App delivery growth: RTA

RTA (RealTime API), which means "real-time advertising API" in Chinese, is a platform that asks advertisers through the RTA interface whether their ads are participating in the bidding during the ad retrieval stage. After receiving the information, the platform will retrieve and recall the ads.

1. Background

The emergence of any new thing must be accompanied by a certain demand background. Let us first understand the demand background of RTA products. In the above article "Things about Marketing API", we mentioned three ways for advertisers to place advertisements. Except for self-built DSP connected to ADX which relies on the advertiser's own data and technology, the remaining two methods: opening an account in the media backend and connecting to MAPI to build a Trading Desk both rely on the capacity building of the advertising platform. Since advertisers need to have deep technical experience to build their own DSP, they usually choose to open an account in the media backend or connect to MAPI to build a Trading Desk to meet their own App promotion needs.

In the early stages of promotion, advertisers can acquire customers through the targeting capabilities provided by advertising platforms. However, with the disappearance of the Internet dividend, it is becoming increasingly difficult for advertisers to acquire customers and costs are rising. The previous extensive delivery is no longer applicable to the current market, and refined delivery has become an indispensable choice for advertisers. Combining the advertiser's data and the technical capabilities of the advertising platform to assist in delivery is an effective solution.

So, before the ad platform came into being, how did it use third-party data to make up for the ad platform’s lack of targeting capabilities? It's very simple. As mentioned in "Things about Marketing API", a very important function of MAPI is data feedback. Whether it is the DMP API or the user behavior data feedback API, advertisers can use it to return the required targeted data to the DMP platform. The advertising platform then connects its own DMP and DSP and uses the data uploaded by advertisers to carry out targeted delivery. In addition to making up for targeting, advertiser data also has an important application scenario, which is OCPX. This is why advertising platforms recommend that everyone use OCPX, so that it can obtain back-link data, enrich its data features, and assist in optimizing the model.

Since we already have MAPI for data transmission, why do advertising platforms want to launch RTA? The considerations are mainly due to the following reasons:

1. In order to protect user privacy and data security, advertisers no longer export data to external platforms through APIs. User data can only be used on their own platforms.
2. There is a lag in data transmission via MAPI, and the real-time nature of the data cannot be guaranteed; using the RTA interface can make up for the lack of real-time directional capabilities of the platform.

2. RTA Advertising Process

During the ad retrieval stage, the advertising platform asks advertisers through the RTA interface whether their ads are participating in the bidding. After receiving the information, the advertising platform performs ad retrieval and recall. A complete RTA advertising process is shown in the figure below.

1. The front end initiates an advertising request and requests the back end advertising service;
2. Advertising services generally go through a number of processes, including ad retrieval, ad recall, rough ad ranking, and ad fine ranking, before the ads are eventually sent to the front end for display to users.
3. RTA request timing: Why RTA is not inserted into any of the above processes? The reason is that RTA is suitable for any link between ad retrieval and ad delivery. However, the processing logic for ad platforms and advertisers is different in different links. Therefore, when connecting to RTA, advertiser developers need to clarify with the other party when to request RTA.
4. RTA timeout control: RTA has very high requirements for timeliness, so when problems such as timeouts occur, the processing mechanisms of different platforms are different. Some platforms will assume that the advertiser does not need this traffic after a timeout; some platforms will do the opposite and assume that the advertiser needs this traffic. Therefore, when developers connect to RTA, they need to clarify the RTA timeout period and timeout processing logic with the other party;
5. RTA control dimension logic: The control granularity of RTA can be to the advertising account, advertising plan and other dimensions, but the control granularity of different platforms is different. The finer the granularity, the more refined advertisers can be. Currently, most platforms are still at the advertising account dimension. When developers connect with RTA, they need to clarify the control dimensions of their RTA with the other party;
6. RTA bidding: The RTA interface of the current mainstream advertising platforms supports setting bids (or price coefficients). Advertiser developers can estimate the value of this traffic based on their own data and take the initiative in bidding. Some RTA interfaces also support quality scores, which are similar to bids and return the value of the current traffic. When connecting to RTA, developers need to clarify with the other party whether their RTA supports bids or quality scores.
7. DPA advertising: Currently, the RTA of mainstream advertising platforms supports returning product IDs (it is necessary to connect to the product library of the advertising platform in advance), which means that DPA advertising can be placed, which is suitable for remarketing and further refined operations on traffic. When connecting to RTA, developers need to clarify with the other party whether their RTA supports DPA ads;
8. Monitoring link: For advertising platforms that support exposure and click monitoring, some platforms’ monitoring macros support RTA_ID. This can distinguish which ads have won through RTA strategies, and monitor the quality of advertising traffic. For example, if an unknown RTA_ID appears in the exposure and click monitoring data, there may be a problem with the traffic. When connecting to RTA, developers need to clarify with the other party whether their monitoring link supports replacing the RTA_ID macro parameter.

3. RTA Access

RTA, as an API that expands real-time targeting capabilities, can improve the effectiveness of advertising. First of all, advertisers certainly want to access it, but advertising platforms generally set access thresholds for data security and timeliness considerations, which is similar to connecting to ADX.

1. Consumption threshold: Since there is a risk of data output when advertising platforms open the RTA function, advertising platforms generally only develop RTA capabilities for KA customers;
2. Technical threshold: RTA is a real-time interface with high timeliness requirements. Therefore, advertiser developers must have good technical experience to connect to RTA.
3. Data capabilities: RTA needs to combine advertiser data and the technical capabilities of the advertising platform to maximize its effectiveness. Therefore, advertiser developers need to have certain data capabilities and be able to score traffic quality based on first-party data and measure traffic value.

