If you don’t know how to do attribution analysis, it’s no wonder that your channel conversion and user growth are getting worse and worse!

If you don’t know how to do attribution analysis, it’s no wonder that your channel conversion and user growth are getting worse and worse!
The APP has been developed, the advertising has been made, the channels have been selected, and the budget has been in place (although it always feels like it’s not enough), but the results of the advertising are not good enough. At this time, we will infer whether there is something wrong with the optimization of the APP product? The advertising copy isn’t tempting enough? Is there something wrong with your channel strategy? Wait, of course all of the above are true. But no one can answer this question: Where is my 50% advertising budget wasted? Or how can the advertising budget be allocated more reasonably? Advertising is the last and most important link in directly facing users. If you cannot trace (attribute) the effect of the advertising (I am not talking about the gorgeous but fake data that the channels show you in the general sense), then it will be really difficult for you to do a good job in the next step of optimization. The efforts made previously may be lost slowly and unknowingly in the final delivery stage. Anyone who is still doing extensive effect evaluation is too passive. Compared with traditional advertising, the data of Internet online advertising is recorded and can be used for data analysis to optimize the effect of delivery. Online media channels have taken advantage of these advantages and harvested advertisers' budgets under the concept of targeted delivery. Data always makes people inexplicably obsessed. With our inherent cognition and limitations, we see that data is beautiful, but the results are far from satisfactory, especially now that channel conversion is getting worse and worse, and user growth is becoming more and more difficult. After the advertisements are put out, the operation staff also spends a lot of energy on effect evaluation, but it is useless. This extensive evaluation method cannot solve more substantial problems at all:
  • We know that the user clicked on the ad from this channel, but did they install the app after clicking on it?
  • Which channel brought about the download volume this time? Which one should I buy next time?
  • In addition to user traffic , who should be credited for this later conversion? What is more scientific and reasonable?
The biggest difference between Internet marketing and traditional marketing is that it is targetable and traceable. The characteristics of traditional advertising are wide coverage, rough distinction of target groups, and inability to track effects. Internet advertising can be targeted and its effects can be tracked, both of which are equally important. I believe that advertisers pay more attention to targeting research but rarely pay attention to attribution. The targeted delivery brought by big data technology is indeed very attractive. For example, when searching for "travel", the platform will show different products and advertisements to different people. Targeting seems to be very cost-effective for advertisers, but people always have limitations. Without attribution, how do you know whether extensive targeting or more refined targeting will be more effective? Most advertisers will use a multi-channel combination promotion approach when launching APPs. For example, when launching an overseas APP, they will use a combination of Facebook, Google , Apple bidding ads or other media channels to purchase advertisements. For advertisers, I need to clarify which channel the users who receive the ads come from, what the quality of the users is, and what the relationship is between ad purchases and user acquisition as well as user behavior within the app. These require a set of methods or systems to verify, which is attribution. In actual applications, attribution operations are more complicated. For example, if a user views an information flow ad , clicks on a social ad, and then completes a download in a search engine ad, how should this conversion be calculated? After a user downloads an APP through the QQ client, there is no other conversion behavior within the APP. How to judge this channel? For example, if an e-commerce user sees an e-commerce advertisement and generates internal conversion behavior, then to what extent will he repurchase and become active? These can be linked together through attribution to achieve more comprehensive and scientific judgment and tracing, thereby effectively optimizing all aspects of delivery and operation. In other words, through a platform's attribution service, the advertising effectiveness data can be clearly identified. For example, you can tell advertisers what the internal payment rate and payment unit price are for APP users brought by the Google channel. You can make a judgment based on this data and know how to place your ads next time. If the volume on Google is good, you can increase the budget for advertising. In addition, if a certain channel brings a large amount of new traffic but has very poor internal conversion behavior, you can subsequently reduce this part of the promotion budget or even directly pass the channel. Traceability provides a strong basis for scientific optimization. Attribution also involves methodology, and multi-touch attribution is the scientific attribution analysis method. Strictly speaking, there are about 10 attribution models, and attribution classification is roughly divided into single-touch and multi-touch. To make it easier to understand, Sister Xi will introduce four common attribution models to you. 1. Last interaction model: 100% is allocated to the media that the user last touched before conversion. This is also easy to measure, but it is a single-touch model and is imperfect. It is suitable for conversion-oriented advertisers. 2. First interaction model: 100% is allocated to the first contact channel, only considering the initial brand awareness and not conversion, suitable for brand new brands. 3. Time decay interaction model: the ratio decreases over time, which is suitable for temporary promotional advertising. 4. Customized interaction model: Customize the ratio of each stage, suitable for advertisements that attach equal importance to sales and brand. In the actual user conversion path, the user saw and clicked on an Apple mobile phone advertisement on Toutiao , and then clicked on a push advertisement in Moments . After returning, the user searched for new Apple models on Baidu on the PC, and then clicked on a JD.com advertisement on Baidu and completed the purchase on JD.com. This is a relatively common process from advertising display to user conversion. If the final interaction attribution model is used, only the JD.com advertisement that appears on Baidu will be evaluated for its effectiveness, while the advertising behaviors generated by the pre-advertising factors will not be calculated, which means that 100% is given to the final contact channel. If we follow the first interaction model, the effect will be attributed to Toutiao ads. Regardless of whether it is the first interaction or the final interaction, both belong to the single-touch attribution model. This attribution method is rather one-sided in multi-channel combination delivery. The various channels that users come into contact with reflect the exposure rate and may have generated conversions for users. Using single-touch attribution is very unfair to other forms of advertising because it also plays a role in the entire purchasing process of users. A more scientific attribution method needs to consider the user's comprehensive behavior, which is what I call multi-touchpoint attribution. It uses a set of algorithms or models to make allocations. For example, a user saw an ad on Baidu but did not complete the purchase. The next day, he completed the purchase through Toutiao. I will allocate 70% of the effect to Toutiao and 30% to Baidu. Multi-touch attribution will conduct a comprehensive evaluation across devices, screens, and channels to give a more reasonable effect distribution ratio. The time decay interaction and custom interaction models mentioned in the above figure belong to multi-touch attribution. The bidding advertising attribution provided by Apple backend is a rough attribution

