How to attract new users to Meituan, Ele.me and other apps

First of all, the media has a lot of traffic (on-site and off-site), which needs to be integrated and put on the booth to sell to advertisers. So we created an ADX platform to manage the traffic and decide to whom the traffic should be allocated.

At the same time, it is impossible for the media to look for advertisers one by one, as the efficiency is too low, so a public platform was created to allow everyone to place advertisements themselves. This platform is the DSP platform. Customers come to the DSP platform to pay for advertising, and can set targeting, budget, bids, etc. to optimize the results.

There must be a database responsible for docking. On the one hand, it tells ADX the bidding results to complete traffic allocation, and on the other hand, it stores the budget bid targeting data set by the customer.

Everyone knows that the media supports uploading crowd packages. Because the algorithm is a learning process, it is definitely more efficient if you prepare a set of "teaching materials" for him in advance rather than letting him learn randomly from scratch. What we use most often is probably uploading packages of converted people for exclusion or expansion.

However, it is also quite troublesome to package the data and upload it, and big customers have a lot of new data every day, so this step has to be repeated, which is too troublesome. At the same time, in some industries, the data is sensitive to customers and it is not convenient to package and upload it. Then RTA is needed!

RTA is a set of interface services provided by the media to advertisers to screen traffic based on the advertisers' data. Simply put, the media sends user information to advertisers: Do you want this user? The advertiser answers whether or not to bid, and the media then decides whether to participate in the bidding.

It should be noted that RTA can only decide not to participate in the bidding, but cannot decide whether the bidding is successful or not.

If the advertiser says: I don’t want this user, then the media will definitely give up the bidding;

The advertiser says: I want this user. The media will not participate in the bidding based on the platform settings and data algorithms (for example, your bid is too low).

Therefore, RTA was first used in the scenario of attracting new users: I already have this user, there is no need to push him to me.

RTB (full name real time bidding) is a word from another dimension, which stands for real-time bidding.

The most mature use of RTB is to do secondary marketing for old users, which is to attract more customers. Customers can calculate click-through rate and cost per click based on their own data capabilities. After calculating eCPM, they can directly give the CPM bid to the advertising platform to make a more reasonable user value estimate. As you can tell from the name, this requires advertisers to have very high data modeling capabilities and data volume.

At that time, what everyone compared was the installed base and daily active users. But after the wild growth, problems were suddenly exposed: the growth was approaching a bottleneck, it was becoming increasingly difficult to attract new users, and a large number of silent users who downloaded the app but never opened it accumulated.

Then the core issue of growth changed from “attracting new customers” to “attracting active customers”.

Taking the online life service industry as an example, under the epidemic, users have developed a heavy dependence on life service platforms, the user service consumption demand has surged, single service apps have gradually become "platformized", and various growth channels have emerged, which has also provided more business growth opportunities for life service platforms such as Meituan and Ele.me. As a result, giants from all walks of life have entered the life service sector, directly pushing the competition to a white-hot level.

However, as the Internet dividend period has passed, it is becoming increasingly difficult to attract new customers!

Since 2021, the scale of China's mobile Internet users has remained at nearly 1.16 billion, with the year-on-year growth rate showing negative growth for the first time in May (Source: QM). Traffic has reached its peak and is becoming increasingly expensive. What can we do?

On the one hand, there is definitely room for growth in attracting new customers, so industry players are increasingly focusing on new ways to attract new customers and identifying high-value potential groups by combining their own data with data from advertising platforms.

On the other hand, we need to continue to do more to attract customers: dig deep into user value based on different user consumption tendencies and preferences, which is also called layered activation!

This is when the role of RTA comes into play.

You may still have questions: Doesn’t RTA also provide some user information to the media? How to solve data security issues?

Since it affects deep conversion, the return of some deep data is inevitable. RTA can minimize the data you want to expose.

For example, if I want to run ads for my 1 million users, the traditional way is to package the data of these 1 million users and upload them to DMP for fixed-time advertising, which is the method with the worst confidentiality. But after I used RTA, the media sent me 1 million user IDs, and I picked out 10,000 of them to invest in. Then the media would know at most that at least one of these 10,000 IDs was my user. This approach also includes estimating user value.

Next, I will take Tencent Advertising as an example to explain how it addresses data security concerns throughout the entire advertising process.

