When the mobile Internet enters the stock era, advertising is falling into a “dilemma”: on the one hand, the average annual increase in customer acquisition costs, and on the other hand, the marginal benefits of manual optimization continue to decline – the top advertisers need to operate tens of thousands of plans every day, and the manual adjustment cost of a single plan even exceeds the income it generates.
In this context, the advertising center has evolved from an “efficiency tool” to a “growth engine”, becoming the key to breaking the game for enterprises. Based on practical experience, this paper dismantles the core architecture, key capabilities and landing logic of the advertising middle platform, and provides a reusable construction framework for product managers.
1. What is the advertising center?
The essence of the middle platform is “the precipitation and reuse of public capacity”. The advertising center extracts the capabilities of “account management, material optimization, data monitoring, and strategy adjustment” that are repeated in advertising to form a standardized module to support the large-scale delivery of multiple businesses (such as e-commerce, games, online earning) and multiple channels (such as Huge Engine, Guangdiantong).
Its evolutionary path is clearly visible:
Initial stage: Build the basic closed-loop capability of “delivery-effect” (can run through, can be measured)
The core of the initial stage is “survival” – you must first run through the minimum closed loop of “delivery-data return-attribution-optimization”, otherwise the delivery will not be sustainable (advertisers will not pay for “unknown” delivery;
Medium-term: Improve the efficiency and refinement of “large-scale delivery” (more runs, faster runs)
The core of the medium term is “efficiency” – solving the problems of “many, miscellaneous, and repetitive” on a closed-loop basis to support scale expansion; [For example, when the scale of the business expands (such as expanding from 1 media to 5, and the average daily plan increases from 10 to 1,000), the core pain point changes from “whether it can be invested” to “whether it can be invested efficiently]
Mature period: Build ecological capabilities of “data-driven + intelligent decision-making” (can run well and become a system)
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
The core of the maturity period is “enabling business” – when advertising becomes the core growth means of the enterprise (such as annual consumption of more than 100 million yuan and covering multiple business lines), the middle office needs to evolve from “tools” to “enabling business”: [For example, real-time monitoring of cards to stop low-quality advertisements and materials; Automatically adjust bids and assets; Estimated investment income, etc.]
2. The core structure and capabilities of the advertising center
1. Growth indicators: anchoring the “North Star” of delivery
All delivery actions need to revolve around clear goals, and the goals often come from the key nodes in the AARRR model: Acquisition, Activation, Retention, Revenue, and Referral.
For example:
- E-commerce business may take “paid conversion” as the core indicator, which needs to be dismantled into a funnel formula of “exposure→ click→ add-on purchase → payment”;
- Tool apps may focus on “activation rate” and need to be associated with behavior data of “download completed→ first opening→ core function use”.
After clarifying the indicators, the middle office needs to build “indicators”
–data
– Strategy” linkage mechanism – When the core indicators deviate from expectations, it can automatically locate problem links (such as high clicks but low activations, it may be that the material is not good, the bid is too low, etc.).
2. Asset Library: The “First Line of Defense” for Ad Performance
Creative is the “first impression” of an ad, and its quality directly determines click-through rate (CTR) and conversion efficiency. The material library of the middle office needs to solve three core problems:
- Multi-scenario adaptation: Different channels have very different requirements for materials (such as Douyin prefers vertical videos, Guangdiantong focuses on image combinations), and needs to support classification and storage according to the “business + channel + product” tags, such as “e-commerce-Douyin-clothing” materials, and automatically match the 9:16 size template when creating advertisements;
- Effect attribution: Associate the full-link data (exposure, click, conversion) through the material ID to generate a “material effect ranking” – for example, if you find that the CTR of the video material of “3 seconds ago patch + price anchor” is 30% higher than the average, you can quickly reuse this mode;
- Lifecycle management: Automatically mark “inefficient materials” (such as CTR below 50% of the industry average for 3 consecutive days) or the Huge Engine API can directly return the material label, trigger an early warning and recommend replacement or pause to avoid invalid consumption.
3. Media side: the bridge connecting the “traffic pool”
The core of the advertising center is “unified control across channels”, and the basis for achieving this goal is the Marketing API that connects with mainstream media.
The core channels that need to be covered include:
- Domestic: Huge Engine (Douyin, Toutiao), Guangdiantong (WeChat, QQ), Kuaishou, Huawei/Xiaomi and other app stores;
- Overseas: Google Ads, Meta (Facebook/Instagram), TikTok for Business.
Through API docking, the middle office can achieve “two-way synchronization”: on the one hand, it pulls media data (exposure, consumption, clicks) in real time, and on the other hand, pushes operation instructions (creating plans, adjusting budgets, and pausing delivery), completely bidding farewell to the inefficient mode of “multi-platform switching and manual data recording”
4. Delivery system: the “operating system” of the middle office
The delivery system is the core functional carrier of the middle office, which needs to cover the entire process from “plan creation” to “effect optimization”, and the core includes five major modules:
Account management: the “nerve” of permissions and resources
Based on the company’s organizational structure, a three-level account system of “enterprise-department-project team” is established, combined with the RBAC (role-based access control) model to achieve refined permission control:
- Operation post: The planned bid can be adjusted, but the total budget cannot be modified.
- Finance post: only view consumption data, no operation permission;
- Administrator: Configure the role permission template to support batch authorization.
