Why Advertising Needs Computation: Fundamentals of Digital Ad Auctions and eCPM

The piece begins with a humorous anecdote about a Chinese professor claiming a formula for predicting the effect of Xiangsheng comedy, then pivots to a lecture on why advertising must be computed, highlighting a deep cognitive gap between traditional marketers and internet engineers.

Traditional advertising is portrayed as a creative, high‑budget activity driven by imagination, whereas internet engineers view ads as a high‑tech revenue engine powered by AI, where even mediocre performers are expected to excel.

Digital media introduces the core problem: selecting the optimal personalized ad for each impression to maximize effectiveness. This leads to the concept of eCPM (expected Cost per Mille), the product of click‑through rate (μ) and average click value (v), which becomes the primary metric to compute.

Estimating μ and v is split between media (which observes clicks) and advertisers (who know conversion value). Most ad platforms charge on a Cost‑per‑Click (CPC) basis, reflecting this division.

To encourage truthful reporting of click value, the article discusses auction theory: second‑price (Vickrey) auctions and the more general VCG mechanism ensure bidders reveal true valuations, stabilizing the ad marketplace.

Click‑through rate depends on three factors—user, ad, and context—necessitating audience targeting (user profiling) based on behavioral data, which raises privacy concerns addressed by differential privacy techniques.

The article outlines two product trends: programmatic advertising, where advertisers compute their own eCPM and submit bids (often settled on CPM), and the opposite trend where the media computes everything and settles on CPA or oCPM, exemplified by Facebook’s approach.

Programmatic ads lower entry barriers for small advertisers but favor large data‑rich clients, while closed‑loop models let media leverage their data advantage to estimate conversion values, sometimes using optimized CPM (oCPM) that optimizes for CPA.

Overall, the article argues that the market has shifted to prioritize the number of advertisers over budget size, and understanding the computational foundations of ad selection, auction pricing, and data privacy is essential for anyone working in modern online advertising.