When building an RTA platform, advertisers need to pay attention to the interface timeout problem. Once the timeout ratio is too high, the platform may shut down the RTA capability. Especially when advertisers are connected to the RTA of major platforms, such as Guangdiantong, Juliang or Baidu's traffic, if RTA requests are made to advertisers every time, it will inevitably lead to a large amount of QPS on the advertiser's server, and the pressure may not be able to be sustained.

So, in order to improve QPS, advertiser developers can set up a cache mechanism and give priority to using cached data. At this point, we need to consider the cache expiration issue. In addition to setting the expiration time, we also need to actively delete the cache based on business logic, in scenarios such as when users install and activate the App, or when users purchase advertised products. The advertising platform can also set up a corresponding caching mechanism, that is, to cache or record users who have been "marked" by advertisers, so that the advertisers do not need to initiate requests next time. As RTA data continues to accumulate, advertising platforms will be able to grasp more data characteristics of advertisers' users, thereby reducing the number of RTAs. At the same time, the benefits of the accumulation of data from advertisers to advertising platforms are also obvious, such as being used to train OCPX models.

When building RTA, advertisers should pay attention to the following indicators:

1. Timeout rate: A high timeout rate is a bad experience for both the advertising platform and the advertiser developer. The platform may shut down the RTA capability.
2. RTA effective return ratio: the ratio of effective RTA results. If this ratio is too low, the RTA consumption will be too low. The platform will think that RTA has not brought significant effect improvement to advertisers and may shut down the RTA capability.
3. RTA consumption ratio: the ratio of consumption participating in RTA to the overall consumption of the market, which is similar to the RTA participation rate;
4. RTA matching analysis: Analyze the final RTA matching situation by combining the RTA_ID data in the monitoring link and the RTA valid return data.
5. …

4. RTA Application

RTA provides real-time orientation capabilities, so in which scenarios can it be maximized? The following are several commonly used scenarios in the e-commerce industry. If readers have better scenarios, you are welcome to discuss them together.

1. Attract new customers

When advertisers promote an App, they will definitely involve attracting new users and acquiring new customers. Since the goal is to attract new customers, if you serve ads to old customers, it will waste budget and affect the user experience of old customers. Therefore, old customers need to be excluded when serving ads to attract new customers. Only advertisers know how to judge whether a user is a new or old user of an App, and the judgment criteria of each advertiser are different, so advertising platforms cannot block these users. Through the RTA interface, advertisers can decide for themselves whether to run new ads for this traffic, thereby improving the effectiveness of advertising.

2. DPA

Currently, mainstream advertising platforms all support returning product information in RTA return results, which means that advertisers can combine internal recommendation data to push the most suitable advertising materials for this traffic. Advertiser developers need to support real-time + offline methods to calculate users' personalized data. This requires considering the time consumption issue. After all, real-time recommendations may not meet the timeliness requirements of RTA.

3. Frequency control

In the advertising industry, there is a consensus that targeting the same user multiple times will not improve the effect. Instead, the frequency of users seeing the ads needs to be controlled within a certain range. When advertisers promote apps, they usually use different channels for placement. Before RTA, advertisers could generally only set frequency control settings for a single channel. With RTA, if all delivery channels support RTA, advertisers can achieve global frequency control, save delivery costs, and improve delivery results.

4. Traffic Tiered Operation

In the App operation system, the traffic value of users with different activity levels is different. The higher the activity level, the greater the corresponding traffic value. Therefore, advertisers generally assign different quality scores to users with different levels of activity and set corresponding operating strategies. Since the RTA interface supports returning traffic quality scores, advertisers can use this field to convey their scores for this traffic to the advertising platform. After receiving this information, the platform adds the dynamic bidding capabilities on the platform side to help advertisers better compete for this traffic and realize layered traffic operations.

5. RTA VS RTB

RTB (RealTime Bidding) means "real-time advertising bidding" in Chinese. Advertisers can estimate the traffic value, CTR and CVR for each ADX traffic, and then decide whether to participate in the bidding. If they do, they need to return the corresponding advertising materials and bids. Advertisers connect to ADX through self-built DSPs. Due to the lack of contextual information (context) on the platform media side, user behavior (user) of user media, and generalized data from other advertisers, under the same technical capabilities, advertisers' CTR and CVR estimation models are also relatively limited.

Although RTB and RTA have only one letter difference in their names, they are two completely different tools. The main differences are shown in the figure below.

1. Advertisement ownership: RTA ads are created on the advertising platform side; RTB ads are created on the self-built platform side;
2. Bid: RTA's bid can be the platform's bid * user quality score, or RTA's real-time bid; RTB's bid is real-time bidding;
3. Targeting: RTA’s targeting combines the targeting on the platform side and the advertiser’s real-time targeting; RTB’s targeting is mainly the targeting function of its self-built platform;
4. Targeting granularity: The targeting granularity of RTA can be at the account and plan level; the targeting granularity of RTB is at the ad level;
5. Data: RTA uses both advertising platform data and advertiser data; RTB only uses advertiser data.

VI. Conclusion

RTA is an advertising technology capability that has emerged in the past two years. Its core point is to better assist in real-time advertising through advertiser data combined with the technical capabilities of the advertising platform, including but not limited to targeted optimization, bid optimization, DPA optimization, etc. With the in-depth discussion and research of RTA by various platforms, it is expected that this industry will provide more ways to play.

Author: jdyx728

Official account: Daily notes on commercial products, a commercial product manager who understands products, operations, and code