conversion from every channel, and counts them repeatedly (if any).

Let us illustrate the above situation with an example. Assumptions: 1. The order value is 100 yuan; 2. The weights of assisted conversion and final conversion are the same.

Assumption 1

The visitor went through four channels A→B→C→D to achieve conversion, so the contribution weight of each channel is the same, which is 25 yuan.

Hypothesis 2

The visitor went through four channels A→B→C→B→D to achieve conversion, but channel B appeared twice in the conversion path. Therefore, the contribution value of channels A, C, and D is 20 yuan respectively, and the contribution value of channel B is 40 yuan.

Assumption 3

The visitor went through four channels A→B→C→D→D to achieve conversion, but channel D served as both auxiliary conversion and final conversion in the conversion path. Therefore, the channel contribution values ​​of A, B, and C are 20 yuan respectively, and the contribution value of channel D is 40 yuan.

Assumption 4

The visitor went through four channels A→B→C→B→D→D to achieve conversion, but channels B and D appeared twice respectively. Therefore, the channel contribution values ​​of A and C were 16.7 yuan respectively, and the channel contribution values ​​of B and D were 33.3 yuan respectively.

In the above assumptions, there is one condition that is worth discussing, that is, the weight of each assisted conversion and each final conversion. In fact, for assisted conversions and final conversions, we cannot judge which channel has a greater impact on users, so we give them the same weight in the above example. However, different companies and different analysts have different perceptions. The weight relationship between assisted conversion and final conversion can be expressed as: each final conversion = η each assisted conversion.

3. How to implement attribution model analysis

This can be achieved through the multi-channel conversion path of GA ( Google Traffic Analysis Tool ), which requires each marketing channel to uniformly mark the traffic introduced to the official website. In addition, you can also use the traffic monitoring tools in the industry. Currently, there are already several agents developing attribution models. I believe that this statistical method will soon become the mainstream in the e-commerce industry.

4. Inspiration from attribution model analysis

1. Based on the auxiliary conversion value and final conversion value of each channel, be familiar with the main role and contribution of each promotion path in the entire company's marketing system.

2. Based on the company's awareness of assisted conversions and final conversions, determine the weight ratio of the two, and then refine the contribution value of each channel to orders to provide a reference for future marketing investment in each channel.

3. Determine the media mix, delivery and advertising schedule for marketing communications based on the media contact choices and sequence of users during the purchase conversion process.

Author: Ai Qijun , authorized to be published by Qinggua Media .

Source: Aichisem