Moloco CARA

领域详情

每个领域旨在解决特定的挑战，并协同工作以实现高收益的投放效果。

广告主可设定其业务目标，例如最大化广告支出回报率 (ROAS)，或推动注册、充值、购买等应用内事件。此外，CARA 智能体还会评估实时广告活动，挖掘优化表现的机会，并将所得洞察反馈给广告主。

CARA accesses ad inventory that can reach over 2 billion daily active users across more than 2.8 million independent apps globally, through direct publisher integrations via our SDK and more than 35 exchange connections.
‍
Our SDK lets us control creative rendering on devices, which unlocks richer ad formats and better signals that sharpen performance across all domains. As we extend our SDK footprint, each new integration strengthens the performance of the entire system because it gives our models richer signal about how users interact with ads. ⓘ

对于每一次展示，CARA 都会根据用户采取特定操作的可能性，并结合该操作对广告主的价值，来评估此次机会的价值。为了实现这一点，我们的模型利用深度学习同时从多个来源获取信号，从而对用户行为、发布商应用、产品、创意格式、内容以及它们之间的相互作用形成关键洞察。

Ad opportunities evaluated by our system exceed one trillion per day, and the AI system makes predictions on each one, only bidding when the unit economics work for the advertiser. Most auctions in the app ecosystem operate as first-price, where the highest bidder wins and pays exactly what is bid, which makes precise pricing critical. CARA simultaneously optimizes bid strategies under competitive auction constraints, allocating budgets across time and ad inventory, and selecting the best supply path. Getting this right requires deep knowledge of auction dynamics, supply path economics, and budget pacing under real-time competitive pressure. ⓘ

每位用户看到的创意都是专门为其选择的：创意是基于每次展示的决策，而非广告系列层面的评估。对于 Moloco SDK，CARA 模型会调整各个广告组件，由 SDK 在展示时将其组装成完整的广告单元。随后，CARA 的选择模型会从庞大的候选空间中，识别出最有可能在该用户和该环境下产生效果的组合。