A Design Journey at the Scale of Ad Tech & Big Data @ Tencent

2023｜Type: Original｜Tag: Experience

Since the second half of 2020, I have been working for Tencent's Advertising Marketing Services as a product designer for the Ad Data team. During these years, I have dived into Ad Data Marketing, Ad Data Governance, and Ad Data Ingestion under the goals of Design for Efficiency, Design for Effect, and Design for Service Improvement.

Ad Data Marketing

Performance-based Advertising

# Improve the touchpoints #

In 2020, in advertising, we found that some advertisements have successfully passed the learning phase but have low cost and low competitiveness, which may have the opportunity to get good impressions by raising the bids. Low competitiveness may be the reason for bids and prediction, low cost may require advertisers to diagnose which elements need to be improved, and this is not a one-shot issue, but advertisers can gradually adjust bids to improve it. Helping advertisers quickly select, optimising bids, and enhancing the credibility of platform recommendations is our key. The design of Potential Ads is one of the advertising strategies to optimise bids to improve impressions.

For Potential Ads, we have iterated multiple versions in a month and a half for the MVP, constantly refining the touchpoints for the Potential Ads to achieve faster and more accurate targeting of Potential Ads.

View the project Ad Delivery Strategy: The Potential Ads

# Into the heart of the system #

In 2021, also effective on bids, the ad delivery strategy is to constantly optimise bids on ads, another strategy is to optimise bids for the audience targeted, differentiating bids for the audience targeted in three ways (labelled audience, audience file, and machine learning samples). This is an ad data strategy. By configuring the strategy and the scope of the application, it can be directly implemented in the calculation of eCPM.

Precisely because it is a data product, involving machine learning positive and negative samples, computational resources, and others that big data has, it is important to go deep into the framework to understand how the data is applied and how the data strategy takes effect in the advertising system. Taking machine learning samples as an example, both whether the volume of data and fields in the table uploaded by advertisers meet the learning requirements and help advertisers to create more efficiently and more friendly are the experience issues that we are concerned about.

View the project Ad Data Strategy: ROAS Lift-based Strategy

# Critical thinking on Strategy Product Design #

There can be several reasons for improved performance in advertising, and there is no way to directly attribute performance to design behaviours. On this basis, design deliveries are more about optimisation at the experience level. A deep understanding of the advertising system is indispensable to identify problem areas. This is especially important in the design of Ads Strategy Products. The strategy products that have been involved so far include four in two categories:

Ads Delivery Strategy: One-Click-Started-Delivering, Potential Ads.
Ads Data Strategy: ROAS Lift-based Strategy, RTA Strategy.

Based on these experiences, some thoughts are recorded:

The article A Brief into Design Thinking on Advertising Strategic Products, talks about the life cycle, design process, design patterns, and design opportunities of ad strategy products.
In the article Lite Insights on Comparison, we analysed the scenarios in which it needs to compare with itself, its family, the industry in which it is located, and the market. The comparison is again at what part of the life cycle, what factors are included in the comparison, how the comparison is meaningful, etc.
In the article Lite Insights on Strategy, we analysed touchpoints, design patterns, confidence disclosure, and performance disclosure.

View the article A Brief into Design Thinking on Advertising Strategic Products

View the article Lite Insights on Comparison

View the article Lite Insights on Tactics

Ad Data Marketing

Brand Advertising

# A compromise between customisation and performance #

In the same year, we also provided annual protocol products on Zhishu, such as 3C industry Kanban and automobile industry Kanban. The solution for annual frame products was elaborated on in the article "Annual Protocol Product Design of Data Management Platform."

From 2023 to 2024, building on the foundations of Zhishu, we reaffirmed the value of brand advertising and introduced the concept of scientific measurement. We also developed a new data marketing platform called Ruyi. In this project, my role was not just limited to design; it also involved fostering collaboration within the latest R&D team and guiding our newly joined designers. Additionally, I took on the responsibility of the Market Review.

View the project Annual Protocol Product Design of Data Management Platform

View the project 5R Scientific Measurement for Brand Advertising

Ad Data Management

# Integration and updates within the life circle #

In 2021, the original DMP was integrated and upgraded to Zhishu. The DMP created by multiple teams lacked a comprehensive perspective. This gives designers a very good breakthrough point to disclose the value of the design and grasp the problems that need to be solved as a whole. We can learn how to prompt the integration in the article Integrated Design of Data Management Platform. Compared with the product managers who are experts in their fields, when it comes to the comprehensive view of the platform, designers often become the core among the product managers. Take the signing of the Data Protection Agreement as an example, which originally involved only data assets. Once the user resigns or fails to sign on, the platform will be affected. Taking this as a breakthrough point, we talked about the thoughts small leak will sink a great ship in this article A Brief into Design Thinking on Signing Data Processing Agreement.

