You know that half of your advertising budget is wasted, but you don’t know where it is wasted; you have conducted a brand promotion campaign, but you don’t know how to measure the impact of the campaign on the company’s brand; you have chosen a marketing matrix suitable for the company and carried out marketing activities in a multi-pronged manner, but you have no way to distinguish the effects of different channels. Maybe this article can give you some inspiration.

In marketing theory, there are 4A, 4R or AIDA models, which correspond to the four stages of users’ attention, awareness, attitude and action from the time they are exposed to marketing information. From the perspective of advertising , brand advertising and performance advertising target different stages of consumer consumption behavior. Data shows that in performance advertising, companies with high brand awareness achieve higher sales conversion rates than other similar companies.

On the other hand, successful companies nowadays generally use a combination of brand marketing and performance marketing:

Use brand marketing to build consumers' early awareness of products, and use performance marketing to encourage consumers to ultimately purchase products. This process is repeated over and over again, pushing the company's brand and sales to new heights.

As the saying goes, what goes around comes around. Compared with performance marketing, brand marketing has always been ambiguous. This article proposes four data models to try to solve the ambiguity problem of brand marketing.

1. Model 1: Assist model

On the court, if an assist player passes the ball to the main player to score a goal, it does not mean that the assist player has no value. The value of an assisting player lies in helping other players score goals through teamwork.

Brand marketing is like an assisting player. Only after a lot of preparation in the early stage will the final user place an order. Even if the order is placed through other channels, it is also an assist contribution of this brand marketing.

The calculation method of assist contribution is not difficult to understand :

If the users covered by a marketing campaign of channel A complete the conversion target in channel B within a period of time, then the successfully converted users are the assist contribution of channel A to channel B. The conversion goal can be any user behavior, such as registration, ordering, posting, completing the novice tutorial, etc., and can be set according to the needs of the product.

Because the impact of each marketing campaign on users decays over time, a time window needs to be manually specified in this calculation (generally no more than 1 month). It can be considered that only users who complete the target conversion within the time window can be considered as real assists in the brand marketing activities.

Finally, I want to say a little more:

• Even for users who have completed the same conversion goal, their commercial value is not exactly the same.
• Taking registration as an example, there is a world of difference between Papi Jiang joining and some passerby registering for the live broadcast platform ;
• Taking consumption as an example, the system should not treat the rich who buy light luxury goods and those who take advantage of the bargains equally. Different users have different values ​​to the system. Users with higher values ​​are given higher weights, and this weight is added to the calculation of assist contribution.

The above model assumes that each converted user is influenced by only a single assist channel. In reality, companies will conduct matrix marketing on multiple channels at the same time. A user may receive information from multiple assisting channels at the same time and ultimately complete the order. We need to do multi-channel decentralization and distribute assist contributions fairly to different assist channels. Re-evaluate the contribution of each brand marketing activity.

I don't recommend taking the approach of simply splitting assist contributions equally across all assist channels. A simple text link on a portal website and an experiential marketing campaign with deep interaction in an offline experience store leave different impressions on users and should not receive the same value; the credibility of a report on 36kr and the company’s introduction on the official WeChat account of a startup company is different. The former is equivalent to obtaining media endorsement…

Factors that may need to be considered to achieve fair distribution:

1. Use the popularity of the channel as the weight, giving greater weight to channels with greater influence.

I would like to add one more thing here: the more popular the channel is, the greater its influence is not necessarily the case. If you are in a niche industry, the users of professional media in your niche may be more precise and have a greater influence. It is like the meaning of Non-Existence Daily to science fiction users and 36kr to Internet practitioners.

2. Use user engagement in marketing activities as weight.

Every marketing campaign can more or less obtain users’ browsing time, number of views, number of swipes, conversions, likes and other behaviors. We assume that the deeper the user engagement, the more effective the marketing campaign will be in attracting user attention and building awareness. Use user engagement as a weight for marketing channels.

3. The time period between the user's participation in the marketing activity and the user's behavior is used as the weight.

Here are some algorithms that you can configure yourself:

• They believe that the first and last marketing efforts have the greatest impact;
• Believe that the last marketing campaign has the greatest impact;
• It is believed that the first marketing has the greatest impact.

2. Model 2: Engagement Model

User engagement refers to the quantifiable behavioral indicators of users during their participation in an activity, such as user browsing time, the number of times users like, forward, comment, etc. Generally speaking, the deeper the user's participation, the more eye-catching the event is, and the more interested the users are in your event.