At the same time, it is necessary to support the centralized management of account assets, including account balances, qualification documents (business licenses, industry licenses), historical delivery records, etc., to avoid compliance risks caused by “scattered accounts and chaotic permissions”.
5. Smart Batch: Ad building accelerator
One of the core pain points of optimizers is “repeated creation plans” – a game company once estimated that it takes 15 minutes to manually create a single plan, while the average daily demand exceeds 1,000, and the pure manual model cannot support it at all.
The intelligent batch function needs to be implemented:
- Policy configuration: Preset templates such as “delivery period, targeted audience, bidding method”, etc., such as “e-commerce promotion template” automatically matches “10-22 o’clock delivery + 25-40 years old female targeting + OCPM bidding”;
- Creative combinations: Automatically generate a combination of “materials + copywriting + landing page”, such as using 3 sets of materials, 2 sets of copywriting, and 1 landing page to generate 6 advertisements in batches;[Advertising allocation rules can be allocated using the formula of remainder + equal ratio series];
- Compliance verification: The pre-created automatic detection of sensitive words in materials and illegal content on the landing page [can be connected to the machine review platform], and the pass rate has increased to more than 95%.
6. Monitoring strategy: the “safety valve” of cost control
The core risk of advertising is “ineffective consumption” – an education customer once wasted more than 100,000 yuan of budget in a single day because he did not discover the “high click and low conversion” plan in time.
The monitoring strategy needs to achieve “automatic monitoring + intelligent adjustment + timely warning”:
- Monitoring dimensions: Coverage cost (CPA/CPM), conversion (click→ consulting → payment), and ROI (cycle ROI, LTV) three core indicators;
- Adjust the logic: Automatic operation based on “data threshold + condition combination”, for example, “When the CPA of a plan exceeds the target value by 120% for 2 consecutive hours, and the click conversion rate is lower than the average value of 50%, it does not meet the standard after being triggered n n times in a row, it will automatically suspend and notify the optimizer”;
- Notification mechanism: Push early warnings through Feishu and SMS, including “abnormal indicators, scope of impact, and recommended actions”.
7. Ad attribution: the “compass” of performance traceability
The core of attribution is to answer “which channel and which material brought conversions” to avoid resource misallocation caused by “data distortion”. There are two common attribution methods:
- Passive attribution: Track the link through the “channel-exclusive link”, for example, users click on Douyin ads→ jump to the landing page with parameters→ complete the conversion, and the link is clear and traceable;
- Matched attribution: For scenarios that cannot be directly tracked (such as users searching the next day after viewing an advertisement first), the attribution coverage rate is increased to more than 90% by matching conversion behaviors such as device IMEI, IP address, and user ID.
- Media Attribution:For example, in a huge engine, the media side completes the attribution, and the advertiser receives the attribution results [This should be the mainstream in the future].
8. Data Reports: “Dashboards” for Business Decisions
At the heart of the report is “on-demand output”, which meets the analytical needs of different roles:
- Material team: You need the “Material Type-Size-Conversion” report, for example, you will find that “the conversion cost of vertical videos is 20% lower than that of horizontal versions”;
- Channel team: Need the “Media-Plan-ROI” report, for example, compare “The ROI of the Huge Engine (1.8) is higher than that of Guangdiantong (1.5)”;
- Management: You need an overview of “Total Budget-Consumption-Target Achievement Rate” to quickly judge the delivery progress.
The report needs to support “real-time update + drill-down analysis”, for example, click “ROI anomaly in a certain channel” to directly drill down to specific plans, materials, and groups to locate the root cause of the problem
9. Testing Methodology: Reduce trial and error costs with ABtest
The essence of ad placement is “trial and error”, and ABtest is the core tool to improve trial and error efficiency. The middle platform needs to have a built-in ABtest module:
- Support variable testing such as “materials, targeting, and landing pages”, such as launching 3 sets of materials at the same time, automatically allocating traffic and counting conversion data;
- Provide a “significance test” function to quickly determine “which set of schemes is better” – for example, with 95% confidence, the CPA of scheme A is 15% lower than that of scheme B, and scheme A can be fully reused.
10. Data Layer: The “Fuel Depot” of the Middle Office
Data is the core driving force of the middle office, and it is necessary to integrate three types of data to form a closed loop:
- User base data: Account ID, device information, registration time, etc.;
- User behavior data: Click, browse, add-on, payment and other behavior trajectories;
- Business data: Order amount, membership level, LTV, etc.
Break down data silos through unified IDs (such as device fingerprint + mobile number hashing) to support attribution analysis, user stratification, and strategy optimization.
3. Current challenges and optimization directions
- Algorithm capabilities need to be strengthenedIn the future, it is necessary to label materials through deep learning (such as “scene: office” and “emotion: anxiety”) to automatically generate high-conversion materials;
- AI-Experience-driven to data-driven: Automatically generate executable strategies through AI data mining, and accurately iterate and update materials.
- Insufficient attribution success rate: At present, due to the protection of user privacy, advertisers cannot smoothly obtain user device information, resulting in attribution failure.
The value of the middle office is not in “multiple functions”, but in “whether to solve business pain points” – from tools to systems to middle office, the essence is to replace empiricism with systematic thinking, so that advertising from “feeling” to “data”, from “individual ability” to “organizational ability”
