The Ad Stack Is Becoming Infrastructure

Ad technology used to be described as a stack because it was assembled in layers, usually one procurement decision at a time. An ad server, then header bidding, then a few more demand partners, then a data platform, then measurement, then whatever the next channel required.

Each decision was defensible on its own. Collectively they produced something closer to infrastructure: a system where configuration in one place changes revenue somewhere else, and where no single team can explain the whole path from impression to payment.

What the modern stack actually contains

• Ad serving — Google Ad Manager or equivalent, carrying the pricing and prioritisation logic that decides what wins.
• Bidding — Prebid, server-side adapters, SSPs and the timeout and floor decisions that shape competition.
• Demand — DSPs, curated marketplaces, direct deals and programmatic guaranteed arrangements.
• Data and identity — first-party data, consent, and the addressability signals buyers use to value inventory.
• Supply chain declarations — ads.txt and sellers.json, which determine whether inventory is treated as legitimate.
• Measurement — verification, attribution and reporting credible enough to defend a renewal.
• New channels — CTV, DOOH and pDOOH, retail media, mobile and gaming, each with different mechanics.
• AI-assisted decisioning — increasingly present in bidding, pricing, packaging and optimisation.

Why adding technology stops working

When something underperforms, the fastest available action is usually to add a partner, a platform or a layer. That produces three predictable outcomes.

Duplication. The same demand arrives through several paths, each taking a fee, and the auction becomes a competition between routes to the same buyer rather than between buyers.

Opacity. Every added layer makes cause and effect harder to trace, so decisions are made on dashboards rather than on architecture.

Fragility. Changes stop being improvements and start being risks, because nobody can predict the downstream consequence.

The underlying issue is rarely the individual technology. It is the absence of a designed relationship between the components — the same pattern described in Revenue Architecture.

Thinking about it as infrastructure

Infrastructure is judged on different questions than tools:

• What is the intended path from impression to revenue, stated explicitly?
• Which components are load-bearing, and which are historical?
• Where does value leave the system, and to whom?
• Which decisions are made by configuration rather than by strategy?
• What would break if a major partner, signal or channel changed?
• Can the same architecture carry the next channel without another rebuild?

Answering those questions usually removes as much technology as it adds. A shorter supply path with fewer intermediaries, clearer floors and honest measurement frequently outperforms a longer one with more partners.

Architecture first

Treating the ad stack as infrastructure changes the sequence of decisions. The commercial outcome is defined first, the architecture is designed to produce it, and technology is selected to serve that design — not the other way around.

Architecture first. Technology second. Commercial outcome always.