The development of technology has enabled us to measure more and more channels. For example, offline advertising can use Wi-Fi or Bluetooth probe technology to understand the audience’s dwell time. An ordinary online media placement can obtain information about user reading time, number and content of user comments, number of user screen flips, etc.

3. Model 3: Propagation Model

In this era where everyone is a self-media , some official marketing activities not only create a strong psychological cognition among the target users, but also evolve into hot topics and spread in different ways.

I recommend focusing on public opinion guidance through channels such as news, Weibo, WeChat, Zhihu, and vertical media and vertical forums in your specific industry. Understand the number of media outlets, articles, user comments, readings, reposts, etc. participating in the topic.

In addition, key nodes also need to be focused on in the propagation model. The attitude of big Vs may influence the future direction of public opinion.

On Weibo, pay attention to the reposts and direct topics posted by big V users. On Zhihu, pay attention to the likes, comments, topic follows, and answers of big V users. Because these behaviors will appear in the information flow and become new trigger points.

Converting it into a data model, we first need to establish account level standards.

For example, how many fans on Weibo is A-level, how many likes on Zhihu is B-level, and so on; then collect the user data participating in the above behaviors, and analyze the level distribution of the account, the circle (industry) the user belongs to, etc.

For information sent by high-influence accounts, sentiment analysis is needed to understand the proportion of positive and negative information. If necessary, it is even necessary to manually analyze the information flow of big V users one by one and then make corresponding processing afterwards.

TIPS: The above communication analysis can be completed with the help of some third-party tools , such as Fireline Public Opinion, Weibo Hot List, Baidu Index Public Opinion Insight (yes, not the overall trend of Baidu Index), and Qingbo Index. If the company has the resources, it can also obtain the original data for analysis. Qingbo, Weibo and Zhihu provide APIs, which contain data from Weibo, WeChat and Zhihu. You can also use crawlers to obtain raw data from more platforms.

4. Model 4: Sentiment Analysis Model

Sometimes, it is not the case that the more users participate and the deeper their participation, the better. We also need to understand the user's psychological feelings and determine whether the activity guides the user's cognition as expected. Here Romanro recommends using the method of sentiment analysis.

For example, Alipay ’s Children’s Day marketing attracted great attention from users, media and Internet practitioners. If user engagement were the only metric, the event was undoubtedly a huge success. However, people's attitudes towards this marketing are also polarized. If combined with sentiment analysis, it is hard to say whether this event is successful or not.

We can obtain data such as user comments, user facial images, or user voice to understand whether the user is satisfied and whether the evaluation is positive or negative.

If it is a viral marketing campaign , it is necessary to obtain media articles and user comments related to the topic, and perform semantic analysis on both the articles and user comments.

TIPS: The development of technology makes our analysis more and more intelligent. Many scholars and research institutions have proposed different text sentiment analysis models, speech emotion recognition algorithms, and facial emotion recognition tools, many of which have been open-sourced and shared. If you can collect data on comment text, user voice, and user facial images, you can consider analyzing user attitudes through open source or self-developed methods.

If you don't want to take up R&D resources, I recommend the universal manual analysis method, analyzing the above data in full or by sampling.

5. Summary

1. Brand marketing focuses on establishing early user awareness rather than sales, but it can bring higher conversion rates for subsequent performance marketing. The assist model can be used to measure the effectiveness of brand marketing.

2. Use engagement models to assess user interest in the product; use communication models to assess public or media interest in the product; use sentiment analysis models to understand the attitudes of all event participants towards the event to prevent the event from producing counterproductive effects.

Specific indicators might be:

• The number of users covered by the own channels, user portraits, number of user comments/likes/reposts, and the positive and negative semantics of the comments;
• The number of news, readings, likes, comments, and the positive and negative semantics of articles and comments in the media;
• The number of people who create topics on Weibo and their profile, the direct post/forward ratio, the number of readers, the total number of comments, the profile of fans who participate in topic interactions, and the positive and negative semantics of topics/comments;
• The number of articles, readings, reposts, and positive and negative semantics of WeChat public accounts ;
• The number of Zhihu questions, answers, followers, likes, comments, and the positive and negative semantics of topics/answers/comments;
• The number of articles, readings, comments, and positive and negative semantics of articles/comments in vertical forums.

Mobile application product promotion service: APP promotion service Qinggua Media information flow

The author of this article is @罗曼罗 Compiled and published by (APP Top Promotion), please indicate the author information and source when reprinting!