In 2024, Zhishu once again integrated and updated to Ruyi. At this point, Ad Data Marketing has gone through a life cycle.

View the project Integrated Design of Data Management Platform

View the article A Brief into Design Thinking on Signing Data Processing Agreement

Ad Data Governance

# Together we face the unknown #

In 2022, we teamed up with experts from various industries to design a suite of solutions for first-party data governance and measurement. The first-party data is highly variable, making it difficult to abstract a framework at the start. Decentralised collection and governance layouts out the first-party data of advertisers in a short period. What data do advertisers care about, how does the industry define the playbook? With this data, combined with Tencent's second-party data, it is expected to provide advertisers with customised services to optimise the phases of the retrieval, scoring, ranking, bids, and so on, to achieve deep optimisation goals and improve ROI.

View the project Data Governance of First-Party Data of Advertisers

# Reducing the information gap between organisations #

In 2022, in addition to first-party data governance for advertisers, we also did second-party data governance for advertising platforms. On the premise of an overall downturn in the market, improving the internal environment of companies has led to the direction of organisational optimisation. Seeing the internal data has become the breakthrough point. It also helped the team to see what each organisation is responsible for, the production process, and the effect of it. The data moved from the data resource side to the data processing side, to the data application side, looking through the data flow to see how the data was being applied in each session. How features are produced, which features are used in models, and at which nodes the strategy works.

We designed a set of criteria and mechanisms from metadata, features, data, models, and strategies, that help the enterprise know its business, reduce the information gap across organisations, and improve the efficiency of delivery.

View the project Data Governance of Second-Party Data of Ad Publishers and Ad Platforms

Ad Data Ingestion

# Consensus is a revolution #

In 2023, rethinking data ingestion after ad data marketing and ad data governance. we try to explore and build a new full-domain data ingestion and application distribution experience. It meets the requirements of deep application, more accurate and real-time data in ads, and also meets the requirements of rich applications and multi-purpose applications in retail. Taking ads and retail data as a bridge, it creates an all-domain marketing data assistant that connects public and private traffic for clients.

View the project Data Ingestion of Advertiser for Tencent Ads & Retail

Individuals and Teams

At the design group of Tencent's Advertising Marketing Services, design is the responsibility. Not only to maintain refined professionalism but also to develop the team.

# Stay professional #

Keep learning in the field of ads and data. Successively make speeches within the team:

A Brief into Design Thinking on Advertising Strategic Products, states the lifecycle of advertising strategic products, the design process, design patterns, and design opportunities. 2021

Data Product Design: From Understanding to Application, outlining the fundamental theories of data warehousing and data streaming, the core framework of machine learning and data analytics, data product form, data design elements, and visualisation specifications. 2022

Blueprint for Data Ingestion, which looks at the pains and opportunities for Tencent Ads in data ingestion from the ads and retail. 2022

The Design of Distribution and Authorisation from Grammatical Structure and Information Architecture, analysing the design thinking on the distribution and authorisation from linguistics. 2022

Competitive Analysis of User Experience in Ad Data Products studied the domestic and international ad data products in the structure of the Five Elements of User Experience. 2023

Industry Research in Visualisation, studied the visualisation definitions, outputs from academia (IEEEVIS/North America/China) and industry (North America/China), and research-based design insights. 2023

Usage Research in Data Visualization, summarising chart usage, colour usage, axes and markers usage, and graphical semantics usage respectively. 2023

The Large Language Model: from Understanding to Application, an overview of the history of the Large Language Model, GPT, and potential applications. 2023

# Build team #

In the three years of Ad Data Design, we have created four platforms together, from one to four team members, and from half to four platforms. Witnessed the development and change of Tencent Ad Data Team. Each stage has its mission.

Win the Team. Deepen the business and consolidate trust. Be the designer who knows the most about ad data, and build personal influence through projects. 2021

Stabilise the Team. Build small teams, establish internal and external collaboration mechanisms, create team documentation, participate in discussions about business goals, set and communicate design goals, point out design challenges for each quarter, and find a breakthrough point for each individual. 2022

Develop the Team. Develop gradient teams and clarify competency models needed for the developing team. Let go of specific design tasks and explore new propositions, balancing personal growth with team growth. 2023