It tracks and attributes app downloads from search ads, primarily through the Search Ads Attribution API. The advantage is that with Apple Attribution API information, you can optimize your CPT and CPA targets for different keywords , ad groups, and audiences.

In terms of results, Apple's attribution API is still relatively simple and rough. It can trace back which keyword each download came from, but it cannot provide any further information. Advertisers not only want to know which keyword each download comes from, but may also want to see the number of registrations, the payment ratio, and even more user portrait information, etc. At this time, they need to use more professional and systematic third-party attribution tools . How to obtain more valuable attribution analysis through Apple bidding ? Only more detailed and scientific attribution data can better guide the optimization of Apple's ASM bidding advertising. As an independent third-party attribution tool, Liangjianghu Attribution aggregates external data by monitoring click data and conversion data. In addition, it collects APP internal data through SDK, and then attributes the data step by step through APP data. It ultimately attributes the user to which keyword brought him, whether he has registered, paid, or even has other in-depth behaviors. Attribution actually plays a connecting role, connecting all the data to obtain more accurate analysis data. In addition to embedding the SDK, it is more important to make a customized tracking solution. It is important to know where to place the embedding points. For example, by embedding a point on the APP registration page, you can get the user registration information. By embedding a point at the membership service, you can know which users have purchased related services, such as the type of service, the amount of the service, and the user's account. Embedding points at various key locations can help to fully record every step of the user's behavior and obtain various types of data. Finally, based on the analysis of the collected data, we can clearly identify which word the download came from. For example, we can know that a word brought in 1,000 users, of which 600 were registered users and 200 completed payment, which can then guide the next launch. From the perspective of channel management, it can be used to judge the quality of the channel, rationally optimize the channel delivery strategy, save promotion costs, and improve channel conversion efficiency. From the perspective of user growth, it can be used to judge the quality and effectiveness of a certain word, thereby optimizing keywords and increasing user growth. ASM attribution is only a small part, and refined operation is the goal. The above only takes Apple bidding advertising attribution as an example. In actual operations, attribution can also be combined with big data technology to carry out more practical and valuable refined operations, such as multi-touch attribution analysis for multiple channels, data anti-cheating, and APP user portrait analysis . Of course, these are all for the purpose of more efficiently controlling costs and improving conversions. 1. Monitor CPA conversion data based on ASM delivery, different channels, and different advertising groups to optimize delivery plans. Some advertisers just want to see the user registration rate . We only need to embed points on the registration page. Some advertisers have different KPI requirements. For example, for e-commerce advertisers, I will place advertising points in the shopping cart and successful purchase stages, and then collect relevant data for analysis. In addition, it can also monitor the CPA conversion data of other non-ASM channels, thereby obtaining the full-path data of promotion delivery, providing real-time customizable data reports and channel quality evaluation and analysis, which will help optimize advertising and maximize the effectiveness of APP promotion . 2. Establish APP user portrait analysis for refined user operations. The user data collected through the SDK in the APP is not particularly large, but if there is a big data foundation, you can analyze the user in combination with other peripheral massive data to obtain a more comprehensive APP user portrait analysis. The so-called more comprehensive user portrait means that 60% of the user portrait you obtain is not enough to support user operations , but if there is a lot of data, the accuracy can be increased to 90%. The amount of data will determine the accuracy of subsequent data analysis. User portrait analysis can be used at all levels of user operation optimization. 