Intelligent Advertising and Growth Platform: Strategies and Technologies from Tencent

Introduction

With the "second half" of the Internet era, traffic growth has hit a bottleneck and user growth has become a hot topic. Different product forms have varying growth demands, requiring distinct AARRR metrics. A growth middle‑platform abstracts common business needs, builds growth capabilities, and supports agile responses to diverse growth scenarios through material mining, intelligent delivery, and RTA/DPA technologies.

1. Growth and Growth Middle‑Platform Challenges

Multi‑channel traffic: the platform purchases traffic for each app across major providers.

Material specifications: each app has different material requirements, increasing complexity.

Bidding strategies: apps at different lifecycle stages prioritize user volume, cost balance, or ROI for paid users.

User group strategies: targets vary from high‑value paid users to activation, registration, retention, requiring adaptable audience strategies.

2. Background of Growth and Growth Middle‑Platform

Traditional growth buying involves multiple agencies and channel‑specific ROI targets. The middle‑platform enriches product flow, business metrics, and fine‑grained audience mining, enabling programmatic delivery, stage‑specific goals, and personalized material generation.

3. Delivery System and Strategies

3.1 Delivery Formula

The delivery system optimizes ad volume (ad_num) using CTR, CVR, LTV, and CPA. It scales ad volume linearly with competitive advantage, introduces qualification rates, and manages ads through lifecycle stages (learning, stable, decay) with real‑time adjustments.

3.2 Delivery System Overview

The system automates placement, selects materials based on performance data, recommends channels and audiences, proposes optimal bidding, and supports batch copy‑and‑tune for high‑performing creatives across platforms.

3.3 Routine Scheduling

Ads are categorized into learning, stable, and decay phases. The engine monitors conversion data, automatically pulls down under‑performing ads, updates stable ads, and retires decayed ads, triggering new cycles to maintain volume.

3.4 Effect Optimization

During the stable phase, the engine applies automatic price adjustments and budget reallocations based on each ad's acquisition capability, user quality, and market competition to sustain optimal ROI.

3.5 Goal Optimization

Attribution strategies combine oCPX, client‑side, and channel‑side data to split or merge user groups, guiding the model to target desired ROI and user characteristics.

4. RTA Delivery Scenarios

RTA bridges the data gap between media platforms and advertisers, modeling users in real time to decide whether to serve an ad, focusing on uplift rather than pure volume.

4.1 RTA Activation Model

Users are segmented into potential, low‑activity, high‑activity, and daily active groups. The uplift model measures the probability lift from ad exposure versus no exposure, reducing cost by 25% in low‑activity user activation.

5. Material Strategy

Materials are decomposed into element pools with tag systems. A template synthesis engine combines elements with strategy engines to produce final creatives, which are then filtered through selection, review, copywriting, and enhancement stages.

6. Pull‑to‑Push Integration

The approach uses clipboard sharing, device fingerprinting, and dynamic packaging to transfer context from ad to landing page, enabling seamless user experience and improved activation and retention.

7. DPA Delivery Scenarios

DPA leverages user behavior histories to build product catalogs and serve personalized creatives aligned with specific goals (volume, duration, payment, retention), achieving massive user coverage and higher secondary retention.

Conclusion

The growth middle‑platform now connects with over eight major domestic media platforms, delivering programmatic, batch advertising at >30 W QPS and 240 B daily traffic for more than 20 apps, demonstrating the effectiveness of automated delivery, material mining, and intelligent bidding strategies.