Tencent Advertising adopts a two-pronged approach to create a dual-circulation reservoir strategy of "attracting new customers + attracting active customers" - while the "water accelerator" accurately brings in new customers, the "high-intensity oxygen generator" performs refined stratified operations on old customers - thus achieving a dual circulation of external water diversion + internal awakening, introducing new sources and injecting new vitality, achieving double growth in users and revenue.

The first thing is to attract new customers. Why do some advertisers still stick to the crude way of attracting new customers?

1. One party has a shallow understanding of data usage and no awareness

This brings us back to the beginning: many people's awareness of their own data is still limited to excluding and expanding uploaded packages. Why do others have lower costs and better conversions than you? It all depends on the use of your own data!

2. The data is multi-dimensional and complex, and there is no hands-on usage guide

I really don't know how to use it. There is so much data, and it is difficult to know which ones are useful and how to connect them.

3. High data security considerations

For some customers, user data is highly sensitive, so even though they know that the conversion effect will be better, they dare not upload it to the media.

4. The efficiency of doing this is not as expected

How useful is it? Do I need to upload the core data?

The above four points basically cover the voices of most advertisers. So can these be solved? Can!

The Joint Zone is the solution launched by Tencent Advertising: Tencent Advertising and advertisers jointly build the "Joint Zone", combine the data of both parties, model high-value users, learn user characteristics, and increase the ability to obtain volume from high-weight groups while deeply optimizing LTV.

Tencent provides high-level access channels, independent storage mechanisms, and autonomous deep data applications. From data access, data storage, to data application, it provides comprehensive data security protection to ensure data security. Advertisers can compile data in real time in a safe environment under their control, and implement the screening function of non-target populations without using core populations and ensuring data confidentiality.

To put it simply, this is a tool platform that is safe, controllable, highly autonomous, and extremely flexible.

Advertisers can upload historical data here, use their own population as positive samples for modeling, conduct manual screening based on their understanding of the user group, score the samples for relevance, and intelligently optimize the delivery model based on the scores. This allows advertisers to easily find people who are more likely to bring high conversions, complete the exploration of potential customer groups and the recruitment of new customers, thereby achieving the goal of reducing costs and increasing efficiency and outputting more ROI.

What is the effect? A set of key data was disclosed: the proportion of high-value people increased from 10% to 25%; CPM increased by 14.78%, and the conversion cost of shallow targets decreased by 9.31%; after strategy binding, the App retention rate of the modeled population increased by 23%, and the account LTV increased by 5 times.

What is the concept? Taking our common bidding side as an example, if you want to reduce the conversion cost by 10% to lower the bid, you will face a great risk of losing volume, and the CPM will most likely drop. It will be even more difficult to increase the second retention rate and LTV on the back end. Now that the gap between the past and the present is so obvious, what we need to think about is no longer whether to use it or not, but how to use it well.

Looking at the pull job, the biggest problem is still:

1. User stratification is not possible

2. How to use RTA after layering?

3. Unclear about the effect

Let’s first understand the principle of tiered activation: Tencent Advertising sets a paid target for CPA, uses RTA to weight users in real time, stratifies the activation population based on customer assessment goals and key user behaviors, and sets a stepped weighting/bidding strategy for different tiered activation populations. While optimizing account payment costs, it increases the volume-obtaining capacity of high-weight groups, driving an increase in overall account consumption.

Simply put, the media interface is open to you, and you give a rough assessment of the activation value of different types of inactive users: if this user has never opened the APP, you can offer a higher price; if this user has opened it but has not used it much recently, you can offer a lower price; then the media platform makes decisions based on the information you provide. This is the basic logic.

A set of data is also disclosed here: the CVR of high-weight people is twice the average CVR of the account; the weighting coefficient of high-weight people is x2. Since the quality of the population is higher than the average level of the account, the payment cost is still lower than the average level of the account; within the same account, compared with the strategy group and the blank control group, eCPM increased by 10%-15%, CVR increased by 15%-35%, and payment cost decreased by 5%-15%.

The effect is here. If you don’t use it, others will, but their cost is lower and the quantity is larger than yours.

Is it difficult to operate the new business? In fact, it is very simple. This is a win-win situation for the media and advertisers. The media will certainly spare no effort to optimize the operating threshold, and simplify the advertisers' delivery operations through various visual operation guides, efficient manual customer service, sales support and other means to meet the different needs of attracting new customers and increasing activity.