3. Provide anti-fraud services for bidding advertising to save operating costs. The anti-fraud mentioned here is mainly for ASM bidding advertising, mainly for the situation where keywords in the bidding are maliciously clicked. During the operation, a threshold is set for real-time intelligent processing. For example, within one minute, if the average display volume is 100 and the number of clicks is 50, when the value exceeds several times the previous average, the system automatically pauses. This can effectively prevent ASM vicious competition and channel cheating , effectively block false traffic , and your real traffic will naturally increase. (Anti-cheating on Android is more complicated, and students who are interested can pay attention to the anti-cheating on the Liangjianghu APP) The realization of the above functions requires a large amount of data accumulation, professional data modeling capabilities, distributed real-time data processing capabilities, and the dimensions of post-data analysis, all of which will make the results more accurate. For example, by using a real-time streaming computing platform, there will be no delay or loss in conversion data notification, and real-time processing will have higher accuracy. I won’t go into details about the technical part. Postscript: Without attribution analysis and tracing, you will not know at which stage the advertising expenditure is wasted. In the era of traditional advertising, it was impossible to conduct scientific effect evaluation. Today, with the gradual improvement of big data technology, there is a soil for attribution analysis, and attribution can demonstrate the value of operations. Attribution is an indispensable tool for refined operations. A single attribution may save 30% of costs and increase conversions by 50%. After all, attribution is about opening up a new operational channel and seizing the neglected but valuable traffic and new additions.

Mobile application product promotion service: APP promotion service Qinggua Media advertising

The author of this article @溪姐 is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting! Site Map

<<:  The most comprehensive APP campus channel promotion strategy in history: 15 campus promotion methods, how many do you know?

>>:  Dalian home improvement mini program agent price inquiry, how much is the Dalian home improvement mini program agent price?

Recommend

APP new user promotion丨How to attract traffic to APP through H5 activities?

When operating an APP, it is bound to involve the...

The most practical seed user operation method

I have been wanting to talk about the seed user o...

How to conceive a marketing plan for a new product?

This is a very common scenario. When you take on ...

How to build a community from 0 to 1?

Communities are usually divided into two dimensio...

How should I operate a new Baidu bidding account? What should I pay attention to? Urgent!

(1). Choice of keywords: It depends on whether yo...

APP activity promotion and operation: Create these 3 emotions to warm up the activity

"Wolf Warrior 2" grossed over 1 billion...

How much does it cost to attack a server, and what is needed to attack a 200g server?

How much does it cost to attack a server, and wha...

How to increase traffic, popularity and playback volume of Tik Tok works?

Those who have made or played Douyin know that Do...

How to optimize single-page website SEO? What are the common SEO problems?

Q: What is the specific process of keyword optimi...

5 key points to master user operations in community operations

The definition and goals of user operations in co...

9 points to analyze Uncle Kai’s storytelling content operation strategy

In 2014, Wang Kai, the host of "Fortune Stor...

16 creative ideas and keyword intent analysis for medical bidding promotion!

We know that each keyword has its own part of spe...

TikTok Promotion: The essence of TikTok is revealed in 3,000 words!

In the past two months, readers have been leaving...

What are the benefits of developing mini programs in Guangzhou? Who to find for development?

2018 is the second year of the birth of Mini Prog...

Xinxiang Mini Program Production Company, how much does it cost to make a ticketing mini program?

How much is the quotation for Xinxiang ticket pro...